A business verdict for every feature your coding agents ship: did it move demos booked, trials started, forms sent? Your agents ask, you decide. It works the same for the changes your team makes, through Google Tag Manager.
You start with a demo project full of sample results.
Works with Google Tag Manager. Don't have it, or someone else manages it? We'll show you how to bring them in.
Needlr is a release impact tool. It tells product teams whether a release moved a business KPI, with split tests, before vs after checks on past releases and feature adoption, set up through Google Tag Manager.
Claude Code, Cursor or any agent that speaks MCP connects in one line. Every feature it ships gets a business verdict.
Your agent asks what's measurable: which KPI, how much traffic, how long a verdict takes. With little traffic it hears so, and goes for a bigger bet instead of guessing.
It asks “did that work?”. For a release that's already live, Needlr answers from up to 13 months of your own data, often right away. Or it runs a split test in code.
Keep, roll back or try again, with the numbers. Your agent can't peek before the verdict, can't publish to your site, and a test that clearly hurts stops by itself.
> Did the shorter signup form I shipped on Oct 3 help?
needlr_validate · released 2026-10-03 · KPI signup_completed
Improved: signups went from 3.1% to 3.8% of visitors
(+0.2 to +1.2 points), likely thanks to the release.
Recommended: keep. Recorded in Needlr.No developer tickets, no tracking plans, no spreadsheets. Setup takes about two minutes.
Open your site, click the thing you want to change, and edit it right there: text, colour, size, position. Or tell Needlr about a release that is already live.
Pick the KPI it should move: a click on “Book a demo”, a visit to /welcome, a form sent. Test it live and watch Needlr catch it in real time. Already track it in Google Analytics? Use those events as KPIs for changes released to everyone.
Choose who sees it and how sure you need to be. One button publishes to Tag Manager and splits your visitors. The verdict comes once the planned sample and a full week are in, and it never flips after.
Most tools always show a winner. Needlr tells you when something worked, when it hurt, when it made no difference, and when it is simply too early to know.
The new button text brought more demo bookings. Keep it.
The popup cost you trials. Good thing you only showed it to half.
No real difference. Spend your energy somewhere else.
Leaning positive, but not enough visitors yet to call it.
Just started. Needlr won't guess before the numbers are in.
Released to everyone. Compared before and after the release date.
→ Booked a demo · +12% (range +4% to +20%)
Every visitor gets version B.
→ Started a free trial · +9% (range +2% to +16%)
→ Sent the contact form · +6% (range +1% to +11%)
Example
Needlr's verdicts are checked two ways: simulated checks where the truth is known, and a test site with effects we planted on purpose.
simulated checks where we know the truth, run through the same code that decides yours.
of checks with no real difference showed one at 95% confidence. The promise is 5%.
It found the planted effects it has finished measuring so far. Some lab checks are still running. Lab validation, 2026-10-08.
Everything an experiment platform does, without the engineering project.
Your site opens with Needlr's editor on top. Click any element and change its text, colour, size or position, add a banner, or hide something. No code, no developer.
Visitors never see the old version swap for the new one. Views Needlr couldn't change in time are left alone and simply not counted.
Needlr only touches the tags it created, publishes with your approval, and removes everything when you delete a project.
Released a feature to everyone? Count how many people actually use it, against a target you set, like 20% of active people within 30 days.
No personal data in events. Needlr respects your consent banner through Google Consent Mode. Raw events are removed after 90 days.
A KPI that stops firing gets a tag wherever it shows, and you get an email when it stops and when it's back. Its results wait in the meantime. A split that comes out uneven holds the verdict. A burst of traffic from one source is left out of the numbers. The check page always says why.
Not everything can be split in half. For features released to everyone, Needlr compares the weeks before and after the release date, and checks a fake release date too, so a trend that was already happening isn't mistaken for your win. Or count how many people actually use it, against a target you set.
Create an account and explore a demo project with seven real-looking results. Connect your own site when you're ready.