How to Compare ID Verification Vendors on Your Own Traffic

A demo shows a vendor at its best. Your own customers show it at its worst, which is what you need to know. A vendor-neutral way to run a side-by-side trial, size it honestly and decide on evidence.

Read

7 min read

Published On

Oct 21, 2026

Read: 7 minutes

Last updated: October 21, 2026

In this guide

Whether you can run a vendor beside your current one is the gating question
How many sessions you need before a comparison means anything
Why agreement is not a score, and what a mostly genuine sample cannot tell you
A scorecard to hold every vendor to, and when not to switch

Compare on Your Own Customers, Not on a Demo

Your current provider works. Your team knows its console and your compliance people have signed off on it. Replacing it is a project, and a new vendor that misbehaves costs you customers you cannot see. That is a good reason not to switch on a promise. It is not a reason never to look.

The most reliable evidence is how a vendor handles your real customers, on the same sessions, at the same time. This guide is about getting that evidence without changing what your customers experience. It applies to any vendor.

Ask the Gating Question First

A side-by-side trial needs the candidate to see the same sessions your current vendor sees, and needs the per-check results to compare. Whether that is possible depends on your current provider's contract and tooling. Ask in writing before you plan anything else, and treat a refusal to discuss it as information.

How a Side-by-Side Trial Works

Your current vendor keeps making every decision, exactly as it does today.
The candidate runs its own checks on the same sessions and records what it would have decided.
Nothing changes for your customers. If the candidate errors, it logs the error and stops. It never delays or blocks the live decision.
Your consent and privacy notices must cover any processing that happens. Ask your counsel before starting.
You compare the two decisions session by session, then read the differences.

How Many Sessions You Need

Volume decides whether a comparison can tell you anything. Suppose you verify about 400 people a month, and assume, as an example, that a vendor wrongly rejects 2 to 5 percent of genuine customers. That is 8 to 20 events a month. Two vendors whose true rates differ a little will produce overlapping counts on that few events. You cannot rank them on one month.

Plan for several months at that volume, or a larger sample, and read the cases instead of the percentages. Decide the length of the trial and the question it answers before you look at any results.

Read the Disagreements

Agreement is not a score to maximise. A disagreement can mean either side is wrong, so open the cases.

Question	What it tells you
Where does the candidate flag something your provider cleared?	Either a miss on your side or a false flag on theirs. Look at the images.
Where does your provider flag something the candidate cleared?	Either the candidate would let through a case you would not, or your provider is stricter than it needs to be. This is the one that matters most for risk.
Where do both agree?	Most sessions. Agreement on easy cases tells you little. The disagreements are the data.

One caution applies to every comparison like this. Real customer sessions are overwhelmingly genuine people. A mostly genuine sample measures agreement and false flags well. It does not measure how many fraud attempts a vendor would catch, because there are few or none in it. Test that separately, with tampered documents the vendor can supply or a public test set, and do not read agreement as proof against fraud.

A Scorecard for Every Vendor

Write the measures down before you see results, and ask each vendor the same questions on the same sample.

Measure	Ask every vendor
Completion rate by device	Where do phone and desktop customers drop off, and what do you do about the handoff?
Share sent to a person	Who decides when a check is unsure: a rule, a model or a human? Can a model decline on its own?
Time to a final decision	What does a customer wait through, and what do they see while they wait?
Clarity of the reason given	Does a failed customer learn what to fix, or see a generic message?
Cost, including white label	What does each verification cost, and is your own branding included or extra?
Data handling	Where is data processed and kept, how long, and what does the consent tell the customer?
Effort to integrate and to leave	Can you keep the results, erase on request and export your verified customers?

If a vendor will not say how it handles unsure cases, treat that as an answer. Do not repeat claims about other vendors that you have not checked against their own published terms. They change.

What Changes Besides the Checker

Question	Why it matters
Whose brand is on every screen?	A vendor's logo in your flow is a trust and conversion question
Who decides a failure?	It shapes the pending states your product must show
Who holds the result?	It decides what you can reuse for the next application

When Not to Switch

Sometimes the right answer is to stay. If your provider performs well on your customers, your reviewers are content and your compliance process is settled, the trial may simply confirm that. That is a useful result.

There are also reasons to switch that are not about accuracy: your own brand on every screen, a person deciding every unsure case, the result held in your records, or one record shared with a funding partner. Weigh them beside the numbers, not instead of them. Find the Step Where Applicants Stall shows how to see where honest customers leave. Embedded Finance Companies: What to Ask has a similar checklist for capital partners.

Disclosure: Slate makes an identity verification product, so weigh this guide accordingly.

Common Questions

Will a side-by-side trial slow my current flow?

It should not. A well-built trial never changes, delays or blocks the live decision, and logs and stops on its own errors. Confirm that before you start.

Does it work with any provider?

It depends on your current provider's terms and tooling, which is why the gating question comes first.

Does my customer know?

Your customer goes through the same flow as before, but your notices must still cover any processing that happens. Ask your counsel.

Do I have to switch afterwards?

No. You can stop after the comparison and keep what you have.