OctoMonic · Sales operations

The Sales Playbook

One intake, two calls, one document, four routes. Everything said out loud on a call, everything written down after it, and the rules that keep the free ten minutes free.

Internal · Prepared for Tom Levi

5-10 min

«שיחת בירור»
the free call, by phone

30 min

«אבחון טכני»
the Technical Diagnostic, on Zoom

₪250

+ VAT, invoiced before the session
refundable up to 24 hours before

routes out of the diagnostic,
plus the honest no and the park

Where the ladder actually stands, 2026-08-22

Two things shipped. One thing still blocks the money.

The design work behind this playbook flagged three blockers. Two are closed. The third is the narrowest point in the whole system and no script fixes it.

Shipped

The names are live and in production. «שיחת בירור» for the free call, «אבחון טכני» / Technical Diagnostic for the paid rung, «דוח האבחון» / The Diagnostic Report for the document. The English free call still carries no product name, on purpose. Treat all of it as settled.

Fixed

The booking validator no longer rejects valid bookings. Name plus one channel is the contract the page prints above the fields, and that is now the contract the service enforces. Verified live. Bookings with an email and no phone go through.

Still blocked

The ₪250 rung cannot take a card. Grow verification is still open, so every diagnostic booking is a request that Tom invoices manually before it confirms. The moment of highest intent, right after the free call, is followed by an invoice and a wait. Some share of that gap is lost revenue.

Backend note: the OctoMonic funnel moved off n8n to a Python service on 2026-08-22. Booking validation, slot rules, confirmation and the Notion write all run there now. Anything in older drafts that describes an n8n workflow as the funnel backend is out of date.

§1 · The ladder

The ladder at a glance

Three rungs, and each one only has to sell the next one. In Hebrew the sequence reads בירור ← אבחון ← בדיקת היתכנות: is there anything here, what is it and what is the right route, then the full blueprint. In English: free conversation, Technical Diagnostic, Viability Assessment.

Step 1 · free

שיחת בירור

No English product name. The CTA reads "Book a free 5-10 min conversation".

5-10 min · phone · confirms instantly

One job: decide whether this person books the diagnostic, or does not. Not selling, not consulting, not scoping.

→

Step 2 · paid

אבחון טכני

Technical Diagnostic

30 min · Zoom · ₪250 + VAT

Invoiced before the session. Cancellation up to 24 hours before is fully refundable. Not publicly bookable: it needs a per-lead token issued at the close of the free call.

→

Step 3 · the deliverable

דוח האבחון

The Diagnostic Report

2-3 pages · next business day, 18:00

The recommended route, why it fits, and what happens next. Theirs to keep, including if someone else builds the work. The report is the ask, not the room.

And then one of four offerings

בדיקת היתכנות

Viability Assessment

₪1,500 to ₪2,500 + VAT

Ends in a build-ready specification. The default destination out of the diagnostic.

הכשרה מקצועית

Professional Training

Priced after a call

Teaches people to operate AI on their own work. The price is set once the group and the format are known.

הקמת סביבת עבודה

Workspace Setup

From ₪2,500 + VAT

One day, one person, their machine and their accounts configured.

ייעוץ וליווי

Consultation & Advising

Scoped per engagement

Technical judgement with no build attached. Someone to argue with before committing.

The rung boundary that protects the ladder

The diagnostic is a read, not a specification. The specification is the Viability Assessment, and the report has to say so in writing every single time. Without that line the ₪250 rung eats the ₪1,500 rung.

Live booking constraint	Value
Days and hours	Sunday to Thursday, 10:00 to 18:00 Israel time. The end of the slot must fall inside the window
Free call grid	10 minutes on a 15-minute grid
Diagnostic grid	30 minutes on a 30-minute grid
Minimum notice	30 minutes
Instant confirm	Free call only. The diagnostic stays a 24-hour hold until Tom acts, because it is invoiced first
Diagnostic access	Requires a per-lead token of 16 characters or more, issued through the step-2 unlock path. Not publicly bookable, so the free-call close must actually send the link
Double bookings	Impossible at the database level. A losing race returns 409 and the page re-offers times
Event log	Every status and slot change is logged automatically. The behavioural half of the learning loop is already being captured

§2 · The free call · «שיחת בירור» · 5-10 minutes

The free call script

The call has exactly one job: decide whether this person books the Technical Diagnostic, or does not. Not selling. Not consulting. Not scoping. Print this section and tape the clock next to the laptop.

Modality

Phone. No screen, no slides. Notes by hand on the capture sheet.

Language

Open in the language they booked in. If they answer in the other one, switch mid-sentence and never comment on it.

Voice

First person. You speak as "I". The website is company voice. A spoken script is not. The written report afterwards is company voice again.

Slots

Sun to Thu, 10:00 to 18:00 Israel time, 10 minutes on a 15-minute grid, 30 minutes notice minimum. Intro slots confirm instantly.

The clock

Time	What happens
0:00 → 0:25	Opening, verbatim. Includes the intake read-back. Sets the whole frame.
0:25 → 1:00	Their one sentence. You say almost nothing.
1:00 → 5:30	The questions, in priority order.
5:30 → 6:30	Read-back, then the one free answer.
6:30 → 7:30	The route: name it out loud, one sentence, with the reason.
7:30 → 8:30	Closing, verbatim. Diagnostic, or the honest no.
8:30 → 9:00	Logistics plus the last question. Hang up.
9:00 → 10:00	Reserve, not content. You take these back to back. This minute stops the day sliding, and it is where the Notion capture happens.

Two hard checkpoints

At 5:30 the questions stop. At 8:30 you are already saying goodbye.

The order, and why it is not the source order

WHAT · NOW · WHY NOW · WHO PAYS · SIZE · STACK

The original design put WHY NOW and WHO PAYS at positions 5 and 6 and marked them "never drop", which is precisely where a real ten-minute call runs out of road. Never-drop questions cannot live at the bottom.

SIZE and STACK moved down because both get re-asked at depth in the paid session anyway, and SIZE usually arrives free. When they say "we do about forty a day" during the walk, that is Q5 answered. Tick it and move on.

Mnemonic: what, now, why now, who pays, then size and stack if there is road left.

When you are behind

At 3:30 and still on NOW: stop the walk mid-sentence with the interrupt line and jump to WHY NOW.

The tired version, for when you are genuinely exhausted or the call is obviously five minutes: WHAT, then WHY NOW, then WHO PAYS, then read-back, then close. Three questions. It still routes correctly most of the time, and it beats a limp six.

Realistically four questions land in ten minutes. Expect the call to run 12 to 14 minutes routinely. Book 15-minute gaps or run the three-question version most days, but do not pretend both.

The router, in one line: listen to the noun in their first sentence

The noun is	Route
a system, a product, a process	Technical Diagnostic, then Viability Assessment
a person, a team, "my staff"	Professional Training, priced after a call
"us", "how we work", "my day"	Workspace Setup, from ₪2,500 + VAT, one day, one person
"whether we should", "which way"	Consultation & Advising
something you would never build	Honest no

The three rules that protect the call

1 One free answer

Exactly one genuinely useful technical answer per call. Real, specific, worth their ten minutes. Then stop. That answer is the brand promise. The second one is unpaid consulting, and it is where the ₪250 close gets harder, because they now have enough to go and try it themselves. The sheet has one line for it. Once it is written, the rule has been used.

2 No quotes

Never price their project on a free call. Ladder prices are public and you can say them out loud. A number for their thing is not public, and inventing one on the phone is the fastest way to be held to it.

3 Technical read only

Feasibility, approach, effort, cost, risk. You do not answer whether people will buy it, whether it is a good business, or whether to raise. Say so out loud when it comes up. The boundary line is written out below.

The opening, verbatim

The most important 25 seconds in the funnel. It buys the right to interrupt, plants the honest no before they need it, and closes the loop on the intake so the tap they made on the booking page was not theatre.

Opening · Hebrew

אהלן [שם], תום מאוקטומוניק. תודה שקבעת.

ראיתי שסימנת ״[the tapped option]״ — נתחיל משם.

השיחה הזאת היא עשר דקות, בלי עלות, ואני שומר על השעון — אז אני אקטע אותך פה ושם, זה לא חוסר עניין.

אני אשאל כמה שאלות קצרות כדי להבין מה יש לך ביד, ובסוף אני אגיד לך מה לדעתי הצעד הנכון — גם אם הוא ״לא לבנות את זה״.

נתחיל?

Opening · English

Hi [Name], Tom from OctoMonic. Thanks for booking.

I saw you picked "[the tapped option]" — let's start there.

This is ten minutes, no charge, and I keep an eye on the clock — so I'll cut in here and there, it isn't rudeness.

I'll ask you a few short questions to understand what you've got, and at the end I'll tell you what I think the right next step is, even if that's "don't build this".

Shall we?

Three variants of line 2. Pick before you dial.

They	Say
tapped an option, wrote nothing	HE: «ראיתי שסימנת ״[option]״ — נתחיל משם.» EN: "I saw you picked '[option]' — let's start there."
tapped and wrote a line	HE: «ראיתי שכתבת ״[their line]״ — נתחיל משם.» EN: "I saw you wrote '[their line]' — let's start there." Quote their words, not the tap. Their sentence always beats your category.
skipped both	Omit line 2 entirely. Do not say "you didn't fill anything in".

Then, immediately, the handover. No pause, no small talk.

במשפט אחד — מה אתה רוצה שיהיה קיים, או מה אתה רוצה להפסיק לעשות ביד?

In one sentence — what do you want to exist, or what do you want to stop doing by hand?

Why the opening is built this way

"I'll cut in here and there" is pre-authorised permission to interrupt, given while they still like you. Without it the ten-minute talker eats the call and you cannot stop them without a cost.

"Even if that's don't build this" is the differentiator, said before they can suspect it is a tactic. It is already the promise printed on the booking page, so you are repeating yourself rather than pitching.

The interrupt line · you will need it two or three times

רגע, אני עוצר אותך שנייה — אני רוצה לוודא שהבנתי לפני שממשיכים.

Let me stop you there a second — I want to be sure I've got that before we move on.

The questions, in order

Where a question has two forms, the fork is idea-stage against operating business. You will know which by the end of Q1.

Q1WhatNever drop

במשפט אחד — מה אתה רוצה שיהיה קיים, או מה אתה רוצה להפסיק לעשות ביד?

In one sentence — what do you want to exist, or what do you want to stop doing by hand?

Tells you which of the four offerings is in play, by the noun rule, and whether they have one problem or a fog.

Write down their exact words. This sentence is the single most valuable thing you capture on this call. It opens the paid session, it opens the report, and it is the field that makes the learning loop worth running at all.

Bad answer sounds like: a three-minute origin story with no object in it · "I want AI in my business" · "it's like [famous app] but for [industry]" with no process underneath · "I can't say until you sign an NDA".

Q2Now

תעביר אותי על מה שקורה היום, שלב-שלב — מי עושה את זה, ואיפה זה יושב?

אם רעיון: מה קיים היום? משהו בנוי, משורטט, כתוב — או שהכול עוד בראש?

Walk me through what happens today, step by step — who does it, and where does it live?

If idea-stage: What exists right now — anything built, drawn, written down? Or is it all still in your head?

Tells you whether there is a real process to automate or a wish. Quietly tells you where the data sits, which is half of feasibility.

Grab SIZE inline here if they volunteer a number. Ask "how many a day is that?" and tick Q5 off without spending a question on it.

Bad answer sounds like: "nothing yet, that's what I need you for", which is fine alone and fatal when WHO PAYS is also empty · an answer entirely in future tense that never says what happens today · "it depends, every time is different" three times over, which is judgement work, and judgement work does not automate cheaply.

Q3Why nowNever drop

למה עכשיו? מה השתנה — ומה קורה אם זה נשאר ככה עוד שנה?

Why now — what changed? And what happens if it stays exactly as it is for another year?

Tells you whether this is a project or a hobby. A real trigger has a date on it: a new hire, a lost order, a season, a system being retired, growth that broke something.

Both halves are asked here and never again. The paid session confirms it in one line and moves on.

Bad answer sounds like: "no particular reason, I've just been thinking about it" · "I'm exploring options" · anything where the honest answer to part two is that nothing happens. Nothing happening is the strongest single predictor of a no-show at the paid session.

Q4Who paysNever drop

מי עוד צריך להגיד כן חוץ ממך — ויש לך מספר בראש, כמה שווה לך להשקיע בזה?

Who else needs to say yes besides you — and have you got a number in your head for what this is worth spending?

Tells you authority and budget reality in one breath, which is less crass than asking about money alone and takes half the time.

If they name any range at all, write their exact words.

Bad answer sounds like: "let's not talk about money yet" · "depends what you quote me" · a number an order of magnitude below the ladder. Someone imagining ₪800 for a build is not being difficult, they are at the wrong stage, and that is a park rather than a rejection · "I need to ask my partner", where the partner has never heard of this.

Q5SizeFirst to drop

כמה פעמים בשבוע זה קורה, וכמה זמן זה לוקח בכל פעם?

אם רעיון: מי בצד השני, וכמה כאלה יש?

How many times a week does that happen, and how long does it take each time?

If idea-stage: Who's on the other end, and how many of them are there?

Tells you the arithmetic behind the honest no. Frequency multiplied by duration is what a build has to beat.

Often already answered inside Q2. For idea-stage it usually produces nothing usable, which is fine and is why it ranks fifth. It does mean track A routes on Q3 and Q4 alone.

Bad answer sounds like: "a few times", with no number and no willingness to estimate one · numbers that inflate on the spot · "everyone", on the idea route.

Q6StackSecond to drop

באילו מערכות הכול יושב היום — ויש משהו שכבר מדבר עם משהו אחר?

What does it all live in today — and does anything already talk to anything else?

Tells you feasibility and integration risk, usually the single biggest cost driver, and the thing you read faster than anyone else they will speak to.

Bad answer sounds like: a closed vendor system with no API and no export · "it's on paper and that's not changing" · "I don't know" from the person who is supposedly the owner. That last one is an authority signal, not a technical one. Mark it against WHO PAYS.

Q7Prior attemptsOnly if something still does not add up

ניסית כבר לפתור את זה פעם? מה קרה?

Have you tried to fix this before? What happened?

Tells you about a burned vendor, an abandoned internal attempt, or quiet resistance inside the company. All three change how the paid session runs.

Bad answer sounds like: three failed attempts with three different people, all of whom were "the problem".

The read-back · 5:30 to 6:30 · this is what makes ten free minutes feel worth it

רגע, תן לי לוודא שהבנתי. [שתי שורות — מה הבעיה, איפה היא יושבת, ומה תלוי בה]. זה מדויק?

Let me make sure I've got it. [Two lines — what the problem is, where it sits, what depends on it]. Is that accurate?

Then, once, the one free answer: the most useful technical thing you can say in thirty seconds. The hard part is stopping after it.

The routing decision guide

Read down. First row that matches wins.

What you heard	Route	Say it as
Clear thing to build, a real trigger at Q3, and Q4 answered with a range or ₪250 shrugged off	Technical Diagnostic. The default	"There's something here, and I'm not going to guess at the shape in the minutes we've got left."
The same, plus a written brief already done and a build-range budget named	Diagnostic anyway, unless Tom has set a standing override	Do not improvise a skip on the phone. See the open decisions list.
No single process pain. The ask is how do we actually use this day to day. One person, hands-on	Workspace Setup, from ₪2,500 + VAT, one day, one person	"What you're describing isn't a build, it's a setup day."
The object is people learning. A team, a group, recurring sessions	Professional Training, priced after a call	"That's training, not a project. The price gets set once I know the group and the format."
They already have technical people. They want judgement, a second opinion, ongoing input, not a build	Consultation & Advising	"You don't need me to build it. You need someone to argue with before you commit."
Q5 multiplied by Q6 fails. Small, rare, or locked in a system that makes it expensive	Honest no (i), the value math	Verbatim below
An off-the-shelf tool already does it	Honest no (ii), it already exists	Verbatim below
Buildable, but not your lane	Honest no (iii), not us	Verbatim below
Q3 empty and Q4 empty. No trigger, no number, no authority	Park. This is not now, not never	Verbatim below. A park is not a loss and must never be counted as one

Close A · the diagnostic · 7:30 to 8:30 · the default

The frame that stops it sounding like an upsell: it is not an extra step invented for them. It is the technical read that normally only happens once a ₪5,000-plus build is already under way, pulled out and priced on its own so the answer arrives before the commitment does. That is the site's published positioning, so you are repeating yourself rather than pitching.

Close A · Hebrew

אז ככה אני שומע את זה: [שורה אחת].

יש פה משהו אמיתי, ואני לא הולך לנחש את הצורה בדקות שנשארו לנו.

הצעד הבא אצלי הוא אבחון טכני — פגישת עבודה של שלושים דקות בזום. היא עולה 250 שקל פלוס מע״מ. זה בדיוק האבחון שנעשה בדרך כלל רק בתוך פרויקט של חמשת אלפים שקל ומעלה, ואצלי הוא עומד בפני עצמו, כדי שהתשובה תגיע אליך לפני ההתחייבות.

חשבונית לפני הפגישה, וביטול עד 24 שעות לפני מזכה בהחזר מלא.

בפגישה בוחרים כיוון, ואחריה אתה מקבל את דוח האבחון — עד סוף יום העסקים הבא: מה הכיוון שאני ממליץ עליו, למה הוא ולא האחרים, מה בגדול צריך כדי לעשות אותו, וכמה זה עולה. הדוח שלך בכל מקרה — גם אם בסוף מישהו אחר בונה.

שאשלח לך קישור וחשבונית?

Close A · English

So here's how I'm hearing it: [one line].

There's something real here, and I'm not going to guess at the shape in the minutes we have left.

My next step is a Technical Diagnostic — a thirty-minute working session on Zoom. It's ₪250 plus VAT. This is the read that normally only happens once a ₪5,000-plus build is already under way; here it stands on its own, so the answer reaches you before the commitment does.

Invoice before the session, and cancellation up to 24 hours before is fully refundable.

In the session we pick a direction, and afterwards you get the Diagnostic Report — by end of the next business day: the route I recommend, why that one and not the others, roughly what it takes, and what it costs. The report is yours either way, including if someone else ends up building it.

Shall I send you a link and an invoice?

Remember the rail

The diagnostic is not publicly bookable and does not auto-confirm. If you say "I'll send you a link", the link has to actually go out, with the invoice, while the intent is still warm. Until Grow verification clears, this is the manual step that decides whether the close converts.

Close B · the honest no, in four flavours

Honest no · all four · Hebrew and English

(i) The value math

אני לא חושב שכדאי לך לבנות את זה, ואני אגיד לך למה: [frequency × duration]. עלות הפיתוח לא תחזיר את עצמה, ולא נעים לי למכור לך אבחון לדבר שאני כבר יודע את התשובה עליו.

I don't think you should build this, and here's why: [frequency × duration]. The build cost doesn't earn itself back, and I'm not comfortable selling you a diagnostic for something I already know the answer to.

(ii) It already exists

יש כלי מדף שעושה בדיוק את זה. תבדוק אותו קודם. אם אחרי חודש משהו שם לא מסתדר — תחזור, ואז יש על מה לדבר.

There's an off-the-shelf tool that does exactly this. Try it first. If something there doesn't fit after a month, come back — then there's something to talk about.

(iii) Not us

אפשר לבנות את זה, אבל לא אצלי — זה לא התחום שאני חזק בו, ואתה תשלם על העקומת למידה שלי.

This can be built, just not by me — it isn't where I'm strong, and you'd be paying for my learning curve.

(iv) The park

אני חושב שעדיין מוקדם. אין טריגר ואין תקציב בראש, והאבחון יהיה מעניין אבל לא ישנה לך כלום כרגע. כשמשהו יזוז — תחזור, השיחה הזאת עדיין תהיה בחינם.

I think it's early. There's no trigger and no number yet, so a diagnostic would be interesting and change nothing. When something moves, come back — this call is still free then.

The awkward cases

"So do you think it'll work?" · the boundary line

אני נותן קריאה טכנית — האם אפשר לבנות, איך, כמה עבודה, כמה זה עולה, ומה הסיכונים. אם יש לזה שוק ואם כדאי לשים על זה כסף — זה לא התחום שלי, ואני לא אמכור לך תשובה שאין לי.

What I give is a technical read — can it be built, how, how much work, what it costs, what the risks are. Whether there's a market for it, or whether it's a good place to put your money — that's not my field, and I'm not going to sell you an answer I don't have.

The last question, always, at 8:30

שאלה אחרונה — איך הגעת אליי?

Last thing — how did you find me?

Not worth the time at all

End warmly at 6:00, make no offer, mark it, and decline a second free booking: students and researchers, anyone who names a competitor and asks how you do things, anyone who wanted the free answer and now has it, and anyone with no authority and no route to whoever has it.

Two things that will not survive contact with a real call

Six questions do not fit in ten minutes. At roughly 40 seconds each including the answer, six questions is four minutes of pure question and answer with no reaction, no rapport and no read-back. Israeli business owners do not answer "walk me through what happens today" in 40 seconds. Four questions land. That is exactly why the two never-drop questions moved to positions 3 and 4.

The one-free-answer rule is right and nearly impossible to hold. You will give three. The mitigation is the sheet: one line, and once it is written the rule has been used.

§3 · «אבחון טכני» · Technical Diagnostic · 30 minutes · ₪250 + VAT

The diagnostic script

A working session, not a sales call. Thirty minutes that end in a named route with a published price attached, and a written report the next business day. Hebrew below is written masculine singular, spoken register. Flip endings for a woman and to plural if two people join.

The script never says the product name out loud

Tom says "the thirty minutes" and «שלושים הדקות». That is deliberate. A future rename does not break the script.

Must be true or the call does not happen

Invoice paid.
Zoom link sent, in Israel time with the zone label, IDT or IST.
Track known: A for idea-stage or B for operating business.
The verbatim WHAT sentence from the free call, on the sheet.

If the last two are missing you start four minutes behind and will not reach the diagnosis. In that case run the compressed variant: skip the outcome block and open straight into the walk.

Five-minute prep, same morning

Their site or Instagram. What they actually sell, roughly how many people, whether they take orders online.
Search the main tool they named plus the word "API". Know before the call whether it has one. This one search is often the whole ₪250 of value.
Write the one-line problem statement at the top of the sheet. You read it back at minute 2.

Minute by minute

Time	Block	What has to happen
0:00 → 0:02	Opening	Frame reset, note-taking consent, scope boundary, what they walk out with and when
0:02 → 0:04	Read-back	Their problem in one sentence, in their words. They correct it. This is the call's anchor
0:04 → 0:07	The outcome	The change, and the measure. Two questions, neither asked on the free call
0:07 → 0:15	The deep walk	The longest block. This is what they paid for
0:15 → 0:21	Feasibility	Access, data, integration, volume confirm, compliance if relevant
0:21 → 0:23	Constraints	Who decides, what drives the timeline, who runs it after
0:23 → 0:24	The pause	"Give me half a minute." Silence. Fill the scorecard
0:24 → 0:28	The diagnosis	Five sentences, out loud, including the price of the route
0:28 → 0:30	The close	Report, when, next step. No decision asked for today

Gate 1 · 21:00

You stop asking. Anything unknown goes into the report as a named unknown. Asking a nineteenth question is how the value gets stolen from the part they paid for.

Gate 2 · 23:00

You are talking, not listening. The scorecard is filled and the diagnosis is being assembled.

Gate 3 · 30:00

You are done. Never more than three minutes over.

What to cut when it overruns

If at	Cut
9:00 and still on the outcome	Drop the measure. Ask it in the report's follow-up line instead
15:00 and the walk is not mapped	Stop it. Ask only "which step in that hurts most?" and go straight to feasibility
19:00 and feasibility is thin	Ask only Access and Data. Compliance only if they handle health, credit or ID data
21:00 and constraints unasked	Ask one: "who besides you has to say yes?" Timeline and budget go in the report as open items
27:00 and no route named	Say the route and the price. Skip why not the others, which belongs in the report anyway

Squeeze the walk, never feasibility

With a talkative operator the deep walk will overrun and feasibility will get squeezed to two questions. That is the wrong squeeze. Feasibility is the part they cannot get anywhere else. If the walk is not done at 15:00, cut it with the harder interrupt line and lose the detail, not the access question.

If you are still deep in questions at 30:00, that is itself the finding. Say it.

"We're at time and we're not finished, and that's the answer to your question. This is bigger than thirty minutes, which is exactly what the assessment format exists for. I'll put the specifics in the report."

״נגמר הזמן ולא סיימנו, וזו בעצם התשובה. מה שיש כאן גדול מדי לשלושים דקות, ובדיוק בשביל זה קיים הפורמט הארוך. את הפרטים אני שם לך בדוח.״

The opening, verbatim · two minutes

Almost the same every time. The repetition is what makes it smooth.

Diagnostic opening · English

"Thanks for making the time. Before we get into it, one minute on how this works.

This is not a sales call, it's a working session. I'm going to ask a lot of questions, some of them technical, and I'll interrupt you sometimes so we cover everything. Don't take it personally, it's the clock.

At the end of the thirty minutes I'll tell you what I think should happen first, roughly what it takes, and what could go wrong. If my read is that you shouldn't build this, you'll hear that too, in the same words.

Afterwards you get the report: what I understood, what I recommend, why that one and not the others, and what it costs. It'll be with you by end of the next business day. The report is yours either way — if you take it to another developer, it's written so they can work from it.

One quick thing: I take notes as we go so the report comes out accurate, and they stay with me. Alright?

And so we don't waste the thirty minutes: what I give you is a technical read. Can this be built, how, how much work, what it costs, what the risks are. I'm not going to tell you whether there's a market for it or whether it's a good place to put your money. That's not my field and I'm not going to sell you an answer I don't have.

Alright. Start here: what do you need to happen?"

Diagnostic opening · Hebrew

״תודה שהתפנית. לפני שנכנס לעניין, דקה על איך זה עובד.

זו לא שיחת מכירה, זו פגישת עבודה. אני הולך לשאול הרבה שאלות, חלק מהן טכניות, ולפעמים גם אקטע אותך כדי שנספיק הכול. אל תיקח את זה אישית, זה השעון.

בסוף שלושים הדקות אני אגיד לך מה לדעתי צריך לקרות ראשון, כמה עבודה זה בערך, ומה יכול להשתבש. אם המסקנה שלי היא שלא כדאי לבנות את זה, גם את זה תשמע, באותן מילים בדיוק.

אחרי השיחה אתה מקבל את הדוח: מה הבנתי, מה ההמלצה, למה דווקא היא ולא האחרות, וכמה זה עולה. הוא אצלך עד סוף יום העסקים הבא. הדוח שלך בכל מקרה — וגם אם תיקח אותו למפתח אחר, הוא כתוב כך שאפשר לעבוד ממנו.

דבר קטן אחד: אני רושם לעצמי תוך כדי כדי שהדוח ייצא מדויק, וההערות נשארות אצלי. בסדר?

ודבר אחרון, כדי שלא נבזבז את שלושים הדקות: מה שאני נותן כאן זה אבחון טכני. האם אפשר לבנות את זה, איך, כמה עבודה, כמה זה עולה, ומה הסיכונים. אני לא אגיד לך אם יש לזה שוק ואם כדאי לשים על זה כסף. זה לא התחום שלי, ואני לא אמכור לך תשובה שאין לי.

יאללה. תתחיל מזה: מה אתה צריך שיקרה?״

Note-taking, not recording

The script asks consent for notes, which is true, needs no policy, and gets the same permission to pause and type. Recording consent copy is deliberately unshipped. Zoom's free tier can record locally, so recording is possible, but it would need a consent line, a storage answer and a privacy-policy sentence. Default: stay on notes until Tom decides otherwise.

The read-back · 0:02 to 0:04 · a hard gate, not a courtesy

Say their problem back in the words they used on the free call, then stop talking

"Before I ask anything, let me say back what I think we're here for, and you correct me. You said [one sentence]. Right, or did I get it sideways?"

״לפני שאני שואל משהו, אני אגיד לך מה הבנתי שאנחנו כאן בשבילו, ואתה תתקן אותי. אמרת ש[משפט אחד]. נכון, או שלא הבנתי?״

Their correction is the single most valuable sentence in the call

Write it verbatim in the QUOTES slot. It goes into the report's opening section word for word. That is the moment the client recognises the document as being about them. Without it the report is about the thing they asked for rather than the thing they actually have, and they do not recognise themselves in it.

The interrupt line · you will need it three or four times · warmly, and do not apologise twice

"Let me stop you there for a second, I want to make sure I have that one right before we move on."

Harder: "I'm cutting in, only because the clock's running and there's a part I have to cover."

״רגע, אני עוצר אותך פה שנייה, אני רוצה לוודא שהבנתי את זה נכון לפני שממשיכים.״

״אני קוטע אותך, רק כי השעון רץ ויש חלק שאני חייב לכסות.״

The question set, in four layers

★ means ask every call, no exceptions. There are eight of them. If everything falls apart, ask only the eight. Nothing here re-asks the free call. Where the free call already answered something, the paid session confirms in one line and goes deeper, and says so out loud, or it reads as repetition.

Layer 1 · The outcome · 0:04 to 0:07

Neither question was asked on the free call. They arrive holding a solution, and this block takes the solution out of their hands without insulting them.

A★ The change

אם זה עובד כמו שאתה רוצה, מה שונה בעסק בעוד שלושה חודשים?

If this works the way you want, what's different in the business three months from now?

Establishes the business outcome behind the feature request. Everything downstream gets judged against this sentence.

Bad answer: they describe the software again. Push once, with "that's the how. What changes for you?" or ״זה ה'איך'. מה משתנה אצלך?״

B★ The measure

איך תדע שזה עבד? על מה תסתכל?

How would you know it worked? What would you be looking at?

Establishes whether success is measurable at all. If they cannot name anything they would look at, the project has no finish line, and the report has to say so.

Layer 2 · The deep walk · 0:07 to 0:15

Open the block with this frame, or it reads as repetition · required line

"I asked you a version of this on the phone and I'm going to ask it again properly — this time I want the boring parts, because that's where the work hides."

״שאלתי אותך גרסה של זה בטלפון, ועכשיו אני שואל את זה ברצינות — הפעם אני רוצה גם את החלקים המשעממים, כי שם מסתתרת העבודה.״

C★ The walk, or what existsTrack B / Track A

HE · B

קח אותי על זה מההתחלה. משהו מגיע, ואז מה. שלב שלב, גם החלקים המשעממים.

A: מה קיים היום? מסמך, מסכים, קוד, חשבון איפשהו — או שזה עדיין בראש?

EN · B

Walk me through it from the beginning. Something arrives, and then what. Step by step, including the boring parts.

A: What exists today? A document, screens, code, an account somewhere — or is it still in your head?

While they talk, mark every step where a human retypes, copies, or checks something with M. The count of Ms is the headline of the report.

On track A, "in my head" is common and is not a problem. Pretending otherwise is.

DWhere it breaks, or the one jobTrack B / Track A

HE · B

כשזה מתפקשש, מה בדרך כלל התפקשש?

A: אם הגרסה הראשונה תדע לעשות רק דבר אחד, מה הדבר האחד שהיא חייבת לעשות כדי שיהיה לזה טעם?

EN · B

When it goes wrong, what usually went wrong?

A: If the first version could only do one thing, what's the one thing it has to do to be worth existing?

On track B, the thing worth fixing is often not the thing they asked about. This question earns the ₪250 more often than any other.

On track A, this answer usually becomes the headline of the report.

EWho touches it

מי עוד נוגע בזה חוץ ממך?

Who else has their hands on this besides you?

How many people have to change habits, which is the hidden cost of every build, and who will fight it.

Layer 3 · Feasibility · 0:15 to 0:21 · the part nobody else asks

F★ AccessBiggest single cost driver

על שם מי המערכת הזאת, הרישיון שלכם כולל גישת API, ואפשר לקבל חשבון בדיקה עוד לפני שמתמחרים משהו?

Whose account is that system under, does your licence include API access, and could I get a test account before anything is quoted?

Three failure modes in one question: no API on that licence tier, a reseller who owns the account, or an IT gatekeeper who has not been told. Any one of them changes the number, and it is almost never asked before a quote.

G★ Where the data lives

HE · B

איפה המידע הזה יושב בפועל? אקסל, מערכת, וואטסאפ של מישהו, מחברת?

A: כדי שזה יעבוד, איזה מידע הוא צריך, ולמי יש אותו היום?

EN · B

Where does that information actually sit — an Excel file, a system, someone's WhatsApp, a notebook?

A: For that to work, what information does it need, and who has it today?

"It's in my head" and "it's in the WhatsApp group" are real answers, and both point at Workspace Setup rather than a build.

For idea-stage, data that must be scraped from someone who will block it, or that does not exist yet, is the number-one killer.

HVolume: confirm, do not re-ask

אמרת [their number] בטלפון — זה שבוע רגיל או שבוע טוב?

You said [their number] on the phone — is that a typical week or a good one?

Write down their number exactly as they say it. Never round it up in the report.

IComplianceOnly if health, card or ID data

יש כאן נגיעה במידע רפואי, תעודות זהות או כרטיסי אשראי?

Does any of this touch health records, ID numbers or card data?

Cheap to design in, expensive to retrofit. If the answer is yes, it belongs in the risks section of the report whether or not they raised it.

Layer 4 · Constraints · 0:21 to 0:23 · escalate, do not repeat

J★ Who decides

הזכרת את [name] בטלפון. אם נגיע לבנייה — זו שיחה או חתימה?

You mentioned [name / 'my partner'] on the phone. If we get to a build, is that a conversation or a signature?

This is also the person the report has to survive being forwarded to. Write the name down.

K★ Who runs it after

נניח שזה נבנה. מי מריץ את זה ביומיום, ומי עונה לטלפון כשזה נופל באחת עשרה בלילה?

Say it gets built. Who runs it day to day, and who answers the phone when it breaks at 11pm?

Establishes whether they understand that software is a running cost rather than a purchase. Also seeds the retainer conversation honestly.

L★ What drives the timeline

יש תאריך שזה חייב לעבוד עד אליו, ומה קובע את התאריך הזה?

Is there a date this has to work by, and what makes it that date?

A date with no driver behind it is a wish. A date with a driver is a budget.

MThe range checkOnly if the free call got no number

המסלולים מכאן נעים בערך מ-1,500 שקל לבדיקת ההיתכנות ועד חמש ספרות לבנייה. לאיזה קצה כיוונת?

The routes from here run from about ₪1,500 for the assessment to five figures for a build. Which end of that were you picturing?

Never quote their project. You are checking a range against published prices, which is not a quote.

NPrior attemptsOnly if the free call did not ask it

ניסית לסדר את זה פעם? מה קרה עם זה?

Did you try to fix this before? What happened to that?

Changes the tone of the report more than its content. A burned buyer needs the unknowns section to be unusually explicit.

The pause · 0:23 to 0:24

Say it out loud, then genuinely go quiet and fill the scorecard

"Give me half a minute — I want to put this together properly rather than think out loud at you."

״תן לי חצי דקה — אני רוצה לחבר את זה כמו שצריך, ולא לחשוב לך בקול.״

The silence is not awkward. It is the most expensive-looking thing that happens in the thirty minutes.

The diagnosis · 0:24 to 0:28 · five sentences, in this order, out loud

#	Sentence	How it lands
1	What this actually is	"What you have here isn't a [what they called it], it's a [what it is]."
2	The route, named, with its published price	"The route I'd recommend is [offering], which is [price] + VAT."
3	Why that one	One reason, tied to something they said in the last half hour.
4	Why not the obvious alternative	One sentence. This is the sentence that buys trust.
5	The one thing that could change the answer	The biggest named unknown, which becomes item 1 of the report's unknowns.

Then stop. Do not add a sixth.

The close · 0:28 to 0:30

The close · English and Hebrew

"That's the read. It's all going in the report — with the reasoning, the unknowns and the next steps — and it'll be with you by end of the next business day. I'm not asking you to decide anything today; read it, and if it makes sense, reply and we'll book the next piece. If you'd rather take it to someone else to build, that's what it's written for."

״זה האבחון. הכול נכנס לדוח — עם הנימוקים, מה שעוד לא ידוע, והצעדים הבאים — והוא אצלך עד סוף יום העסקים הבא. אני לא מבקש ממך להחליט משהו היום; תקרא, ואם זה מסתדר — תענה לי ונקבע את השלב הבא. ואם תעדיף לקחת את זה למישהו אחר שיבנה, בשביל זה הוא כתוב.״

Never ask for the sale in the room

The report is the ask. It arrives when they are alone, unpressured, and can forward it to whoever else has to say yes, which is exactly the person question J identified.

§4 · «דוח האבחון» · The Diagnostic Report

The report

The deliverable the whole ladder converges on. The scripts are spoken and first person. The report is a written company document, so the voice changes: write "the recommendation is", never "I recommend", and «ההמלצה היא», never «אני ממליץ».

Length

Two to three pages. Longer is not more thorough, it is less read.

Turnaround

By 18:00 the next business day, Israel time. Promised in the opening and printed on the booking page, so breaking it is a public failure.

Delivery

A branded mockup URL on an octomonic.com subdomain, carrying the OctoMonic and client logo lockup top-centre. A PDF export goes with it.

What it is not

Not a specification. The specification is the Viability Assessment, and section 9 has to say so explicitly every time.

The structure: nine sections, fixed order

#	Section	HE	Content	Length
1	Header	—	Name, date in Israel time with the zone label, "Technical Diagnostic, 30 minutes", one line on what this is and what it is not	3 lines
2	What we understood	מה הבנו	Their words, including the correction from the read-back, quoted	1 short para plus up to 3 bullets
3	The recommendation	ההמלצה	One line. The named offering, its published price, and + VAT	1 line, boxed
4	Why this route	למה המסלול הזה	Three bullets, each tied to something they said	3 bullets
5	Why not the others	למה לא האחרים	One line each for the three not chosen. Never skip this.	3 lines
6	How it would work	איך זה עובד	The shape in plain language: what exists, what gets built, where the data moves, where humans stay in the loop	4 to 8 lines or a small diagram
7	What is still unknown	מה עדיין לא ידוע	Named unknowns plus what would answer each. This is what makes it honest and what sells the next rung	2 to 4 items
8	Risks	סיכונים	Two to four things that could go wrong, in their language rather than yours	2 to 4 lines
9	Next steps	הצעדים הבאים	Numbered, price and + VAT on each, plus the take-this-elsewhere line and the scope boundary verbatim	3 to 5 lines

Hard rules inside the report

Every number is either their number, quoted as they said it and never rounded up, or a published ladder price. Nothing else. No invented hours, no invented savings, no "you'll save 40%".
Every price shows + VAT.
Section 5 is non-negotiable. A recommendation without three rejections is a pitch.

Section 7 is non-negotiable. A document with no unknowns is a document that was written before the call.
No client is ever named. Not theirs, not anyone else's.
If the recommendation is "don't build this", the report is shorter, not apologetic, and sections 3 to 5 become what we would not do and why.

Budget the writing time honestly

The report will not take 45 minutes at first. Budget 90 minutes for reports 1 to 3 and put the block in the calendar the same day as the call, or the next-business-day promise breaks on the first busy week. The Hebrew report is real writing, not translation: both worked examples below were written in Hebrew rather than translated, and it shows in the register.

The standing footer, verbatim, on every report

Standing footer · English

This is a technical read — feasibility, approach, effort, cost and risk. It is not market validation, investment advice or business strategy. It is not a specification: a build-ready specification is the Viability Assessment (₪1,500–2,500 + VAT). This report is yours to keep, including if the work is built by someone else. Prices are valid for 30 days and become binding only in a signed quote or invoice. OctoMonic is operated by Tom Levi, Osek Murshe, Tax ID 302148911.

Standing footer · Hebrew

זהו אבחון טכני — היתכנות, גישה, מאמץ, עלות וסיכונים. הוא אינו בדיקת שוק, אינו ייעוץ השקעות ואינו ייעוץ עסקי. הוא גם אינו מפרט: מפרט מוכן לבנייה הוא בדיקת ההיתכנות (₪1,500–2,500 + מע״מ). הדוח נשאר שלכם, גם אם העבודה תיבנה על ידי מישהו אחר. המחירים תקפים ל-30 יום ומחייבים רק בהצעת מחיר חתומה או בחשבונית. OctoMonic מופעלת על ידי תום לוי, עוסק מורשה, מספר עוסק 302148911.

The full worked example, both languages

Two fictional clients, one on each track, each report written out in full so the shape is copyable. Every person, company and detail below is invented. Open the block you need. All four expand automatically when the page is printed.

▸ Worked example 1 · idea-stage · English Fictional

Fictional. Invented person, invented details, written to show the shape. Not a real client.

דוח האבחון · Diagnostic Report

Noa Bar-On · 21 August 2026 · Technical Diagnostic, 30 minutes

A technical read on what you described. Not a specification, and not a view on whether the market wants it.

1 · What we understood

You want an app where dog owners find and book sitters, take payment through the app, and both sides leave a rating. In your words on the call: "I want it to feel like ordering a taxi, not like posting on a Facebook group."

When we said it back, you corrected one thing, and it turned out to be the important thing: "actually the hard part isn't the booking, it's that I need to know the sitter isn't a lunatic."

Today nothing is built. There is a name, a logo a friend made, and a list of about fifteen sitters you know personally.

2 · The recommendation

Viability Assessment — ₪1,500–2,500 + VAT.

3 · Why this route

The product you described is not one build, it is three, and they are wildly unequal. Listings and booking are small and well understood. Payments — specifically holding a customer's money and paying out to a private individual in Israel — is a regulated, provider-dependent piece that determines both the cost and the timeline. Identity and trust is a third. Thirty minutes was enough to see the split; it was not enough to size the second one, and the second one is the whole number.
You said you would rather know the real cost before committing than start and discover it. The assessment exists precisely for that: it ends in a build-ready specification with architecture, integrations, timeline and budget, and it costs a fraction of discovering the same thing by building.
The specification is yours to keep. If you take it to another developer, or to a partner, or to your bank, it is written to be worked from.

4 · Why not the other routes

Straight to a build. You would be paying a developer to discover the specification while writing code, which is the most expensive order to do those two things in.
Professional Training teaches people to operate AI on their own work. It does not build a marketplace.
Workspace Setup configures AI on one person's machine and accounts for one day. Not what you need.
Consultation & Advising is technical judgement without a build attached. You do want a build; you want to know what it costs first.

5 · How it would work

In shape, not in detail:

A listings and booking core. Sitter profiles, availability, a booking with a state (requested → accepted → done). Ordinary, well-trodden, the cheapest part.
A payments layer. The customer pays at booking; the money is held; the sitter is paid out after the booking completes; a fee is taken in between. In Israel, holding and disbursing money to private individuals is the constrained part — it depends on which provider will onboard the business and on what terms, and that answer is a fact to be established, not designed around.
A trust layer. This is the thing you corrected us on. Ratings alone do not solve it at the beginning, because a new marketplace has no ratings. What actually solves it early is a manual step — someone verifies each sitter before they appear — and manual is fine at fifteen sitters and impossible at fifteen hundred. That transition is a design decision, and it should be made deliberately in the specification rather than discovered later.
What runs it. A phone-first web app is enough for a first version. Native apps double the surface and the app-store overhead and buy nothing you named.

6 · What is still unknown

Unknown	What would answer it
Which payment provider will onboard this model, and what they charge to hold and pay out to private individuals	Direct enquiries to two or three providers with the actual model described. Part of the assessment
Whether sitters need identity verification to be listed, and what that costs per sitter	A short review of what comparable services do plus provider pricing. Part of the assessment
Whether your fifteen sitters would list on day one, and under what terms	A conversation you can have this week without paying anyone. It changes what the first version must do
Whether the first version needs ratings at all	Design decision, resolved in the assessment

7 · Risks

The payments piece is the schedule. If a provider declines the model, the timeline moves regardless of how the rest is built. It should be settled before any code is written.
The empty-marketplace problem is a data problem before it is a software problem. A first version with no sitters in it cannot be shown to anyone. The build has to be planned around getting real sitters in, not around launch day.
Regulated data. If identity documents are collected, they are held to a different standard than a phone number. Cheap to design in, expensive to retrofit.
Scope drift. Every item in this document is buildable. All of them together, in a first version, is how idea-stage builds run past their budget.

8 · Next steps

Viability Assessment — ₪1,500–2,500 + VAT. The idea mapped end to end into a build-ready specification: architecture, integrations, data flows, timeline and budget. The exact figure inside that range is set by how much of the payments question has to be resolved during it. Reply to this document and it is scheduled.
Before that, and free: ask three of your fifteen sitters whether they would list, and what cut they would accept. The answer changes what the first version has to be, and no one needs to be paid for you to find it out.
A build is quoted from the finished specification and decided separately. Nothing here commits you to one.

▸ Worked example 1 · idea-stage · Hebrew Fictional

בדוי. אדם בדוי, פרטים בדויים, נכתב כדי להראות את הצורה. לא לקוח אמיתי.

דוח האבחון

נועה בר-און · 21 באוגוסט 2026 · אבחון טכני, 30 דקות

קריאה טכנית על מה שתיארת. לא מפרט, ולא עמדה בשאלה אם השוק רוצה את זה.

1 · מה הבנו

אתם רוצים אפליקציה שבה בעלי כלבים מוצאים וקובעים שמרטפים, משלמים דרך האפליקציה, ושני הצדדים מדרגים זה את זה. במילים שלכם בשיחה: ״אני רוצה שזה ירגיש כמו להזמין מונית, לא כמו לפרסם בקבוצת פייסבוק.״

כשחזרנו על זה, תיקנתם דבר אחד — והתיקון הזה התברר כעיקר: ״האמת שהחלק הקשה הוא לא ההזמנה, אלא שאני צריכה לדעת שהשמרטף הוא לא מטורף.״

היום לא בנוי כלום. יש שם, יש לוגו שחבר הכין, ויש רשימה של כחמישה־עשר שמרטפים שאתם מכירים אישית.

2 · ההמלצה

בדיקת היתכנות — ₪1,500–2,500 + מע״מ.

3 · למה המסלול הזה

המוצר שתיארתם הוא לא בנייה אחת אלא שלוש, והן לא באותו סדר גודל בכלל. רשימות והזמנות זה חלק קטן ומוכר. תשלומים — ובמיוחד להחזיק כסף של לקוח ולהעביר אותו לאדם פרטי בישראל — זה חלק מפוקח שתלוי בספק, והוא זה שקובע גם את העלות וגם את לוח הזמנים. אמון וזיהוי זה חלק שלישי. שלושים דקות הספיקו כדי לראות את הפיצול הזה; הן לא הספיקו כדי לתמחר את החלק השני, והחלק השני הוא כל המספר.
אמרתם שאתם מעדיפים לדעת את העלות האמיתית לפני שמתחייבים, במקום להתחיל ולגלות אותה תוך כדי. בדיוק בשביל זה קיימת בדיקת ההיתכנות: היא מסתיימת במפרט מוכן לבנייה — ארכיטקטורה, אינטגרציות, לוח זמנים ותקציב — ועולה חלק קטן ממה שעולה לגלות את אותו דבר תוך כדי בנייה.
המפרט נשאר שלכם. אם תיקחו אותו למפתח אחר, לשותף או לבנק — הוא כתוב כך שאפשר לעבוד ממנו.

4 · למה לא המסלולים האחרים

ישר לבנייה. תשלמו למפתח שיגלה את המפרט תוך כדי כתיבת קוד, וזה הסדר היקר ביותר לעשות בו את שני הדברים האלה.
הכשרה מקצועית מלמדת אנשים להפעיל AI על העבודה שלהם. היא לא בונה מרקטפלייס.
הקמת סביבת עבודה מגדירה AI על המחשב והחשבונות של אדם אחד, ליום אחד. לא מה שאתם צריכים.
ייעוץ וליווי הוא שיקול דעת טכני בלי בנייה. אתם כן רוצים בנייה — אתם רק רוצים לדעת קודם כמה היא עולה.

5 · איך זה עובד

בצורה, לא בפרטים:

ליבה של רשימות והזמנות. פרופילי שמרטפים, זמינות, והזמנה עם מצב (התבקשה ← אושרה ← בוצעה). רגיל, שחוק, החלק הזול ביותר.
שכבת תשלומים. הלקוח משלם בהזמנה, הכסף מוחזק, השמרטף מקבל תשלום אחרי שההזמנה הסתיימה, ובאמצע נלקחת עמלה. בישראל, החזקת כסף והעברתו לאנשים פרטיים היא החלק המוגבל — הוא תלוי באיזה ספק יסכים לקלוט את העסק ובאילו תנאים, וזו עובדה שצריך לברר, לא משהו שאפשר לתכנן סביבו.
שכבת אמון. זה הדבר שתיקנתם אותנו עליו. דירוגים לבדם לא פותרים את זה בהתחלה, כי במרקטפלייס חדש אין דירוגים. מה שכן פותר את זה בשלב הראשון הוא שלב ידני — מישהו מאמת כל שמרטף לפני שהוא מופיע — וידני זה בסדר גמור בחמישה־עשר שמרטפים ובלתי אפשרי באלף וחמש מאות. המעבר הזה הוא החלטת תכנון, וכדאי שהיא תתקבל במודע במפרט ולא תתגלה אחר כך.
על מה זה רץ. אפליקציית ווב שמכוונת לנייד מספיקה לגרסה ראשונה. אפליקציות נייטיב מכפילות את השטח ומוסיפות את החנויות, ולא קונות שום דבר ממה שתיארתם.

6 · מה עדיין לא ידוע

מה לא ידוע	מה ייתן תשובה
איזה ספק תשלומים יסכים לקלוט את המודל, וכמה הוא גובה על החזקת כסף והעברה לאנשים פרטיים	פנייה ישירה לשניים־שלושה ספקים עם המודל האמיתי. חלק מבדיקת ההיתכנות
האם שמרטפים חייבים אימות זהות כדי להופיע, וכמה זה עולה לשמרטף	סקירה קצרה של מה ששירותים דומים עושים, יחד עם תמחור ספקים. חלק מבדיקת ההיתכנות
האם חמישה־עשר השמרטפים שלכם יירשמו ביום הראשון, ובאילו תנאים	שיחה שאפשר לעשות השבוע בלי לשלם לאף אחד. היא משנה מה הגרסה הראשונה חייבת לעשות
האם הגרסה הראשונה בכלל צריכה דירוגים	החלטת תכנון, נסגרת בבדיקת ההיתכנות

7 · סיכונים

חלק התשלומים הוא לוח הזמנים. אם ספק לא יאשר את המודל, התאריכים זזים בלי קשר לאיך שנבנה כל השאר. כדאי לסגור את זה לפני שנכתבת שורת קוד.
מרקטפלייס ריק זו בעיית מידע לפני שהיא בעיית תוכנה. גרסה ראשונה בלי שמרטפים בפנים אי אפשר להראות לאף אחד. את הבנייה צריך לתכנן סביב הכנסת שמרטפים אמיתיים, לא סביב יום השקה.
מידע מפוקח. אם נאספים מסמכי זהות, הם מוחזקים בסטנדרט אחר לגמרי מטלפון. זול לתכנן מראש, יקר להוסיף אחר כך.
התרחבות תכולה. כל פריט במסמך הזה ניתן לבנייה. כולם יחד, בגרסה ראשונה — ככה בניות בשלב רעיון חורגות מהתקציב.

8 · הצעדים הבאים

בדיקת היתכנות — ₪1,500–2,500 + מע״מ. הרעיון ממופה מקצה לקצה למפרט מוכן לבנייה: ארכיטקטורה, אינטגרציות, זרימת נתונים, לוח זמנים ותקציב. המספר המדויק בתוך הטווח נקבע לפי כמה משאלת התשלומים צריכה להיסגר במהלכה. תשובה למסמך הזה מספיקה כדי לקבוע.
לפני זה, ובחינם: תשאלו שלושה מתוך חמישה־עשר השמרטפים אם היו נרשמים, ואיזו עמלה היו מוכנים לשלם. התשובה משנה מה הגרסה הראשונה חייבת להיות, ואף אחד לא צריך לקבל תשלום כדי שתדעו אותה.
בנייה מתומחרת מתוך המפרט המוגמר ונסגרת כהחלטה נפרדת. שום דבר כאן לא מחייב אתכם אליה.

▸ Worked example 2 · operating business · English Fictional

Fictional. Invented company, invented details.

דוח האבחון · Diagnostic Report

Eitan Shamir · [company] · 21 August 2026 · Technical Diagnostic, 30 minutes

A technical read on what you described. Not a specification.

1 · What we understood

Orders arrive by email from about thirty regular trade customers, in free text, and someone on your team retypes each one into Priority. In your words: "we hired a person basically to copy emails into a system."

When we said it back, you corrected it: "it's not the typing that kills us, it's that when someone types it wrong we find out three days later from an angry customer."

Your numbers, as you gave them: about 40 orders a day, roughly 3 minutes each, one person doing it, plus a second person checking on busy days.

2 · The recommendation

Viability Assessment — ₪1,500–2,500 + VAT.

3 · Why this route

The automation itself is well-understood work: read the mail, extract the order, write it into Priority. The reason it cannot be quoted from this call is a single fact neither of us has yet — whether your Priority licence includes API access. With it, this is an integration. Without it, it becomes file-based or interface-driven work, and the effort and the fragility both change materially. That one fact is worth establishing before anyone quotes.
The problem you corrected us to is not retyping, it is silent errors discovered three days later. That changes what gets built: the valuable part is not the typing robot, it is the check that refuses a bad order at the door and puts it in front of a human immediately. That is a different design, and designing it is what the assessment is for.
Order emails from thirty customers are not one format, they are up to thirty. How many distinct shapes there really are decides whether this is a small build or a medium one, and it is answered by looking at a real sample — which is assessment work, not call work.

4 · Why not the other routes

Straight to a build. Two open questions — the licence and the format spread — each swing the effort by a factor. A quote made now would be a guess with a number on it.
Workspace Setup configures AI on one person's machine for a day. It would not touch the order path.
Professional Training would teach your team to work with AI. It would not stop the retyping.
Consultation & Advising is a second opinion without a build. You do not want an opinion, you want the retyping gone.

5 · How it would work

A watcher on the order mailbox. New mail from a known customer is picked up automatically. Nothing is deleted and nothing is moved; the mailbox stays exactly as it is today.
Extraction. Customer, items, quantities, delivery date, reference. Modern extraction handles free-text order emails well — the difficulty is not reading them, it is knowing when it has read one wrong.
Validation before the write. This is the part that answers what you actually complained about. Does the customer exist in Priority, does the item code exist, is the quantity within a sane band, is the delivery date possible? Anything that fails does not get written.
An exception queue. Everything that fails validation lands in one list a human clears — probably the same person, spending a fraction of the time, on the small number of orders that are genuinely ambiguous rather than on all forty. This is the design that turns "we find out three days later" into "we find out in four minutes", and it is the single most valuable thing in this document.
The write into Priority. Via API if the licence allows. If not, the alternatives exist and cost more, which is why the licence question is item one below.
Nothing about how you sell changes. Customers keep emailing exactly as they do now. They are never asked to use a portal.

6 · What is still unknown

Unknown	What would answer it
Whether your Priority licence tier includes API access, and who owns the account	One email to your Priority reseller. You can send it today, and it is the biggest single unknown in this document
How many genuinely distinct order-email formats exist across the thirty customers	Reviewing a sample of ~100 real order emails. Part of the assessment
What share of orders would fail validation and land in the exception queue	The same sample gives a defensible estimate. It decides whether one person keeps a slice of this job or none of it
Who owns the exception queue day to day	A decision for you, and it is worth making before the build rather than after

7 · Risks

The licence question can change the shape of this project entirely. Every other estimate here sits on top of it.
Format drift. A customer changes their template and extraction quality drops. Validation is what stops that becoming a silent failure, which is why it is in the design rather than added later.
Silent failure is the failure that matters. A system that stops loudly is an inconvenience; one that writes wrong orders quietly reproduces the exact problem you described, faster. Anything built here has to alert on the way out, not just log.
The person whose job this is. Someone currently spends most of a day on this. That is a conversation to have deliberately, before the system arrives and has it for you.

8 · Next steps

Today, free: email your Priority reseller and ask whether your licence tier includes API access, and whether a sandbox or test company is available. Forward the reply.
Viability Assessment — ₪1,500–2,500 + VAT. The order path mapped end to end into a build-ready specification: the extraction, the validation rules, the exception queue, the Priority integration, timeline and budget. The figure inside the range depends on how many distinct order formats the sample turns up. The specification is yours to keep.
The build is quoted from the finished specification and decided separately.

▸ Worked example 2 · operating business · Hebrew Fictional

בדוי. חברה בדויה, פרטים בדויים.

דוח האבחון

איתן שמיר · [חברה] · 21 באוגוסט 2026 · אבחון טכני, 30 דקות

קריאה טכנית על מה שתיארתם. לא מפרט.

1 · מה הבנו

הזמנות מגיעות במייל מכשלושים לקוחות סיטונאיים קבועים, בטקסט חופשי, ומישהו אצלכם מקליד כל אחת מהן מחדש לתוך פריוריטי. במילים שלכם: ״גייסנו בן אדם בשביל להעתיק מיילים לתוך מערכת.״

כשחזרנו על זה, תיקנתם: ״לא ההקלדה הורגת אותנו, אלא שכשמישהו מקליד לא נכון אנחנו מגלים את זה שלושה ימים אחר כך מלקוח כועס.״

המספרים שלכם, כמו שמסרתם אותם: בערך 40 הזמנות ביום, כ-3 דקות כל אחת, אדם אחד עושה את זה, ובימים עמוסים עוד אחד בודק.

2 · ההמלצה

בדיקת היתכנות — ₪1,500–2,500 + מע״מ.

3 · למה המסלול הזה

האוטומציה עצמה היא עבודה מוכרת: לקרוא את המייל, לחלץ את ההזמנה, לכתוב אותה לפריוריטי. הסיבה שאי אפשר לתמחר אותה מהשיחה הזאת היא עובדה אחת שאין לאף אחד מאיתנו עדיין — האם רישיון פריוריטי שלכם כולל גישת API. אם כן, זו אינטגרציה. אם לא, זה הופך לעבודה מבוססת קבצים או ממשק, והמאמץ והשבירות משתנים מהותית. את העובדה הזאת שווה לברר לפני שמישהו מתמחר.
הבעיה שתיקנתם אותנו אליה היא לא ההקלדה אלא טעויות שקטות שמתגלות אחרי שלושה ימים. זה משנה מה בונים: החלק בעל הערך הוא לא רובוט הקלדה, אלא הבדיקה שמסרבת להזמנה שגויה כבר בכניסה ושמה אותה מול בן אדם מיד. זה תכנון אחר, ובדיוק בשביל לתכנן אותו קיימת בדיקת ההיתכנות.
מיילים של הזמנות משלושים לקוחות הם לא פורמט אחד אלא עד שלושים. כמה צורות שונות באמת יש — זה מה שמחליט אם זו בנייה קטנה או בינונית, וזה נענה מהסתכלות על מדגם אמיתי, שזו עבודה של בדיקת היתכנות ולא של שיחה.

4 · למה לא המסלולים האחרים

ישר לבנייה. שתי שאלות פתוחות — הרישיון ופיזור הפורמטים — כל אחת מהן מזיזה את המאמץ בפקטור. הצעת מחיר עכשיו היא ניחוש עם מספר עליו.
הקמת סביבת עבודה מגדירה AI על המחשב של אדם אחד ליום. היא לא נוגעת במסלול ההזמנות.
הכשרה מקצועית תלמד את הצוות לעבוד עם AI. היא לא תפסיק את ההקלדה מחדש.
ייעוץ וליווי זו חוות דעת בלי בנייה. אתם לא רוצים חוות דעת, אתם רוצים שההקלדה תיעלם.

5 · איך זה עובד

מאזין על תיבת ההזמנות. מייל חדש מלקוח מוכר נאסף אוטומטית. שום דבר לא נמחק ושום דבר לא מוזז; התיבה נשארת בדיוק כמו היום.
חילוץ. לקוח, פריטים, כמויות, תאריך אספקה, אסמכתא. חילוץ מודרני מסתדר טוב עם מיילים של הזמנות בטקסט חופשי — הקושי הוא לא לקרוא אותם, אלא לדעת מתי הוא קרא לא נכון.
בדיקה לפני הכתיבה. זה החלק שעונה על מה שבאמת הפריע לכם. האם הלקוח קיים בפריוריטי, האם מק״ט הפריט קיים, האם הכמות בטווח הגיוני, האם תאריך האספקה אפשרי. מה שנכשל — לא נכתב.
תור חריגות. כל מה שנכשל בבדיקה נוחת ברשימה אחת שבן אדם עובר עליה — כנראה אותו אדם, בחלק קטן מהזמן, על מספר ההזמנות שבאמת דו־משמעיות במקום על כל הארבעים. זה התכנון שהופך ״מגלים אחרי שלושה ימים״ ל״מגלים אחרי ארבע דקות״, וזה הדבר בעל הערך הגבוה ביותר במסמך הזה.
הכתיבה לפריוריטי. דרך API אם הרישיון מאפשר. אם לא — קיימות חלופות, והן עולות יותר, ולכן שאלת הרישיון היא סעיף ראשון למטה.
שום דבר באופן שבו אתם מוכרים לא משתנה. הלקוחות ממשיכים לשלוח מייל בדיוק כמו היום. אף אחד לא מתבקש להתחיל להשתמש בפורטל.

6 · מה עדיין לא ידוע

מה לא ידוע	מה ייתן תשובה
האם רמת הרישיון שלכם בפריוריטי כוללת גישת API, ועל שם מי החשבון	מייל אחד למשווק פריוריטי שלכם. אפשר לשלוח אותו היום, וזה הנעלם הכי גדול במסמך הזה
כמה פורמטים שונים באמת של מיילי הזמנות קיימים בין שלושים הלקוחות	בחינה של מדגם של כ-100 מיילי הזמנות אמיתיים. חלק מבדיקת ההיתכנות
איזה אחוז מההזמנות ייפול לתור החריגות	אותו מדגם נותן הערכה שאפשר לעמוד מאחוריה. זה מחליט אם אדם אחד נשאר עם חלק מהתפקיד או עם כלום
מי אחראי על תור החריגות ביומיום	החלטה שלכם, וכדאי לקבל אותה לפני הבנייה ולא אחריה

7 · סיכונים

שאלת הרישיון יכולה לשנות את כל צורת הפרויקט. כל הערכה אחרת כאן יושבת עליה.
סחיפת פורמטים. לקוח משנה תבנית ואיכות החילוץ יורדת. הבדיקה היא מה שמונע מזה להפוך לכשל שקט, ולכן היא בתכנון ולא תוספת מאוחרת.
כשל שקט הוא הכשל שמשנה. מערכת שנופלת ברעש היא מטרד; מערכת שכותבת הזמנות שגויות בשקט משחזרת בדיוק את הבעיה שתיארתם, רק מהר יותר. כל מה שנבנה כאן חייב להתריע בדרך החוצה, לא רק לתעד.
האדם שזה התפקיד שלו. מישהו מבלה היום כמעט יום שלם על זה. זו שיחה שכדאי לנהל במודע, לפני שהמערכת מגיעה ומנהלת אותה במקומכם.

8 · הצעדים הבאים

היום, בחינם: שלחו מייל למשווק פריוריטי ושאלו אם רמת הרישיון שלכם כוללת גישת API, והאם יש סביבת בדיקה או חברת ניסיון. העבירו לנו את התשובה.
בדיקת היתכנות — ₪1,500–2,500 + מע״מ. מסלול ההזמנות ממופה מקצה לקצה למפרט מוכן לבנייה: החילוץ, כללי הבדיקה, תור החריגות, האינטגרציה מול פריוריטי, לוח זמנים ותקציב. המספר בתוך הטווח תלוי בכמה פורמטים שונים המדגם יעלה. המפרט נשאר שלכם.
הבנייה מתומחרת מתוך המפרט המוגמר ונסגרת כהחלטה נפרדת.

§5 · The booking intake

What a visitor answers when booking

Friction at the booking moment is the most expensive friction on the site, so every field costs bookings. The whole intake is therefore one new tap and one re-specified field that already ships. Nothing else.

Where it sits in the flow

Top of step 0, above the name field, on the free-call path only. Someone who already had the free call must never be re-asked.

Step 0, not step 2. A field costs most when it breaks a mode and least when it sits inside one the user is already in. Step 0 is already a form. Step 2 is the check-it-over and confirm rhythm, the most conversion-sensitive beat on the page, and an unanswered-looking control between the visitor and the confirm button manufactures a perceived requirement.
Above the name, not below. It is the only field answerable with zero typing, so leading with it gives an immediate sense of having started without opening a keyboard. Costs one row of about 90px, and the contact fields stay above the fold at 375px.
Never on step 1. That step auto-advances 420ms after the slot tap. Anything there is scrolled past.

The booking flow as it stands

Step 0 details: name (required), phone, email, with one of the two required, plus an optional notes textarea.
Step 1 time, auto-advancing 420ms after the tap.
Step 2 review and confirm.

The free-text field already ships. It is on the page today, capped at 2000 characters, and under-specified. Re-specifying it costs zero new fields. The validator now enforces name plus one channel, matching the contract printed above the fields.

The question, exact copy

It reuses the three situations the visitor already read on the homepage, which turns the intake from a classification task into a recognition task, the cheapest cognitive act there is. It is also the only axis that actually routes the funnel, because each option lands on a different offering.

Intake question · English

Where are you starting from?
So the call starts in the right place — skip if none fit.

◻ An idea — nothing built yet
◻ An operation that runs by hand
◻ AI that should already be running

Intake question · Hebrew

מאיפה אתם מתחילים?
כדי שהשיחה תתחיל במקום הנכון — אפשר גם לדלג.

◻ רעיון — עוד לא נבנה כלום
◻ תפעול שרץ ידנית
◻ AI שכבר היה אמור לעבוד אצלכם

No fourth "something else"

The question is optional, so not tapping already means none of these. That is cleaner than an escape hatch, which would dilute the aggregate.

The skip is explicitly licensed

It costs answer rate and it buys away the stall. Bookings rank above data. Expect a meaningful share to skip, and never build routing, reporting or prep that assumes the tap is there.

Option 2 is broadened

The homepage card can be vivid about spreadsheets. The intake is a router, and it must not make a business whose manual work is not spreadsheet-shaped hesitate.

The re-specified notes field

	Now	Becomes
EN label	`Notes (optional)`	`Anything worth knowing (optional)`
EN placeholder	`Anything worth knowing before the call?`	`One line about the idea or the process — a link to your site works too.`
HE label	`הערות (לא חובה)`	`משהו שכדאי לדעת (לא חובה)`
HE placeholder	`משהו שכדאי לדעת לפני השיחה?`	`שורה אחת על הרעיון או על התהליך — גם קישור לאתר שלכם עוזר.`

"Notes" is a bureaucratic word that invites nothing. The new placeholder asks for exactly two things and names both, which is the difference between an 8% fill rate and a 30% one. It also recovers the company-URL field at zero cost. Expect under a third even after the re-spec, which is still far more useful context than the current label earns.

What got cut, and why

Field considered	Verdict	Reason
Starting point. Three options, tap, optional	Ask	Decides which of three frames Tom opens with, and on one of the three the ₪250 diagnostic may not be the right next step at all
One-line free text	Keep	Already ships. A specific sentence in their own words beats every dropdown
Private vs business	Cut	Forks meaningfully on one path only, and only into which offering. Tom asks it in five seconds as his opener
Role or job title	Cut	Adds nothing the starting point does not imply. Pure B2B-qualification smell on a free call
Idea / solo / corporate	Cut as worded	Mixes two axes, stage and size. Someone with an idea who works at a corporate stalls, and hesitation at the booking moment is the exact cost being avoided
Company or website URL	Cut, folded in	Genuinely useful for pre-call research, but it is a typing field and it excludes idea-stage entirely. The new placeholder recovers it at zero cost
Conditional "you or a team?"	Cut	Real routing value, but it changes nothing Tom does in the five minutes before a ten-minute call. He asks it in three seconds

Wiring: ship the cheap one first

v0 · zero backend change, shippable tomorrow

Prefix the tap answer into the existing notes string on submit, using tags of [idea], [manual-ops] and [ai-enablement].

It arrives in the WhatsApp ping, the approval email, the booking-requests row and the Notion Notes property, all of which already carry notes end to end. Tom sees [manual-ops] we retype orders from email into Priority on his phone. Costs one front-end change and nothing else.

v1 · a proper field, when the site session is next open

The booking payload gains startingPoint, one of idea, manual, ai or empty.
The funnel service validates it against that allow-list and carries it into the claim payload. This now lives in the Python service, not in an n8n node.
Migration: add a nullable starting_point text column to the booking-requests table.
The request-prep step writes the Notion Starting point select and adds one line to the WhatsApp text.

Ship v0 now, v1 whenever the site session is next open. The learning loop works off Notion either way. v0 just means one regex on the Notes field for the first month.

§6 · Routing

Signals to offering

The free call routes on the noun and on the trigger. The diagnostic routes on what the thirty minutes could and could not settle. Read each table downward and take the first row that matches.

Out of the free call

Signals	Recommended next step
Noun is a system, a product or a process. Q3 has a dated trigger. Q4 names a range or shrugs off ₪250	Technical Diagnostic, ₪250 + VAT
Noun is a person, a team or "my staff". The object is people learning	Professional Training, priced after a call
Noun is "us", "how we work", "my day". One person, hands-on, no single process pain	Workspace Setup, from ₪2,500 + VAT
Noun is "whether we should" or "which way". They already have technical people	Consultation & Advising
Q5 multiplied by Q6 does not clear the build cost	Honest no (i), the value math
An off-the-shelf tool already does it	Honest no (ii), try it first and come back in a month
Buildable, outside your strengths	Honest no (iii), not us
Q3 empty and Q4 empty	Park. Not now, not never. The free call stays free when they return

Out of the diagnostic

What the thirty minutes established	Recommended offering
A real build, but one or more facts that swing the effort by a factor are still unknown. Access unresolved, format spread unknown, or the product splits into unequal parts and the expensive part could not be sized	Viability Assessment, ₪1,500 to ₪2,500 + VAT. The default
Data lives in someone's head or a WhatsApp group. The ask is how to actually work this way day to day. One person, one set of accounts	Workspace Setup, from ₪2,500 + VAT
The blocker is people rather than systems. A group, a format, recurring sessions	Professional Training, priced after a call
They have technical capacity and want judgement, a second opinion or ongoing input, with no build attached	Consultation & Advising
A written specification already exists, access is confirmed, and nothing material is unknown	Build, quoted from the existing specification. Rare
Frequency times duration does not clear the build cost, or an off-the-shelf tool already does it	Honest no, written up shorter and without apology

Tie-breaks, in order of precedence

1. Access beats everything. If question F left the licence, the account owner or a test environment unresolved, the route is the assessment and never a build quote, however clear the rest looked. Every other estimate sits on top of that one fact.

2. Follow the correction, not the request. When the read-back correction moves the problem, the route is chosen against the corrected problem. A typing robot and a validation queue are different builds, and only one of them is what they actually bought.

3. Where the data lives beats what they asked for. Data in a head or a chat group means Workspace Setup, even when they arrived asking for a build.

4. Unequal parts mean assessment. If the product splits into pieces of very different sizes and the largest could not be sized in thirty minutes, that is the assessment's job by definition.

5. When two routes still tie, name the cheaper one. The report is theirs to keep and the next rung is decided separately, so the cheaper honest route costs nothing and buys the trust that makes the larger one sellable.

6. Never quote their project in the room. Only published ladder prices are said out loud. Anything else is a guess with a number on it, and you will be held to it.

§7 · The learning loop

What gets captured, and what gets reviewed

The loop dies if capturing data is a separate chore, so it is not one. One printed sheet per call does three jobs: the notes taken during the call, the outline the report is written from, and the fields tapped into Notion afterwards.

Honest starting position

The funnel database holds four rows, three of which look like tests. At this volume the question "which questions predict a sale" is unanswerable and will stay unanswerable for months. So: capture from call number one, review qualitatively from month one, and run the predictive cut only at 30 or more free calls with 8 or more paid conversions. Anyone who hands Tom a statistical dashboard before that is selling him a toy.

Where things live

Surface	Holds	Written by
Notion, Leads and Funnel	The judgement data: route, signals, outcome, their words	The funnel service on booking, then Tom within 90 seconds of each call
Postgres booking tables	The behavioural data: requested, confirmed, rescheduled, cancelled, expired, timestamps, auto-confirm flag, reschedule count	Automatic. A database trigger already logs every status and slot change
The paper sheets	The verbatim words, the M-counts, the working	Tom, during the call

No ETL, no sync job, no dashboard. Not yet. Postgres already captures no-shows, reschedules and cancels for free and joins to Notion on email. Building a pipeline between them for four rows is exactly the disproportionate friction this playbook exists to flag. Revisit at 30 leads.

The two capture sheets

Free-call sheet, one side of A5

NAME ____________  DATE/TIME ________  LANG he/en
START (tap): idea / manual / ai / —
THEIR LINE (intake): ____________________________

Q1 WHAT  (verbatim — travels the whole funnel)
  "______________________________________________"

Q2 NOW  ______________  [ ] real process  [ ] fog

Q3 WHY NOW  __________  [ ] dated trigger [ ] none
   if nothing changes for a year: ______________

Q4 WHO PAYS  _________  [ ] number named
                        [ ] no authority

Q5 SIZE  ____ × ____    [ ] real [ ] thin [ ] n/a
Q6 STACK ____________   [ ] open [ ] locked [ ] n/a
Q7 PRIOR ____________   [ ] burned         [ ] n/a

FREE ANSWER GIVEN: ______________ (one. only one.)

ROUTE: diagnostic / workspace / training / advising
       / no(value) / no(exists) / no(not-us) / park
LENGTH: ____ min   HOW THEY FOUND ME: ___________

Diagnostic scorecard, filled in the 0:23 pause

TRACK  A idea / B operating
M-COUNT (manual steps): ____
WHAT IT ACTUALLY IS: ____________________________
THE CORRECTION (verbatim — opens the report):
  "______________________________________________"
THE ONE JOB / THE PAIN STEP: ____________________
ACCESS: API? y/n/unknown  owner: ______  test? y/n
DATA LIVES: _____________________________________
BREAKS WHEN: ____________________________________
RUNS IT AFTER: ______ DECIDES: ______ DATE: ______

ROUTE: viability / training / workspace / advising
       / build-quote / honest-no
BECAUSE: ________________________________________
NOT __________ BECAUSE __________________________
UNKNOWN #1 (the one that moves the number): _____
  answered by: __________________________________
RISKS: 1 __________ 2 __________ 3 ______________
FREE HOMEWORK FOR THEM: _________________________

Everything in the scorecard maps one to one onto a report section. Writing the report becomes transcription plus prose, which is what makes a next-business-day turnaround achievable at all.

The six Notion properties to add first

Property	Type	Options	Written by
`Starting point`	select	`idea`, `manual-ops`, `ai-enablement`, `not-answered`	The service on v1, or Tom from the tag prefix on v0
`Track`	select	`A-idea`, `B-operating`	Tom, 1 tap
`Green flags`	multi-select	`Q1-clear-object`, `Q2-real-process`, `Q3-dated-trigger`, `Q4-number-named`, `Q5-volume-real`, `Q6-open-stack`	Tom, 0 to 6 taps
`Red flags`	multi-select	`Q1-fog`, `Q2-no-process`, `Q3-no-trigger`, `Q4-no-authority`, `Q4-wrong-stage`, `Q5-volume-thin`, `Q6-locked-stack`, `Q7-prior-burn`	Tom, 0 to 8 taps
`Route recommended`	select	`diagnostic`, `viability`, `training`, `workspace`, `advising`, `honest-no`, `park`	Tom, 1 tap
`Their words`	rich text		Tom, paste the Q1 sentence

The flag design is the whole trick

Recording an answer-shape for six questions per call is too much friction and would be abandoned by week three. Recording only the flags that actually fired is 2 to 4 taps, and it is exactly the data that answers the question. A question earns its place if its flag fires often enough to matter and separates outcomes. A flag that fires on everyone, or on no one, is dead weight, and that shows up in this data with no extra work.

Add later, once the first six have survived a month of real use: Questions asked, Call length, Honest-no reason, Report sent, Diagnostic outcome, Deal value, Objection, Found me via. Two additions to the existing Stage vocabulary close the loop: Report sent between Discovery done and Spec sold, and Parked alongside Lost. A park is not a loss and must not be counted as one.

The 90 seconds after every call

Non-negotiable, and it happens in the reserve minute while the call is still in his head, on his phone, in the Notion mobile app.

Open the lead row. It is already there, created at booking.
Set Track, Route recommended, Stage.
Tap the flags that fired. Only the ones that fired.
Paste the Q1 sentence into Their words.

This is the single highest-risk component of the whole system. Post-call admin decays by week three in every organisation that has ever existed. It survives only if it is on his phone, under 90 seconds, and done in the reserve minute rather than batched at the end of the day. If it is being skipped after a month, cut it to three fields, being route, their words and the red flags. Three fields kept beats twelve abandoned.

The monthly review, 20 minutes, first Sunday

Read in this order. The first two are the only ones that matter before 30 calls.

Read every Their words from the month, out loud, in one sitting. Nothing else in this system will teach more. Patterns in how people describe their own problem are what the questions should be shaped around, and no statistic surfaces them.
Every call that went nowhere: which question would have caught it earlier? If the answer is none of them, that is a missing question, so add it. If it is the same question three months running, move it earlier in the order.
Flag frequency. A flag that fired on zero calls is badly worded or asking about something that does not happen. A flag that fired on every call carries no information. Both are candidates to cut, but never before three months of data.
Conversion by starting point. Free to diagnostic, and diagnostic to sale. If idea-stage converts at a fraction of operating-business, the track A close needs different words, not a different funnel.
The behavioural read from Postgres, which is free and already captured: requests, confirmed, expired, cancelled, rescheduled, and average lead hours by booking type. Average lead hours is the one to watch, because bookings made for today or tomorrow convert differently from bookings made for next week, and that is a scheduling decision rather than a script decision.
The turnaround audit. Report sent minus the call date. If it slips past one business day even twice, the promise comes out of the booking page before it damages anything.

The predictive cut runs only at 30 or more free calls and 8 or more paid conversions. For each flag, compare the conversion rate when it fired against the base rate. Retire a question when its flags fire on 20% or more of calls and the conversion gap between fired and not-fired stays under 5 points across two consecutive quarters. Below those thresholds, changing questions on the evidence is superstition.

§8 · Open items

What Tom must decide

Two of the original blockers are closed. What follows is the live list, ordered by what blocks what.

Decided

The names. «אבחון טכני» / Technical Diagnostic, «דוח האבחון» / The Diagnostic Report, «שיחת בירור» as the Hebrew name for the free call, and no English name for the free call. Shipped and live. Do not let a translator "fix" the asymmetry.

Decided

The booking validator. Name plus one channel, matching the contract printed above the fields. Fixed and verified live. Bookings with an email and no phone go through.

#	Decision	What is at stake
1	Grow verification	The narrowest point in the ladder. The entire system converges on a ₪250 close that cannot take a card, so every diagnostic booking is a manual invoice chase at the moment of highest intent. PayPlus is the priced fallback and has been sitting unactioned since 2026-07-01. No script fixes this.
2	The intake question and its placement	One optional tap, three options, top of step 0, free-call path only, plus the notes label and placeholder re-spec. And approval to ship v0, the tag folded into notes with a front-end change only, rather than waiting for a backend change.
3	The report turnaround promise	Recommended: by 18:00 the next business day, Israel time. It goes in the paid script's opening and on the booking page. Say no now if it is not always keepable. An unkept written promise on the paid rung is worse than a vaguer one.
4	Record the paid session, or not	The script currently asks consent for notes. Recording is possible on Zoom's free tier but needs a consent line, a storage answer and a privacy-policy sentence, for roughly one paid session a week. Default: stay on notes, and revisit at three a week.
5	Can the diagnostic ever be skipped?	Someone arrives with a written brief, prior discovery done, and a named build budget. Straight to the Viability Assessment, or always through the ₪250 first? The routing table currently says always through. Confirm or override, but the override must be a standing rule, never a judgement call made live on the phone.
6	Refund policy when the answer is "don't build this"	The Terms cover cancellation up to 24 hours before. Nothing covers the outcome. Recommended: no refund, made fair by promising "don't build this" as a possible outcome in the opening, which the script already does. Confirm.
7	The six P0 Notion properties and the two Stage additions	Report sent and Parked. These are the difference between a funnel that can be reviewed and one that cannot.
8	The 90-second post-call capture	Tom commits to it or the loop is fiction. If he will not, say so now and cut it to three fields, rather than discovering the decay in November.
9	Book 15-minute gaps, not 10	A calendar decision, and the difference between the script working and the day sliding. Ten-minute calls run 12 to 14 minutes routinely.

Friction that is not worth its value, cut before it gets built

The conditional "you or a team?" follow-up tap. Real routing value, changes nothing Tom does.
Private vs business, and role. Qualification smell on a free call, on the exact page a rebuild just stripped of friction.
Recording plus transcription plus auto-summarisation. Needs a recording decision, a consent line, a storage policy and a privacy-policy edit, for about one paid session a week.
A Notion to Postgres sync for the learning loop. Four rows. No.
All twelve Notion properties on day one. Ship six.
A funnel dashboard. Not until the monthly review has been done by hand three times and Tom knows which six numbers he actually looks at.