Rigorous experimentation for every domain
Run experiments across
every part of your business.
Decision Process is a general-purpose experimentation platform — built for product teams, retail operations, agricultural researchers, and manufacturing engineers alike. If you have a question that data can answer, we have the tools to answer it rigorously.
A/B → 20-arm
Any experiment design
Bayesian posteriors
Not p-values
Causal inference
Adjusts for confounders
No SDK required
Field, process & physical trials
Where it runs
One platform. Any domain.
Digital & Web
Checkout flow simplification
+18% conversion rate
See examples →
Retail & Commerce
End-cap vs. aisle placement
+14% units sold
See examples →
Healthcare
Discharge education protocol
−19% readmission rate
See examples →
Education
Active vs. lecture learning
+24% assessment score
See examples →
Agriculture
Fertilizer rate comparison
+22% yield per acre
See examples →
Marketing
Email subject line test
+37% open rate
See examples →
Manufacturing
Maintenance schedule A/B
−31% defect rate
See examples →
HR & People Ops
90-day onboarding program
+28% 1-yr retention
See examples →
Process
How it works
Design
Pick a domain, define conditions, choose metrics. Use a template or build from scratch. Works for A/B, multi-arm, crossover, and adaptive designs.
Run
Collect observations manually, via CSV, or through your existing data systems. No SDK required for field trials, process experiments, or physical interventions.
Analyze
Bayesian inference with credible intervals, effect sizes, and plain-language summaries. Causal adjustments when confounders matter.
Decide
Get a clear recommendation: which condition wins, by how much, and with what confidence. Every result is documented — an audit trail for the decision, not just the data.
Why it matters
What changes when you experiment rigorously.
Reduce failed rollouts
Test before you scale. Catch underperformers at the pilot stage, before they cost real money.
Audit trail for every decision
Every experiment has a complete record — what was tested, the conditions, the result, and why you chose what you scaled.
Eliminate false positives
Bayesian credible intervals don't inflate under repeated peeking. You get honest uncertainty, not statistical theater.
No statistics team required
Results come with plain-language recommendations and probability-of-improvement. Your team decides, the math handles itself.
Any intervention, any context
From web traffic to field plots to production lines. The platform adapts to your data, not the other way around.
True causal effects
Adjust for confounders, estimate effects under intervention, and distinguish correlation from causation — built into every analysis.
Example
A fertilizer trial at 12 farms
A cooperative tests three nitrogen application rates across 12 farm plots over a growing season. Decision Process handles the randomized assignment, collects yield and input-cost observations, and delivers a Bayesian comparison of all three conditions — no statistics degree required.
- ✓3-arm design: Low N (control), Standard N, High N
- ✓Metric: yield per acre (lbs) + input cost per acre (USD)
- ✓12 farm plots, 1 growing season = 12 observations per arm
- ✓Result: Standard N achieves 94% of High N yield at 62% of cost
// Results: yield_per_acre
Standard N outperforms control with 97% probability. Effect size: d = +0.82 (large).
Positioning
Not just another A/B testing tool.
| Capability | Generic A/B tool | Decision Process |
|---|---|---|
| What you can test | Webpages only | Any domain |
| Metric types | Conversion rate | Binary, continuous, count |
| Design types | A/B only | A/B, multi-arm, crossover, adaptive |
| Data collection | JavaScript SDK | SDK, manual, CSV, API |
| Analysis | p-values | Bayesian posteriors + causal inference |
| Deployment | Web traffic | Physical locations, batches, people, plots |
Nonprofits & academic institutions get full access free.
Rigorous causal inference shouldn't be gated by budget.
Built on Reinforce OS
Part of a family of learning systems.
DoOperator builds systems that learn from evidence. Decision Process is the enterprise layer — the same inference engine, the same causal reasoning, applied to operational decisions at scale.
The platform
Adaptive decision systems, causal inference, and reinforcement learning — the technical foundation everything is built on.
Personal science
What works for you, specifically. Individual-level experiments and habit tracking with Bayesian feedback.
Civic experimentation
Public institutions learning openly. A registry of civic experiments and shared methodology for policy and community decisions.
Enterprise experimentation
Operational decisions, at scale, with rigor. The enterprise layer of the family.
Ready to run your first experiment?
We're onboarding teams in private beta. Tell us about your use case and domain.