You priced delivery risk and you are good at it. The risk that costs you is the other one: a client whose idea had no buyer behind it. You ship on time, the launch goes quiet, and what you lose is the renewal, the case study and usually part of the final invoice. Screening the idea before the contract is margin protection, and 5 of the 10 questions below can be answered by the client today.
The failure that does not show up in your process
Your retrospectives measure delivery. Scope creep, estimates, handover, bugs after launch — all real, all improvable, and none of them is the expensive failure.
The expensive one looks like success from the inside. The build was clean, the client was happy at handover, and six months later there is no traffic, no revenue and no second phase. The project is not in your failure column because it did not fail. It just never became anything you can sell with.
Two costs follow, and neither is billable. The first is the pipeline: a launched product that went nowhere is not a case study, and a portfolio of those is a portfolio of screenshots. The second is collections — a client whose product did not work is a client who re-reads the invoice.
That second one is not a guess about your business. Of the strongest problems our sweeps have documented in the agency vertical, the money mechanics come up at least as often as the work itself:
The strongest agency records in our corpus on 2026-09-07, split by whether the theme names money. Four limits travel with it: nearly every row in the corpus was seen exactly once, so this says what we have documented and nothing about which problem is most common; the records come from 19 source pages rather than 25 independent ones; 22 of 25 are really agencies when read by hand; and the split is decided by a published word list rather than by sorting them by feel.
Quote-to-cash, collections, revenue visibility — 52 per cent of what we have documented for agencies is about the money arriving, and a client whose product went nowhere sits upstream of a lot of that.
What the client can settle before you quote
Here are the ten questions our own discovery gate asks every candidate idea, with the column that matters for a proposal: who can answer it, and how.
| The question | Who answers it | What you ask for |
|---|---|---|
| Is there a real, documented problem behind it? | The client, today | Three public posts, threads or reviews describing the problem, written by someone who is not the client |
| Is there a path to money that repeats? | The client, today | Who pays, how often, and what happens on the second payment |
| Is there a repeatable way to reach the buyer? | The client, today | One channel where they have already reached this buyer, with a number attached |
| Can it charge enough to be worth one person's time? | You, in an afternoon | A comparable product with a published price, and the customer count that price implies |
| Can it run without hand-holding every customer? | You, in an afternoon | Who answers the tickets after handover, and how many minutes per customer per month |
| Does it survive if one platform changes its rules? | You, in an afternoon | Which third party can end this with a policy change, and what the fallback is |
| Is there a product left if an AI vendor ships this feature? | You, in an afternoon | What is left of the product if the model vendor ships this as a feature |
| Can you, specifically, sell and operate it? | The client, today | Who operates and sells this after you hand it over, by name |
| Can it launch without licences or compliance audits? | You, in an afternoon | Any licence, audit or certification the launch depends on, named |
| Can it sell without legal review and a sales team? | The client, today | Whether the first ten buyers can pay without legal review |
The questions are the published checks our discovery gate runs, in its own words. The other two columns are ours and are a judgement — argue with a row rather than with the idea. Note the split: 5 belong to the client and 5 you can settle yourself without asking anyone.
None of these asks the client to prove the idea will work. They ask for artefacts that either exist or do not: a link, a number, a name. That distinction is what makes this a screening sheet rather than an argument — you are not disputing their judgement, you are collecting what is already there.
The first row does most of the work. Across 23 runs of our own gate, 438 of 706 rejected candidates failed exactly that question, and it is the one a client can answer in an email or not at all. Those runs are our own accounts rather than customers', so that count describes how our gate behaves and not how a market behaves.
How to put it in a proposal without losing the deal
The objection is obvious: a client who came to get something built does not want to be told to validate first, and the agency that says so loses to the one that says yes. Both halves of that are true, which is why the move is structural rather than rhetorical.
- Sell it as phase zero, priced. A short, paid, fixed-scope week that produces the evidence above and a written go or no-go. It is billable, it is fast, and it converts an awkward conversation into a line item.
- Make the artefacts a precondition, not an opinion. "We start when we have the three links and the channel number" is a process requirement. "I am not sure your idea is good" is a fight.
- Write the stop condition into the contract. What phase zero would have to find for phase one not to happen, agreed before anyone is invested. The point is to make it survivable to be wrong.
- Let a no-go still be a deliverable. The client leaves with a written account of what was missing and what would change it. That is a referral more often than a lost client, because it is the rarest thing in this market: someone who told them the truth before taking the money.
What screening cannot do for you
It cannot tell you the idea will work. Everything above rules things out; nothing in it predicts. An idea that clears all ten questions can still fail on execution, timing, or the founder losing interest — and an agency that promises otherwise has just taken on a risk it cannot price.
It also cannot settle the client's own fit. Whether they will sell and operate the thing after handover is the question you are least able to answer and most affected by, and the honest version lives in the contract rather than in a checklist: who, by name, does the selling when you leave.
And a client determined to build anyway will build anyway. Screening is for deciding whether you want the work at your price — not for saving people from themselves.
The short version
- The expensive failure looks like a success: a clean build with no buyer behind it. It costs you the case study and slows the last invoice.
- Money mechanics are 52 per cent of the strongest records we have documented for agencies — as large a share as the work itself.
- Ten published questions settle most of it before a contract exists. 5 are the client's to answer and 5 are yours, all of them from artefacts rather than opinions.
- Price it as phase zero with a written stop condition, and let a no-go be a deliverable rather than a lost sale.