Testimonials are the highest-leverage copy on your landing page. Most SaaS founders ship the worst three they have.
The problem we ran into
We were running a normal SaaS landing page. Twelve testimonials. Pretty headshots. Real customers.
Conversion was flat.
When we A/B tested swapping the testimonial block in and out, the lift was within margin of error. We started reading the actual reviews. Two of them were great. The other ten were variations of "the product is good and the team is friendly."
Polite. True. Useless.
We were doing what every SaaS founder does: dumping every review we could collect into a section and hoping volume = trust. It doesn't.
What we shipped
trustfolio — a testimonial platform that collects reviews from your customers, scores each one on a 0-100 conviction scale using an LLM, and embeds the highest-converting ones on your site via a widget that ships in under 5KB.
Three things make this different from the spreadsheet you're using now:
1. A conviction score, not a star rating
Five stars tells you a customer didn't hate you. It doesn't tell you whether the review will convert anyone.
We score every incoming testimonial against four axes:
- Specificity — does it name a measurable outcome ("cut onboarding from 3 weeks to 4 days") or a vague feeling ("really helpful")?
- Objection-handling — does it dismantle a doubt a prospect would have?
- Identifiability — does the reader picture themselves in the customer's shoes?
- Tension and resolution — does it tell a story, or is it a sentence?
A 92-conviction testimonial belongs on your hero. A 41 belongs in the archive.
2. Smart highlights instead of full quotes
A 180-word testimonial buries the line you want. We surface the 12 words that actually do the work, link them to the full quote on click, and serve the rest only if the reader keeps reading.
This is one of the largest lifts we see in customer A/B tests: average widget click-throughs went up 2.3x when we replaced full quotes with smart highlights.
3. A 5KB widget that respects your Lighthouse score
Most testimonial widgets are 60-200KB of React + dependencies. They tank LCP and tag along third-party fetches.
Ours ships as a single self-contained ESM bundle: 4.8KB gzipped, zero runtime deps, lazy-loads the rest only when the widget enters the viewport.
<script async src="https://cdn.trustfolio.dev/v1/widget.js"
data-project="your-slug"
data-layout="card-row"></script>
That's the whole install. Six layouts available, all under the same size budget.
What we learned building this
A few things were not obvious until we shipped it:
Conviction scoring needs anchoring examples. The first version of our scorer drifted: the same review would get a 71 one day and an 86 the next. We fixed it by anchoring each score to a fixed reference set of customer-graded examples. Drift dropped to single digits.
Customers will give you better reviews if you ask better questions. The default "leave a review" prompt produces "great product." A 3-question collection form (situation → friction → outcome) produces reviews that score 30+ points higher on conviction, on average.
Video is hard, but solvable. We chose to do auto-transcription + AI-pulled timestamps for the moment in a video where the customer says the thing. Video testimonials with auto-pulled highlight clips outperformed text on our own funnel.
Pricing
- Free — 15 testimonials, one widget layout, basic AI analysis. Good for landing pages with light traffic.
- Pro — $29/mo — unlimited text, 50 video/mo, all 6 layouts, full AI features.
- Business — $79/mo — unlimited projects + video, A/B testing, API, white-label.
No credit card on free.
Try it
If you have a landing page and ten or more existing testimonials, drop them into trustfolio in 5 minutes and see what your highest-conviction quote actually is. Most founders are surprised.
trustfolio.dev — Show customer reviews, convert 34% more.
If you ship this on a real page, I'd love to hear about your lift in the comments. We're collecting before/after numbers across the launch cohort.
Top comments (0)