Maintenance
Part of Social proof: methods, tools and useful context
Social proof ethics: responsible use and warning signs
Social proof ethics for anyone publishing a count or a testimonial: what a number has to carry to be honest, and the four ways a true figure still misleads.
Publishing a true number is not automatically honest. The question is what the reader will believe after seeing it, and whether that belief is one you would defend to them.
What to take away
- You are responsible for the inference, not just the figure. A count arranged to produce a false impression is a false statement told in arithmetic.
- Four failures cover nearly everything: the missing denominator, the selected sample, the stale count, and the undisclosed payment.
- The test is whether you would print the method next to the number. If that would spoil it, the number was doing work it had not earned.
What an honest count carries with it
A denominator. "Four hundred customers switched" means one thing beside four hundred and one, and something else beside forty thousand. A numerator alone is not a measurement, and publishing it alone is a choice.
A period. Counts accumulate. A total since launch and a total this quarter are different claims, and the first will keep growing forever regardless of what is happening now.
A method. How were these people counted, and could anyone have been excluded? A review score assembled from everyone who bought and one assembled from everyone the seller invited are not comparable, and the reader cannot tell them apart from the outside.
A disclosure of interest. If a person was paid, given the product, or is connected to the seller, the reader needs to know before they weigh the endorsement. The Federal Trade Commission's endorsement guides work through what counts as a connection and where a disclosure has to sit, and the answers are less permissive than most people assume.
The four failures
| Failure | What is published | What the reader concludes |
|---|---|---|
| Missing denominator | A count with no base | The proportion is large |
| Selected sample | Real quotes, chosen | This is representative |
| Stale count | A true historic total | This is current |
| Undisclosed payment | A genuine endorsement | This was independent |
None of these requires anyone to write a false sentence, which is why they are common and why "we never said anything untrue" is not a defense. Each one leaves the reader holding a belief the publisher knew they would form and knew was wrong.
The selected sample is the one worth dwelling on, because almost every organization does it without noticing. You collect feedback, some of it is good, and you publish the good part. Nothing is fabricated. But the reader is reading the published set as a sample of the whole, and you know it is not, and the gap between those two facts is the entire problem.
The fix, in each case
- For the denominator: publish the base or publish a rate. If the base is embarrassing, that is information the reader wanted.
- For selection: say how the quotes were chosen. "Five of the last twenty responses, unedited" is a sentence that costs you nothing and changes what the quotes are worth.
- For staleness: date the number. A dated figure ages honestly; an undated one lies more every month.
- For payment: disclose it beside the endorsement rather than in a policy page. A disclosure the reader has to go looking for is not one.
Every fix here is disclosure, and that is not a coincidence. Influence that survives the audience learning how it worked is the legitimate kind, which is the standard the whole of social influence basics is built on.
When the honest number is weak
This is where the pressure actually sits. Nobody struggles to be honest about a good figure.
If you have eleven customers, the honest options are to say eleven, to say nothing, or to give a different kind of evidence. The dishonest option is the vague one: "a growing community", "trusted by many", "join others like you". These sentences exist because they cannot be checked, and the reason they cannot be checked is that they are not claims. Where the line sits between a vague boast and a representation a business has to be able to support is covered in plain terms by the Federal Trade Commission's advertising guide for small business.
Saying nothing is underrated. A product page with no crowd numbers and a clear account of what the thing does is not a weaker page; it just stops borrowing authority it does not have. And a small number stated plainly often reads better than a large one stated vaguely, because the reader can tell which one is a measurement.
What the reader is owed when the news is bad
An organization that publishes its good counts and buries its bad ones has a policy, whether or not it has ever been written down. The workable standard is symmetry: if a rise would be published, the fall is published in the same place, at the same size.
This is harder than it sounds and it is the whole of the ethics. A publisher who can point at a case where they reported a number that hurt them has earned the confidence they are asking for, and one who cannot is asking to be trusted on the strength of never having been tested. That asymmetry is exactly the thing readers should look for, and the field guide for spotting it is under manipulation resistance.
The mechanism these obligations attach to, and the five questions a reader should ask of any published figure, are set out under social proof. Where the pressure is about fitting in rather than about evidence, the relevant account is normative influence, and where a crowd is genuinely being used as a source of knowledge, informational influence covers when that is sound.
Common questions
Do I have to publish a denominator if nobody asks for one?
If the number would read differently with the base attached, yes. The test is not whether anyone asked; it is whether you know the omission is doing work.
Is it dishonest to show only positive reviews if they are real?
It is misleading, which is the thing that matters. Real quotes selected for one outcome are a sample presented as a picture, and the fix is to say how they were selected rather than to stop using them.
We pay reviewers but we do not tell them what to write. Is that a connection?
Treat it as one. The reader's question is whether money changed hands, not whether it came with instructions, and the guidance takes the same view.
What should I do about an old number that is still true?
Date it. A figure with a date on it lets the reader decide how much it has aged, and that decision belongs to them rather than to you.
