flags & experiments for teams without a data teamflags & experiments, no data team
Flags and experiments for teams with a database and no data team.
Every number is safe to read the day you look at it. No warehouse, no data engineer, no “chance to win.”Every number is safe to read the day you look at it. No warehouse, no data engineer.
billing.
- 01
The interval you can read every morning.
Most p-values are only true the one time you were supposed to look. Ours is valid at every look — that’s what “sequential” means, and it costs a wider interval, not a lie.
- 02
Locked before it starts.
A hypothesis, a metric, a decision rule — written down before there’s data to bend them. You can’t edit an experiment’s design once it’s running; you can only stop it.
- 03
Reads the events you send. Doesn’t ask for a warehouse.
Send events from the SDK and Quoin computes the metric from them — nothing leaves your infrastructure that you didn’t choose to send.
how it works
- 01
Define a flag in code.
One line: defineFlags({ 'checkout.one_click': flag.boolean({...}) }). Ship it dark.
- 02
Set rules — rollout, force, or start an experiment.
Roll out from the web app or attach an experiment — no deploy, no waiting on CI.
- 03
Ask if it worked.
Define a metric from the events you send, run an experiment, get a report that says “clear,” “not yet,” or “broken” — never a bare p-value.
day 1 · day 23 · day 42
The same sentence, checked again.
Day 1: not enough data yet. Day 23: not yet — the interval still crosses zero. Day 42: clear, +21.6% [+5.2, +37.9]. Same test, same rule, more data.Day 1: not enough data yet. Day 23: not yet. Day 42: clear. Same test, same rule, more data.
- day 1
converted.14d — too early to tell.
- day 23
converted.14d +14.2% [−6.5, +38.0] — not yet.
- day 42
converted.14d +21.6% [+5.2, +37.9] — clear.
No chart of confidence shrinking over time. Just the same honest sentence, checked again.
Runs inside the boundaries you already trust.
- sequential interval
Every interval is valid to read at any moment, not just at a planned end date — so checking early never inflates a false positive.
- undoable writes
Every rule change is a revision with an id and a note, and any of them can be undone.
- no warehouse
Runs on events you already send. No export job, no nightly sync, nothing to stand up first.
pricing
Free to start. Talk to us when you need more.
Starter
$0
- Flags
- up to 5
- Running experiments
- 1 at a time
- Tracked exposures a month
- 2,000
Team
Price not published yet
- Flags
- unlimited
- Running experiments
- unlimited
Scale
Price not published yet
- Flags
- unlimited
- Running experiments
- unlimited
- Talk to us
Enterprise
Custom
FAQ
Questions, answered plainly.
- Do I need a data warehouse?
- No. Quoin computes a metric from the events you send it from the SDK. Nothing is copied into a warehouse you don’t already have.
- What’s the difference between a rollout and an experiment?
- A rollout ships a percentage of traffic and watches guardrails — informative, not proof. An experiment locks a hypothesis and a metric before it starts and gives you an interval you can act on.
- Why a sequential interval instead of a p-value?
- Because you’re going to look before the planned end — everyone does. A p-value is only honest at one look; ours stays honest at every one, for a somewhat wider interval.
- What happens if I check a running experiment every day?
- Nothing bad. That’s what “sequential” buys you — the interval is built to stay valid at every look, not just the last one.
Read your first honest number this week.
No card for Starter.