Running
Day 9 / 14
Landing page value proposition
+2.1 pts
Sign-up rate · 11.8% → 13.9%
p = 0.08 · not significant yet, keep collecting
Business experiments, run like a lab
Let the evidence decide.
Write a hypothesis that can be wrong, plan how to measure it, run it, and keep the result where the next person will find it. Detonative Labs is the lab notebook for product, marketing and sales experiments.
Free for teams up to 3. No card needed.
Example · Website project
Verdict · Shorter sign-up form
Win
Hypothesis not falsified
- Lift
- +3.5 pts
- Relative
- +29.2%
- p-value
- 0.023
- Confidence
- 95%
95% interval, change in sign-up rate
The whole interval sits above zero.
Matomo · 1–14 Oct
How it works
Every experiment follows the same five stages.
The same fields, the same review, the same verdict labels. That is what makes a result from March comparable to one from October.
- DraftWrite the bet
- In reviewA peer checks the plan
- ApprovedLocked before launch
- RunningResults sync every morning
- ConcludedVerdict on file
Write it so it can be wrong
The hypothesis builder asks for a change, a metric, a direction and a minimum effect. If the evidence cannot refute it, it is not ready to run.
If we cut the sign-up form to 3 fields, then sign-up rate will increase by at least 2 pts.
Decide the sample before the first visitor
Enter your baseline and the smallest effect worth knowing about. Detonative Labs works out the sample per variant and how long to run, so nobody stops early on a good day.
- Baseline
- 12.0%
- Minimum effect
- +2.0 pts
- Power · α
- 80% · 0.05
- Sample per variant
- ≈ 4,435
- At 600 visitors a day
- 15 days
Three verdicts, named plainly
Every concluded experiment gets one of three outcomes, with the falsification term beside it and a sentence on what the numbers mean.
Win (not falsified)
Loss (falsified)
Inconclusive (undetermined)
Features
An instrument panel, not a marketing dashboard.
Quiet chrome, loud data. Numbers are set in mono so they line up and read as measurements.
Experiment record
Hypothesis, metric, guardrails, procedure, result and verdict on one page. No free-form notes, no missing fields.
Sample planner
Sample per variant, power and minimum detectable effect from your baseline, plus the run length at current traffic.
Matomo sync
Results import from Matomo every morning, or on demand. Running cards show lift, trend and p-value with a plain-language reading.
Library
Start from templates for common tests: sign-up flows, pricing pages, onboarding, cold outreach. Save your own for the next team.
Projects
Group experiments by website, onboarding or outbound sales. Each project keeps its own colour, owners and history.
Peer review
A colleague approves the plan before it runs. Roles for owners, editors and reviewers, one tenant per install.
Archive
Losses stay on file too.
A falsified hypothesis is a result. Search past experiments by hypothesis, metric or verdict before you spend a sprint on an idea someone already tested.
Outcomes are blue and amber rather than green and red, and each has its own icon and word, so they read the same for everyone.
| EXPERIMENT | DATES | RESULT |
|---|---|---|
Shorter sign-up form |
1–14 Oct | Win +3.5 pts |
Checkout page redesign |
8–21 Sep | Loss −4.0 pts |
Demo qualification script |
1–28 Sep | Inconclusive +0.6 pts |
Activation onboarding checklist |
7–27 Nov | Running +2.4 pts |
Hosted or self-hosted
Run it on our servers or on yours.
Same app either way. Your experiments stay readable and exportable whichever you choose.
Hosted
lab.detonative.comFreefor teams up to 3
- Unlimited experiments and projects
- Sample planner and Matomo sync
- Larger teams: pricing on request
Self-hosted
MIT licensedFreeon your own infrastructure
One container stack, one tenant per install. From the repository root:
$ docker compose up -d
Start
Your next idea is a hypothesis. Treat it like one.
Create an account in a minute and log your first experiment today.
Free for teams up to 3Unlimited experimentsMIT licensed