Trust Center

The 0–100 Trust Index

The Trust Index is a single composite score per service, calculated from five components with fixed weights. Weights sum to 1.0 and changes to the formula are versioned. The current formula is v2026.1.

If the rating component is unavailable (not enough weighted reviews), its 40% is reallocated proportionally across the remaining four components so the score stays on a 0–100 scale. We never artificially inflate to 100 when data is missing. If the signal floor isn't met, we show "Not yet rated" instead.

Four review tiers

Community reviews are classified by the strongest verification signal present at submission. Weights are spread by roughly an order of magnitude so evidence-backed reviews dominate the aggregate and anonymous ones cannot swing it.

Flagged reviews are floored to 0.1×. Automated fraud detection (burst, duplicate, scattergun signals) drops any review to near-invisible weight until moderation clears it. Flagged reviews are still displayed (we don't hide disputed input); they just don't meaningfully move the number.

What counts as evidence

For a review to reach tier-4 (evidence-backed), the reviewer must complete a read-only OAuth connection to one of the following providers. The evidence then enters a verification queue as pending; the review is promoted to this tier only after an admin marks that evidence verified:

• Meta Ads (Facebook & Instagram)
• Google Ads
• TikTok Ads
• YouTube Analytics
• X (Twitter) Ads
• LinkedIn Ads

Connections are read-only. We never post, modify, or spend on your behalf. Connection can be revoked at any time; after revocation we retain only the aggregated metrics that backed the review, not the underlying account data.

Claiming a profile

Business owners can claim their profile to respond to reviews, correct factual errors, and gain the Claimed badge. Verification uses one of three methods, all standard across the industry:

1. Email at domain. Click a confirmation link sent to an address at the business's domain (e.g. owner@yourdomain.com).
2. DNS TXT record. Publish a DNS TXT record at the business's domain with our challenge value.
3. HTML file. Upload a file at https://yourdomain.com/.well-known/howsociable-verify.txt.

Once verified, an admin reviews the claim, usually within one business day, and either approves it or contacts the claimant for additional evidence. A business starts a claim from its own listing on our review pages: an unclaimed listing links straight into the claim form.

Reporting a service

If you believe a listed service is fraudulent, has fabricated reviews, or is misleading in a material way, you can file a report. We read every report and reply within three business days. Upheld reports affect the Trust Index; dismissed reports do not. Filing a report does not remove a listing; removal only follows a factual or legal finding.

Aggregation rules

Aggregate ratings are only published when weighted volume clears a threshold of 3 effective reviews. Until then, we display the raw reviews we have, and where the aggregate would otherwise be we show the plain average with a note that the weighted rating is unavailable.

We do not Bayesian-smooth toward a global prior. That's the right statistical move for a site averaging over tens of thousands of services, but here the reader-facing promise is "reviews you can verify". A smoothed rating borrows information from unrelated products and undermines that claim.

Limits of the data

No scoring system eliminates gaming risk entirely. We publish our aggregation rules specifically so attackers know the tradeoffs, and legitimate reviewers know why their submission matters. Known limits:

• A brand-new service with no reviews scores "Not yet rated". Absence of bad news is not evidence of good service.
• A coordinated client-invite campaign can move the aggregate legitimately. We publish the tier breakdown on every profile so readers can judge the mix.
• Trust Index values are not snapshotted or averaged over time. Each score is computed from the inputs present when it is calculated, so a new review, a claim, or a new report moves the number instead of being smoothed out.

Change log

• v2026.1: initial published formula. Five components, weights 40/15/20/15/10, signal floor of 3 distinct signal categories.