Testimonials are among the most persuasive assets a company can hold and among the most reliably wasted. The waste rarely happens in production. It happens in the ask.
A customer is emailed a request for a few words about their experience, they write something kind and general, and the result is indistinguishable from copy the marketing team could have written. Which is the problem: a viewer cannot tell it apart either.
The distinction that governs everything
Two things get filmed and called the same thing, and they sit under different rules.
A testimonial
A real customer describing their actual experience, in their own words. Its persuasive power comes entirely from being unprompted and true, and every step you take toward controlling it removes some of that power.
Commissioned content
Somebody producing content to your brief, paid for the work. Legitimate, useful, and a different product — the value is in the demonstration and the volume, not in the endorsement.
The rules follow the distinction. Where a material connection exists between the speaker and the brand — payment, free product, a discount, an ongoing relationship — most markets require that connection to be disclosed clearly and prominently. Paying somebody for their time to sit for an interview is normal and disclosable. Paying for a favourable opinion, or writing the words and presenting them as an unprompted view, is not something disclosure fixes.
We turn down work in this category regularly for exactly that reason, and what we will and will not accept sets out where the line sits and how it is enforced rather than merely stated.
Why most testimonials are unusable
Three causes, and all three are decided before anybody presses record.
The question invited praise. "What do you like about working with us?" produces a compliment. Compliments are not evidence, and a viewer discounts them automatically because they know how the sentence was obtained.
The customer was over-prepared. Sending questions a week ahead produces polished answers with the texture removed. Some preparation helps; a rehearsal does not.
Nobody asked about the difficult parts. A testimonial containing no hesitation, no alternative considered, and no moment of doubt reads as an advertisement, because that is the only kind of story that contains none of those things.
The sentence that makes a testimonial believable is almost always the one where they say what nearly stopped them.
The questions that work
Structure the conversation around what happened rather than how they feel about you.
What was the situation before?
The problem in their words, with the details that made it theirs. This is the part a viewer recognises themselves in, and it is usually the strongest thirty seconds in the finished edit.
What did you try first?
Alternatives considered, including the option of doing nothing. It establishes that a real decision took place rather than a purchase.
What nearly stopped you?
Price, timing, an internal objection, a worry that turned out to be unfounded. This is the credibility question and the one most often cut from an interview guide.
What actually changed?
With a number if one exists. "It saved us time" is worth nothing; "it took four days and now it takes an afternoon" is worth the whole video.
Who would you not recommend this to?
Uncomfortable and enormously effective. An honest limitation makes everything else believable, and it pre-qualifies the audience.
Formats, and what each costs
| Format | Typical cost | Best for |
|---|---|---|
| Self-recorded on their phone | Cost of asking | Volume, authenticity, social placements |
| Remote interview, edited | $500 – $2,500 | B2B, considered purchases, longer cuts |
| On-site filming | $3,000+ | Hero assets, environments that carry meaning |
| Event capture | $1,500 – $5,000 | Several testimonials in one day |
Self-recorded footage looks worse and frequently performs better, particularly in social placements where polish signals advertising. A phone at arm's length in a real room carries more credibility than the same words in a lit studio.
If you go that route, send three instructions and no more: film in landscape or portrait, whichever you need; find somewhere quiet, because bad audio is unwatchable in a way that bad video is not; and answer in full sentences that include the question, since the interviewer will not be in the edit.
Running an interview that produces material
The mechanics of the session decide how much of the hour is usable, and three of them matter more than the rest.
Record the audio separately, or at least deliberately. A quiet room with the window shut beats an expensive microphone in a hard-surfaced office, and a testimonial with unusable audio is unusable entirely — viewers forgive a soft image and abandon bad sound within seconds.
Do not conduct the interview with somebody the customer has a commercial relationship with. An account manager in the room, however friendly, changes what gets said; the difficult answer that makes the whole piece credible does not arrive in front of the person who sold them the thing.
And film for longer than you need. Forty minutes of conversation yields two usable minutes at a good rate, and the strongest thirty seconds is almost never in the first ten. Booking a twenty-minute slot to produce a two-minute film is the most common way an otherwise well-planned session comes up short.
Consent, before rather than after
The release is the part everybody intends to sort out later and occasionally cannot.
Get it signed before filming, and have it state four things: the channels the footage may appear on, the term, the territory, and whether paid advertising is included. Organic use on your own website is a smaller ask than a face in a media buy, and treating them as one agreement is how a good relationship becomes an awkward email eighteen months later.
For business customers there is a second approval nobody remembers. The individual can agree to appear; their employer usually owns whether the company's name can be used, and legal or communications teams frequently have a view. Ask early, because a testimonial you cannot publish is a filmed afternoon.
Who to ask, and when
The timing matters more than the selection.
Ask at a moment of demonstrated success: a renewal, a result they mentioned themselves, a support conversation that went well, the point where onboarding finished and something worked. Requests sent at random points in a relationship get polite refusals; requests sent a week after somebody told you the thing worked get enthusiastic yes.
Choose people whose situation resembles your prospects rather than your most impressive logo. A viewer needs to recognise their own problem, and a household name using your product at a scale nobody else operates at does not provide that.
And ask a specific person rather than a company. Testimonials are made by individuals with a story, and the account manager is not always the one who has it.
Where commissioned creators fit instead
Sometimes the honest answer is that you do not need a testimonial.
If you need volume — thirty short clips demonstrating the product across markets and demographics — that is content production, not endorsement, and it should be briefed and paid as such with disclosure attached. How a campaign brief becomes usable footage covers what to specify, and what it costs to source creators against a brief you control sets out the rates.
If you need somebody demonstrating a use case you have no customer for yet, that is also commissioned content and it must not be dressed as a customer's experience.
The two work well together, and they should be visibly different things: real customers carrying the credibility, commissioned creators carrying the coverage. What a UGC creator actually is covers the distinction from the production side.
Commissioned content buys reach and demonstration. Only a real customer buys belief, and only for as long as the audience trusts that the difference is being observed.
Testing before committing
A testimonial is creative, and it fails in the same ways other creative does — the viewer misses what the product is, or the story does not land outside the market it was filmed in.
Two rounds of comprehension checks on the rough cut catch most of it, and the sample needed is small because the failures are large. What human pre-testing answers that performance data cannot covers the question set, and the useful version here asks what the viewer thought the product was, what the customer's problem had been, and whether they believed it.
That last question is the one worth running before a testimonial goes anywhere near a media budget. Disbelief in a testimonial is fatal in a way that disbelief in an advertisement is not, because the whole asset was the claim to be genuine.
What to avoid
Four things, each common enough to name.
Scripted lines. A customer reading your copy is an actor without the training, and viewers detect it immediately.
Undisclosed incentives. Covered above and worth repeating, because it is the one that creates liability rather than merely disappointment.
Stock footage of people who were not there. Reconstructions of customer environments, actors standing in for real users, footage licensed to represent a testimonial that occurred elsewhere — all of it undermines the only thing the asset had.
And composites. Assembling several customers' words into one narrative and presenting it as a person's account is fabrication regardless of how accurate the composite is.
For anything filmed by people you did not hire directly, the evidence standard matters as much as the consent: who filmed it, when, and against what brief. What counts as evidence and how to specify it up front covers the formats, and how a brief becomes reserved capacity and verified footage describes the mechanism where the people are distributed rather than in your office.