In April 2025, we worked on the offer of a financial institution that sells its services to other businesses. The promise had been decided, and its economic effect had been named. We lacked three pieces of data to put a figure on it. Published as it stood, the wording of that effect would have left the sales team with no answer to a buyer’s first request for the calculation. In the working document, we classified each element by its status: observed, inferred or projected. We kept that wording for internal use while the data was missing.
This classification rests on three separate notions: the promise, the impact and the proof. Each answers a different question. This article gives a test for each.
The promise: what becomes true for the client
The promise states what becomes true for the client, something the client could not achieve alone. It commits to one result. Take an example reconstructed for this article: a firm that provides outsourced finance director services. Its promise: “take back control of your figures in ninety days”.
A list of benefits commits to no result. “Compliance, time saved, peace of mind, support”: none of these words can be disproved, so none can be checked.
The promise test. Replace your name with a competitor’s. If the sentence is still true, the promise has not yet been decided.
The impact: the sign that will show the result
The impact names the observable sign that will show the promise has been kept. It can be seen in the client’s business, at five levels: a decision taken earlier, a changed organisation, faster execution, a measured result, an instruction a newcomer applies without help. In the firm’s example, the impact is this: the managing director can say at any time how many months of cash the company has left.
The first confusion is to write the impact as a second promise. “More peace of mind” and “more agility” describe a state of mind that nobody can observe. Our method, the Clear Offer Framework, calls this flaw “phantom impact”. Presenting an activity as a result causes the same flaw. “We keep your books” describes a task done by the provider; the impact is what that task changes for the client.
The impact test. Ask where and when a third party would see the change. If there is no place and no date, the impact is still to be defined.
The proof: the answer to a specific doubt
Proof removes a doubt the prospect actually has. That doubt concerns the promise, the mechanism that produces the result, the impact, or whether the result can be repeated in the prospect’s own business. A row of client logos answers one question only: have other companies bought? It says nothing about what they obtained or how.
The second confusion is between impact and proof. The impact is observed in the client’s business, after the engagement. Proof works before the purchase: it gives the prospect reasons to expect that impact for themselves. The doubt can also concern how the result is measured, and it can remain after the contract is signed.
In September 2026, the marketing head of a B2B information services publisher returned our signed proposal to us with her questions. One concerned how we measure her brand’s presence in generative AI engines, whose answers vary from one query to the next. That same day, we sent her our protocol. Each search intent is put to each engine three times: a main wording, a rewording, then a follow-up that names the brand. A reading that varies from one pass to the next is run again and flagged as unstable. Each reading is kept as a time-stamped screenshot and export, so that the publisher can run the test again. This answer addresses the doubt she raised: the reliability of a measurement that varies. A March 2026 study of Perplexity, SearchGPT and Gemini measures how the cited sources vary and criticises single readings.
In the April 2025 engagement, by contrast, no document held by the client yet contained the figures that would have demonstrated the economic effect that had been named.
The proof test. For each piece of proof on the page, write down the doubt it removes. If you cannot find that doubt, remove the proof or link it to an impact.
The order of work and the three-column table
The three notions depend on one another, in a fixed order. The impact turns the promise into observable signs, which requires a decided promise. Proof supports the impact or the mechanism, so both must be named before it. The work therefore starts with the promise and ends with the proof.
Before putting an offer page online, draw three columns:
- Promise: the single sentence, run through the competitor test.
- Impact: for that promise, two or three observable signs, each with its place and date.
- Proof: for each sign, the element that removes the prospect’s doubt, with its status: observed, inferred or projected.
We applied this table to the reconstructed firm, using our Clear Offer Framework assistant. The starting wording we submitted: “Your financial co-pilot: compliance, time saved, peace of mind, support”, backed by a row of logos and the line “more than 40 companies supported”. According to the assistant, this wording fails the competitor test: the sentence remains true for most firms on the market. It then proposes three impacts and their proof, all with the status “projected”: these are hypotheses for the reconstructed case, to be checked against a real firm’s data.
| Projected impact | Where and when to observe it | Proof to provide |
|---|---|---|
| The managing director receives a usable management dashboard on a fixed date. | At the management committee, every month, a set number of days after the close. | A dated record of six to twelve closes, showing the actual delivery date of each dashboard, before and after the engagement. |
| Cash tensions appear early enough to be dealt with before they become urgent. | In the cash-flow forecast, at each weekly or monthly review, over a set horizon. | An anonymised client case: the dated forecast, the gap detected, the decision taken and its result. |
| Financial and regulatory obligations are met without incident or last-minute reminders. | In the deadline calendar, at each deadline over a given period. | The deadline log, with planned and actual dates, and documented evidence that there were no delays or penalties over the period. |
The logos and the number of clients are observed elements. They prove none of these three impacts.
In your own table, an empty cell in the Proof column marks data still to be obtained. Flag it internally, as we did in the April 2025 engagement. Until you have it, remove from the page the wording it was meant to support. If your sales pitch follows the Advantage, Proof, Benefit structure, the same rule applies to its proof: it must remove a doubt that the advantage raises.
This work is part of the Clear Offer Framework. It assumes the offer has already been decided.
Frequently asked questions
What is the difference between an offer’s promise and its impact?
The promise states what becomes true for the client thanks to the offer. The impact names the observable sign that will show this change has happened: a decision, a deadline, a cost, a task handed to someone else. A promise without an impact cannot be checked.
What makes proof useful in a B2B offer?
Useful proof removes a specific doubt the prospect has: about the promise, the mechanism, the impact or whether the result can be repeated in their business. A logo or a figure without context removes none of these doubts. For each piece of proof, you should be able to write down the doubt it removes.
In what order should you work on the promise, the impact and the proof?
The promise first, the impact next, the proof last. The impact turns the promise into observable signs, and the proof supports those signs. Proof chosen before the impact is tied to no observable sign.