Ledger and Statsig
Ledger or Statsig?
Statsig runs controlled experiments on your product and tells you what each variant did. Ledger keeps every decision your team makes, tested or not, with what you expected and what actually happened.
new_checkout_flow
Experiment · Decision: ship test group
Hypothesis: the shorter checkout raises purchase conversion.
Purchase conversion rose in the test group, and the result is significant.
Summary written by the growth team.
Move pricing to annual-first
- Paid conversion≥ 4%
- Watching: refund requestsNo worse
- Rolled out to everyone at onceLow confidence
- Check onDec 1
The numbers arrive from Stripe every day
Where the two part ways.
- 1
Ledger records the calls you can't A/B test
Pricing changes, a new hire in support, a policy change in claims and a vendor switch all get the same record: the number it should move, the date to check it and the lesson afterward.
- 2
It's built for the whole company
Finance, operations, sales and support write decisions in plain language. Nobody needs an SDK in the code for the record to exist.
- 3
It's honest about how a change rolled out
Each decision says whether it went to everyone, part of the way or against a held-out group, and Ledger shows how much the result can be trusted because of it.
- 4
Numbers come from the tools you already use
Connect Stripe, Amplitude, Pipedrive or any tool with an MCP server, and readings arrive hourly, daily or weekly without anyone pulling them.
- 5
Lessons outlive the experiment
When a decision settles, someone writes what the team now believes, and that lesson stays searchable next to the decision it came from.
When Ledger is the right choice
- Most of your decisions happen outside the product code.
- You want one record of every call, with what you expected written down first.
- You want the numbers to arrive on their own from the tools you already pay for.
- You want a team record of how often expectations held, with no one ranked.
Job by job
What gets recorded
Statsig: Experiments and feature rollouts, each with a hypothesis.
Ledger: Any decision, with the numbers it should move and a date to check them.
Who writes it
Statsig: Mostly product, engineering and data teams.
Ledger: Anyone in the workspace, in plain language.
Where the numbers come from
Statsig: Events logged by Statsig's SDKs, or your data warehouse.
Ledger: Feeds from connected tools, your code, a pasted spreadsheet or a typed reading.
How a result is judged
Statsig: A statistics engine compares test and control groups.
Ledger: Ledger shows the numbers against the target, and a person marks it confirmed, mixed or missed.
Guarding against side effects
Statsig: Secondary and guardrail metrics on each experiment.
Ledger: Watched claims that must hold no worse than the weeks before the change.
Getting back to it
Statsig: Each experiment has a target duration, and its summary records the decision.
Ledger: The person who wrote the decision gets a reminder when its date arrives.
Test in Statsig. Remember in Ledger.
Statsig tells you what a product change did to the people who saw it. Ledger keeps that result beside every other call the company made about the same number.
- Record a decision as rolled out against a held-out group, and Ledger reads its result with higher confidence.
- Add the experiment's result as evidence on the decision it settles, with the experiment's name as the reference.
- Connect Statsig's MCP server, and Zero can read your experiments when it drafts a decision.
Questions
- Can a Statsig metric feed Ledger?
- Ledger feeds read from any tool connected over MCP, and Statsig publishes an MCP server. You can also send readings from your code with an API key.
- How does Ledger show how well the team decides?
- Ledger shows how often the team's expectations held across the whole workspace, and it never splits that by person.
Try Ledger with your domain.
Write down the next call your team makes that no experiment will cover.
No website? Describe what you do
If you'd like help with the first changes, talk to us.