Flywheel
Your teams plan, run and report experiments and research with whatever tools they use, including AI assistants. Flywheel checks each piece against the standards your organisation sets, and shows leaders which results passed and which did not.
What gets checked
Trusted by
Decision Intelligence Stack
Flywheel covers each layer of the Decision Intelligence Stack, for experiments and research alike.
Can the experiments and research be trusted?
Can what you know be found and connected?
Does the evidence reach decisions?
The standard your organisation sets, applied to every experiment and study at every layer.
Governance
Flywheel suits organisations where experiments and research already happen, but the results do not reliably reach the people making decisions. It is likely to fit if one or more of these is true.
You run 50 or more experiments a year but cannot show their effect on strategy. Winning tests are reported, but nobody tracks whether they delivered value once released.
Your research team's work sits in repositories nobody reads. Product decisions are made without reference to studies that already exist.
Experiments and research are managed across Jira, Airtable, Notion and spreadsheets. No single place shows the whole picture.
Leadership relies on the team to report on its own work. There is no independent view of quality.
Not sure yet? Start with an evidence audit. It shows where you stand before you commit to a platform.
Decision influence
Decision influence
Each decision records who made it, which experiments and research informed it, when it was implemented and whether it delivered the expected outcome.
Evidence integrity
Every experiment receives a governance score based on how it was planned, run and documented, so you can see which results are sound enough to act on.
Decision influence
Dashboards show decisions informed by evidence, governance health and outcomes compared with predictions, instead of test counts and win rates.
Experiments and research
Evidence integrity
Evidence synthesis
Decision influence
Works where your teams work
Flywheel connects to AI assistants that support the Model Context Protocol (MCP), such as Claude. Before suggesting a test, the assistant checks what your organisation has already learned. What it records goes into the same Flywheel records that leaders see.
Illustration. The company and figures are fictional.
Governance
They record what people choose to enter. Fields are optional, and a hypothesis can be edited after the results are known without anyone seeing the change.
They run experiments well. They do not hold research, and they do not link results to the decisions they inform. Flywheel connects to them; you keep using them.
An internal build or an AI workflow gives full control, but it takes engineering time to build and maintain, and the team whose work is checked ends up setting the standard.
Governance
Most organisations are live within four to eight weeks. The total time needed from your team is about one to two days.
Governance design. A half-day workshop with your executive sponsor and the people who will use Flywheel. Together we agree the required fields, approval steps and quality checks.
Set-up. We configure Flywheel, connect your testing tools, import past experiments and research, and assess their governance so your dashboard shows real data from the first day.
Training. Separate sessions of about two hours for practitioners and researchers, and for executives.
First 90 days. We review the first experiments and studies entered, run a health check at week eight, and deliver your first quarterly governance report at week twelve.
Onboarding is included in the subscription.
Governance
Flywheel is priced by user licence. Programme managers, who set the standards, are priced higher than practitioners. Executives and other viewers are free. Single sign-on is included. Request a call for a quote.
Governance
With an evidence audit. We trace three recent decisions back to the experiments and research behind them, and send you a written finding for each. If the findings show you need checks on an ongoing basis, Flywheel is the next step. If not, you keep the report.
Yes. Your teams upload research files such as PDFs, slides and transcripts. Flywheel drafts the findings with AI, and a researcher verifies them before anyone else sees them. Experiments and research sit in the same record and are checked against the same standards.
No. Flywheel sits alongside Optimizely, VWO, AB Tasty or whichever tool you use, and adds the record, the checks and the link to decisions.
Any assistant that supports the Model Context Protocol (MCP). Flywheel works with Claude, ChatGPT and Gemini. What your teams share with an assistant is covered by your organisation's agreement with that provider.
They keep running. We import them in their current state, along with your past experiments and research.
The standards are agreed with your teams in the design workshop, not imposed on them. Executive sponsorship matters, because standards hold when they have support from above.
Nothing. Executives and other viewers use Flywheel free of charge, and single sign-on is included in the price.
In the European Union.
No. Nothing you put into Flywheel is used to train AI models.
We export your data and return it to you, then delete it from Flywheel within 30 days.
Yes. Flywheel subscriptions run for a minimum of 12 months.
Flywheel
Leave your details and Manuel will reply by email to arrange a call.