Applitools vs Percy

Both attach to an existing test suite, so unlike a Storybook-based tool neither forces you to change how you work. The real differences are the comparison engine, the billing unit, and how much pricing information you get before you commit.

At a glance

The comparison engine is the real difference

Pixel comparison with a tolerance forces a trade you cannot win: tight enough to catch a subtle layout break is tight enough to fail on antialiasing, and loose enough to be quiet is loose enough to miss real regressions. Every threshold you set is a position on that spectrum.

Perceptual comparison sidesteps it by modelling what a human would actually notice. Applitools sets match levels in the SDK rather than per-project thresholds, and groups similar differences across browsers and pages into one review — so a single approval updates every affected baseline. On a wide matrix that is the difference between reviewing forty diffs and reviewing one.

Percy runs pixel comparison with its own Visual AI engine layered on and DOM snapshotting underneath. The DOM approach helps with a different problem — responsive re-rendering — rather than with noise.

The checkpoint trap

This is the single most expensive misunderstanding in an Applitools evaluation, and it is not hidden — it is just easy to skim past.

The Starter tier at $667/month billed annually includes 100,000 component checkpoints or 1,000 page checkpoints. Those are not close to the same budget.

A team with 20 routes at 2 viewports uses 40 page checkpoints per run. That is 25 runs a month before the Starter allowance is gone — roughly one pull request a day. If you are attaching Applitools to an end-to-end suite that drives real routes, model against the 1,000 and not the 100,000, and do it before signing an annual commitment.

Percy bills in screenshots with no such split, which is simpler — but the rate is not published, so you cannot do the equivalent arithmetic at all.

Pricing transparency

Applitools is a hybrid: the Starter price is published, Professional and Enterprise are contact-sales, starting from 150 and 350 active pages respectively. So you can at least model the entry point, which you currently cannot with Percy.

This is a planning fact rather than a product criticism. Unpublished pricing generally signals negotiated, enterprise-shaped deals — fine if you have procurement, awkward if you are a small team trying to forecast.

Cross-browser: Ultrafast Grid is the real feature

Applitools' strongest differentiator after the comparison engine. You run your test once; the grid renders that single capture across Chrome, Safari, Firefox and Edge in the cloud. You do not maintain browser infrastructure and you do not re-execute the suite per engine.

The cost consequence is direct, though: each browser is a separate checkpoint. A four-browser matrix multiplies your consumption fourfold, against that 1,000 page-checkpoint allowance. The savings are in CI time and infrastructure, not in the bill.

Whether you need four engines at all is worth questioning first — Chrome, Edge, Opera and Brave all run Chromium, so the genuinely divergent set is three. See cross browser visual testing tools.

Operational gotchas

One from Percy's documentation that will cost you an afternoon the first time it happens:

The axis neither competes on

Both require an SDK call for every state you want covered. Applitools needs eyes.check(); Percy needs percySnapshot(). In both cases a page nobody wrote a test for is invisible, and every new route is a new line of test code somebody has to remember.

That maintenance — not the licence, not the comparison engine — is what usually ends a visual testing programme. No comparison page in this category treats it as a dimension, including the vendors' own. It is also what we built VisualMax to remove, so discount the framing accordingly; the test is simple enough to run yourself. Add a new route, change nothing else, and see whether either tool notices. If neither does and that matters, a whole category of tools now generates its own coverage.