Short answer
Read feedback by level — opinion, interest, request, payment, delivery — and always next to the number of relevant people reached. Five messages from one person are one person, and silence after a weak channel is inconclusive, not a verdict.
The problem with "the feedback was positive"
After a first test, most people summarise everything they heard in one sentence: "The feedback was really positive." That sentence hides almost everything that matters. Positive how? From whom? Out of how many? Did anyone do anything, or did they just say something nice?
Reading feedback well means separating the kinds of signal, keeping the denominator visible and looking as hard for evidence against your idea as for evidence for it. Done that way, a small test can support a real decision. Done the usual way, it supports whatever you hoped to hear.
Separate the levels of evidence
| Level | Event | What it shows | What it doesn't show |
|---|---|---|---|
| 1 | Opinion or click | A first reaction | Willingness to pay |
| 2 | Relevant reply or meeting | Engagement with the problem | A commercial request |
| 3 | Sample or quote request | Concrete interest in your offer | That they'll pay |
| 4 | Payment made | Commitment at this price | That others will pay |
| 5 | Delivery accepted | Your work met their criteria | Repeat demand |
| 6 | Repeat purchase | Ongoing value in that relationship | That every segment will repeat |
Each level is stronger than the one before. A single payment usually tells you more than twenty compliments. Keep promised payments and completed payments as separate events — "we'll definitely buy next month" is level 3 at best.
Keep the denominator visible
Every result needs the number of relevant people it came from. Without it, numbers can't be compared or trusted.
| Summary without denominator | Honest summary |
|---|---|
| "3 people want it" | "3 quote requests from 40 relevant contacts" |
| "Lots of interest" | "12 replies from 40 contacted; 2 meetings" |
| "Nobody replied" | "0 replies from 6 contacted through one channel" |
The last example is important. Six messages through one weak channel is usually inconclusive, not proof that nobody wants your offer.
Count people, not messages
Deduplicate before you count. One enthusiastic person who replies five times is one person. Two contacts at the same company are often one buying decision. Count unique businesses and decision makers, and note when several people at one business were involved.
Look for counter-evidence on purpose
After every test, write down what argues against your hypothesis, even if the overall result looks good.
- Who declined, and why?
- Who said the problem is already solved?
- Did anyone say it's not a priority?
- Did the people who said yes match your target buyer — or were they friends, peers or people outside the segment?
A decision that ignores counter-evidence isn't a decision; it's a hope. Writing it down also helps you design a better next test.
Consider alternative explanations
Before you conclude, ask what else could explain the result.
| Result | Your interpretation | Alternative explanation |
|---|---|---|
| High reply rate | Strong demand | A personal connection or novelty made people reply |
| Quote request | They want to buy | They're collecting quotes for a one-off need |
| No replies | No demand | Wrong channel, wrong timing or a weak message |
| Paid pilot | Validated | One unusual customer; may not repeat |
You won't always be able to rule alternatives out. Naming them keeps you from over-claiming and tells you what to test next.
Silence is not a verdict
No replies is the most misread result. Before deciding anything, check:
- Did you reach enough relevant people? (Fewer than 8–10 is usually too few.)
- Was the channel one your buyers actually read?
- Was the message specific and personal?
- Was the timing reasonable — not a holiday week or a peak season for them?
If any answer is no, call the test inconclusive and fix that variable first.
Turn feedback into a decision
Compare what happened with the rule you wrote before the test:
- Continue — the continue threshold was met with enough relevant people.
- Change — signals were real but something specific blocked progress; change one variable.
- Stop — the stop condition was met through a channel that works, or a hard constraint appeared.
- Inconclusive — too few relevant people were reached; fix the reason and repeat.
Then write the single next task. "Offer a 400-row option to 10 more shops with 300+ products" is a next task. "Keep going" isn't.
A worked example
| Field | Value |
|---|---|
| Hypothesis | Small UK shops will request a paid audit after seeing a sample |
| Rule | Continue if ≥ 2 quote requests; change if ≥ 3 replies and 0 requests |
| Result | 10 contacted · 3 replied · 1 meeting · 1 quote request · 0 payments |
| Supporting evidence | Quote request for 400 rows (original attached) |
| Counter-evidence | One shop says its platform already handles this |
| Alternative explanation | The request may be tied to a one-off promotion |
| Decision | Change — test a 400-row tier with 10 new shops of 300+ products |
Reading common phrases
Some phrases appear in almost every test. Here's how to log them.
| What they say | Likely level | How to treat it |
|---|---|---|
| "That's a great idea." | Opinion | Note it; it's not demand |
| "Let me know when it's ready." | Opinion | Polite interest without commitment |
| "Can you send me more information?" | Interest | A reply — follow up with a concrete next step |
| "How much would it be for our size?" | Request | Strong signal — answer with a clear price |
| "Can we start next week?" | Request, close to commitment | Confirm scope and terms in writing |
| "We'll pay once it's done." | Promised payment | Log separately from a completed payment |
| "We already use a tool for this." | Counter-evidence | Ask what they like and dislike about it |
| No reply | Unknown | Count it; don't interpret it alone |
When the numbers are small
With ten or fifteen relevant contacts, a single person can change the picture. That's fine — a first test is a practical decision, not a statistical study. Two habits keep small numbers honest: report the exact counts rather than percentages ("2 of 12", not "17%"), and look at who said what, not only how many. One request from a buyer who matches your profile perfectly can be worth more than three from people outside it.
Common mistakes
- Averaging levels. Mixing compliments and payments into one "interest" number.
- Dropping the denominator. It's the first thing that disappears in summaries.
- Rewriting the rule afterwards. The rule was written to protect you from exactly that.
- Over-reading one customer. One payment is strong evidence for that buyer, not yet for the segment.
How Proofoak helps
Proofoak's evidence ledger records each event with its level, date, channel, person or business and the original words or file. People are deduplicated, promised and completed payments stay separate, and the summary always shows the denominator. The decision report lists supporting and counter-evidence, alternative explanations and one next task, and your choice is saved separately from the recommendation.
Frequently asked questions
How do I know if feedback is positive enough?
Compare it with the decision rule you wrote before the test, not with how the conversation felt.
What counts as strong evidence?
A paid pilot or a signed request at a stated price is stronger than any opinion.
How do I handle conflicting feedback?
Segment it: different buyer types often want different things. Decide which segment you are testing.
Sources and further reading
How we wrote this
Written and reviewed by the Proofoak Research Team on 4 November 2026. Examples are synthetic and labelled. Rules of thumb are our working recommendations, not benchmarks or legal advice. Outreach rules depend on your country and recipient — check them before you contact anyone.