Customer quotes almost never arrive ready to publish. They contain filler words, half-finished clauses, the wrong tense, the wrong subject-verb agreement, and the kind of conversational drift that reads well in a transcript and badly on a landing page. The editing decision is harder than it looks: edit too lightly and the quote feels unprofessional, edit too heavily and the quote loses the voice that made it credible. This piece is a practical editing protocol for testimonial copy — what to fix, what to leave alone, and how to make the editing decisions defensible if the customer asks why the published quote looks different from the recording.
The underlying principle is that editing a testimonial is not the same as editing a press release or a feature paragraph. The reader of a testimonial is doing a credibility check, not a comprehension check. The reader is implicitly asking "does this sound like a real person who actually uses the product, or does this sound like the vendor wrote it." Every editing choice either strengthens or weakens the answer to that question.
What customers expect to be cleaned up
There is a baseline cleanup that customers and readers both expect. When a customer agrees to a published testimonial, the customer is implicitly agreeing that filler words, false starts, and verbal tics will be removed. Nobody wants their published quote to read "um, yeah, so, like, the thing is, the dashboard, the dashboard, it really helped us." Removing this layer is not voice manipulation. It is the difference between a transcript and a quote.
The baseline cleanup safely covers: filler words (um, uh, like, you know, I mean), false starts and self-corrections ("the dashboard — well, the analytics dashboard"), repeated phrases when the customer was searching for the right wording, and audible breath markers from the recording. Standard transcription cleanup removes these by default, and customers almost always sign off on the cleaned version without complaint.
Grammar correction sits at the next layer up and is where the protocol matters. A customer says "me and my team was working on the migration." The grammar fix is "my team and I were working on the migration." Three things changed: pronoun case (me → I), pronoun order (me and my team → my team and I), and subject-verb agreement (was → were). All three are defensible corrections, but together they make the customer sound like a different person. The protocol question is which corrections preserve the customer's voice and which corrections replace it.
The voice-preservation hierarchy
Order the corrections by how much they alter voice, and apply only the ones that survive the test "does this still sound like the customer." From least invasive to most invasive:
First, subject-verb agreement and clear grammatical errors that distract the reader. If the customer says "the metrics was clear," changing to "the metrics were clear" does not alter voice. Readers register agreement errors as transcription noise and not as the customer's intent.
Second, pronoun case errors that the customer would self-correct in writing. "Me and the team" becoming "the team and I" is a defensible correction because most customers, given the chance to review their own published quote, would make the same correction themselves. Pronoun case is one of the cleanups customers thank you for.
Third, tense consistency within a single quote. If the customer drifts between past and present tense mid-sentence — "we were behind on the rollout, and the team is struggling with the timeline" — pick one tense and apply it consistently. Tense drift is a recording artifact that does not survive on the page.
Fourth, sentence-fragment repair where the fragment confuses the reader. "The dashboard. Game changer." can stay if the rhythm is intentional and the customer is making a stylistic point. "The dashboard. Because the team needed something." cannot stay because the fragment is the customer searching for the next clause and the reader does not have the audio context to fill in the gap.
Fifth — and this is the line — word substitution that changes the customer's vocabulary register. Replacing "stuff" with "functionality" or "thing" with "capability" is voice replacement, not editing. This is where vendor copy bleeds into customer quotes and makes the testimonial sound like marketing wrote it. Hold the line here. If the customer said "stuff," the published quote says "stuff."
Punctuation choices change how the quote reads
Punctuation is where most editors over-edit without realizing they are doing it. The reader reads punctuation as rhythm, and the rhythm signals voice. A long sentence with one period reads as confident. The same sentence broken into three sentences with periods reads as hesitant. Comma splices read as casual. Em dashes read as deliberate. Semicolons read as formal. Pick the rhythm that matches the customer's actual cadence, not the rhythm that matches your house style guide.
Em dashes are the most useful punctuation for testimonial editing because they preserve the spoken cadence where a customer interrupts themselves to add a qualification. "The migration was fast — we were live in three weeks — and the team adopted it without much training." The em dashes here are doing the work that the customer's vocal pause did in the recording. Replacing them with commas flattens the rhythm. Replacing them with parentheses makes the qualification feel like an afterthought rather than an integrated point.
Comma splices ("the dashboard was clear, we knew what to do") are technically wrong but often read better in a quote than the technically correct alternatives (period, semicolon, or coordinating conjunction). The comma splice signals that the two clauses are tightly bound — the second clause is the consequence of the first. A period would suggest the two clauses are independent thoughts, which is not what the customer meant. Use the comma splice if the customer's cadence supports it.
Ellipses are dangerous. An ellipsis in a published testimonial signals to the reader that content was removed. If you used an ellipsis to mark a trimmed verbal tic, the reader does not know that — the reader assumes you removed something substantive and starts looking for what was cut. Remove the verbal tic without the ellipsis. Reserve ellipses for cases where you genuinely trimmed substantive content and want to be transparent about the cut.
Quotation marks within the quote (the customer quoting somebody else) often arrive without clean attribution. "And my CFO said this is the cheapest line item we have." Convert the embedded quote to indirect speech ("my CFO told me it was the cheapest line item we had") or to a clean direct quote with attribution ("my CFO said, 'this is the cheapest line item we have'"). The half-quoted state is a transcription artifact and does not survive on the page.
The contraction question
Contraction handling is the editing decision that signals voice fastest. "We are happy with the rollout" reads as a vendor statement. "We're happy with the rollout" reads as a customer statement. The difference is one apostrophe and it changes the entire credibility frame.
The default for testimonial editing is to preserve the customer's contractions exactly as the customer spoke them. If the customer said "we're," the published quote says "we're." If the customer said "we are," the published quote says "we are." Editors are trained to expand contractions in formal copy and have to consciously override that instinct in testimonial copy. The customer's spoken cadence is part of the credibility signal, and contractions are how spoken cadence appears on the page.
The exception is when the contraction creates ambiguity. "It's been a year" is fine. "Its been a year" is a typo, and a published testimonial with a typo damages the testimonial's credibility. Fix the typo. The exception is narrow — apostrophe-related typos and homophone confusion (its/it's, your/you're, their/they're). These are not voice corrections. They are typo corrections, and customers expect them.
The protocol in practice
The editing pass moves through the testimonial in this order. First, remove filler words and false starts. Second, fix subject-verb agreement and pronoun case where the correction is invisible. Third, repair tense consistency within each quote. Fourth, evaluate punctuation against the customer's actual cadence — preserve em dashes, comma splices, and contractions where they carry voice. Fifth, fix typos that damage credibility. Sixth, leave word choice alone.
Then the protocol adds one more pass that most editors skip: read the edited quote aloud against the audio. If the edited quote reads aloud in a different cadence from the customer's actual cadence in the recording, the editing has crossed the line from cleanup to voice replacement. Walk back the changes until the cadences match.
For related coverage of testimonial editorial standards, see testimonial character count best practices and testimonial anonymization guidelines.
Customer review and the "do I sound like myself" test
The final test before publishing is the customer review. Send the customer the edited quote and ask one specific question: "does this still sound like you." Do not ask "is this accurate" or "are you happy with this." Both questions get yes answers from customers who are not actually reading the quote against their own voice. The "sound like you" question forces the customer to do the voice check that the reader will do.
When customers come back with edits, customers almost always make the quote longer and looser, not shorter and tighter. Customers add qualifications, add context, and restore filler-style hedges ("kind of," "pretty much," "for the most part"). Resist the instinct to strip the hedges out again. The hedges are the customer's voice, and the customer reinstated them deliberately because the cleaner version did not sound like them anymore.
Document the protocol in the editorial style guide so that the next editor on the next testimonial applies the same hierarchy. The risk in testimonial editing is not that one editor over-edits one quote. The risk is that the house style drifts toward over-edited quotes across the whole testimonial library, and the cumulative voice flattening makes every customer sound the same. The protocol prevents that drift.
Why this matters for conversion
Testimonial conversion impact is driven by the reader's credibility check, not by the reader's comprehension check. Over-edited testimonials read smoothly and convert poorly because the reader's credibility check fails — the quotes do not sound like real customers, they sound like the vendor's marketing voice. Under-edited testimonials read poorly and also convert poorly because the reader's comprehension fails before the credibility check even runs.
The editing protocol's goal is a published testimonial that passes both checks. The reader understands the quote on first read, and the reader believes a real customer wrote it. The hierarchy described above is calibrated to keep both checks passing. The protocol is boring and the editing decisions are small, but the cumulative effect on testimonial conversion is real and measurable.