The trust signal hierarchy

Trust & Social Proof is weighted 15% of the overall roast.page score. It also has the largest gap between the median page and the top quartile of any dimension: median 4.2/10, top 25% averages 7.1. That’s a 2.9-point dimension gap — large enough to swing the overall score by 4–5 points before any other dimension changes.

The dimension is wide because 38% of the pages we sampled have no testimonials at all. No customer logos. No counts. No reviews. A founder photo and a vibe. Those pages are gambling that the H1 alone will earn the click. Usually it doesn’t.

The pages that do have trust signals vary wildly in how effectively they use them. So we tagged 1,200 pages for 11 distinct signal types — counts, logos, testimonials (named and anonymous), star ratings, review counts, case studies, press mentions, awards, certifications, and founder photos — and ran bivariate correlations against the page’s Trust dimension score and overall score. The ranking is below. Two of the rankings are surprising.

The most common trust signals are also the least effective. The most effective ones are also the most under-used.

Three things stand out. First, the most common signal (logo bar, 51% of pages) is mid-pack on effectiveness. Second, the second-most common pattern (anonymous testimonials with stock photos, 23%) correlates negatively with score. Third, the top-ranked signals (customer counts, named testimonials, case studies) are used by minorities of pages — leaving most pages competing on the dimensions that don’t move the needle.

Logo bars work via name recognition. If a visitor sees a logo they recognize, the implicit thought is: “That company wouldn’t use this if it didn’t work.” That’s a real effect — pages with recognizable logos score 4.7/10 on Trust versus 3.9/10 for pages without. The 0.8-point lift is meaningful.

The problem is that recognition is unevenly distributed. A SaaS founder selling to engineering managers gains a lot from a GitHub logo and not much from a logo their target buyer doesn’t recognize. A creator-economy tool gains from a MrBeast endorsement and not much from showing a Fortune 500 client. The logo bar works for a subset of visitors, but it’s passive — the visitor has to recognize the logos for the signal to land.

A quantified count is active. “20,000 teams use us” works without requiring the visitor to recognize any specific team. The number itself is the proof. If the cohort is big enough to need a comma, the implicit thought becomes: “Lots of people figured this out before me; I’m not the first one to take this risk.” That’s a different psychological lever, and our data suggests it’s the stronger one.

What if my customer count is small?

State it anyway, with framing that turns small into specific. “Used by 47 founders since January” communicates recency, growth, and the kind of customer (founders). It’s honest in a way that “trusted by industry leaders” isn’t. We’ve watched early-stage pages score meaningfully higher on Trust with this exact pattern than with a generic “trusted by” bar of three logos.

23% of the pages with any testimonials in our sample have the same identifiable pattern: a quote attributed to a first name and a job title (“Sarah, Marketing Director”), paired with a stock photo. Reverse image search confirms the photo is from a stock library on 87% of these. The pattern is so common, and so transparent, that experienced buyers register it as a fabrication signal within seconds.

Pages with this pattern averaged 3.6/10 on the Trust dimension. Pages with no testimonials at all averaged 4.4/10. Adding fake testimonials made the pages worse than honestly empty ones. The correlation with overall page score is mildly negative (r = −0.12) — small but real.

An empty trust section is honest. A fake one is a tell. Visitors notice. Our data does too.

The repair is straightforward: delete the section, or replace it with something verifiable. If you have one customer willing to be quoted by full name with a real photo, that’s worth more than six anonymous testimonials. If you don’t have any yet, use the space for something else — a specific founder note, a quantified count of trials or signups, a link to your changelog or roadmap. Anything verifiable beats anything generic.

How to source real testimonials when you’re early-stage

The pattern that consistently works: email three customers, ask for a two-sentence quote about a specific outcome they had, ask for a real photo (LinkedIn headshot is fine), confirm their full name and company. Most early customers will say yes. The exercise is more uncomfortable than fabricating, but the resulting page scores higher and stays honest.

We tagged each trust signal’s placement: above the fold, mid-page, or footer. Then we looked at the relationship between placement count and Trust dimension score.

The two-position pattern works because it addresses different visitor questions. Above the fold, the visitor scans for “do these people exist?” — a logo bar or quantified count answers that quickly. Before the CTA, the visitor scans for “are they really going to convert me?” — named testimonials or case study links carry more weight in this second moment because the visitor is closer to action.

The strongest pattern we observed: a quantified count above the fold (“Trusted by 12,847 teams”), then 2–3 named testimonials with company affiliations just before the pricing or signup section. Pages following this exact pattern scored 7.1/10 on Trust — the same level as pages we’ve individually critiqued as standout examples (Stripe, Linear, Vercel).

Most landing pages can move their Trust dimension score by 1.5–2.5 points in one focused session. That maps to a 3–5 point overall page swing — meaningful, but more importantly, the trust block is the part of your page that builds compounding confidence as visitors return. Do this once and you don’t have to do it again until you’ve grown.

1. Add a quantified count above the fold. Use whatever number you can defend: customers, signups, trials, downloads, books shipped, projects completed. Phrase it as “{number} {cohort} {action}” — “47 founders building with us since January” works at any scale.
2. Replace anonymous testimonials with named ones. If your only options are anonymous, delete the section entirely. Email three current customers this week. Ask for a two-sentence quote and a real photo. Most will say yes.
3. Add review counts to any star ratings. “4.8/5” alone is weak. “4.8 from 1,200 reviews” is strong. Pull the count from G2, Capterra, the App Store, your CRM — wherever you have it.
4. Add one case study link with a metric. “See how Acme cut churn 47%” works even if the link goes to a one-paragraph case study. The metric in the link text does most of the trust work.
5. Audit the logos. Every logo on your page should be a current paying customer. Remove any that are integration partners, mentions in press, or one-off conversations. 32% of pages in our sample fail this check.
6. Place trust signals in two positions. Above the fold (logos or count) AND before the primary CTA (testimonials or case studies). 18% of pages do this. Those pages average 6.7/10 on Trust. Most pages don’t.

Run a fresh score afterward with the social proof audit tool or a full page analysis. Watch the Trust dimension move. It will. This is the most reliable single-session lift we’ve seen.