Hypotheses that convert.
Structured experimentation programme with 12 to 24 simultaneous tests. Hypotheses driven by behavioural economics, Bayesian frameworks and quarterly governance review.
Experimentation as a discipline, not a one-off tactic.
Poorly executed CRO is dangerous: false positives lead teams to implement changes that degrade revenue. Caporal structured its experimentation programme over 12 years serving banks and FMCG. where misreading a test costs contracts. Our methodology combines ICE prioritisation with behavioural economics frameworks to generate hypotheses with high probability of positive impact.
What we deliver
Prioritized hypothesis roadmap
Experiment backlog ordered by ICE score, with behavioral rationale documented for each hypothesis, impact estimate and test window.
Execution of 12-24 simultaneous tests
Technical setup in VWO, Optimizely or GrowthBook. Precise segmentation by cohort, device and channel. Daily monitoring of guardrail metrics such as revenue, bounce and funnel performance.
Bayesian analysis & documentation
Reading through a Bayesian framework to reduce type I and type II errors. Complete documentation for each test: hypothesis, result, learning and recommendation to roll out or discard.
Governance review trimestral
C-suite review of the test portfolio, accumulated win rate, attributed financial impact and priority recalibration according to business goals.
Playbooks por canal
Learning library organized by channel, such as landing pages, email, product and checkout, to accelerate future cycles and onboard new teams.
How we operate
Funnel diagnosis (week 1)
Audit of existing analytics, mapping of funnel drops by cohort and identification of priority friction points. Deliverable: funnel map with ranked opportunities.
Hypotheses & prioritization (weeks 2-3)
Workshop with product and marketing teams to build the backlog. ICE scoring with behavioral rationale. Approval of the first 5 hypotheses for immediate execution.
Technical setup (weeks 3-4)
Implementation of the chosen testing platform, conversion event setup, definition of guardrail metrics and approval of variation designs.
Continuous execution (monthly)
18-day cycle per hypothesis: build, QA, launch, monitoring and reading. Up to 24 simultaneous tests in an enterprise program. Dedicated Slack channel with daily updates.
Analysis & roll-out (per cycle)
Bayesian reading, roll-out or discard decision and learning documentation. Winner implementation coordinated with the client tech team.
Governance review (trimestral)
Executive report on win rate, accumulated financial impact, priority recalibration and planning for the next 90 days.
Frequently asked questions
How many tests can run simultaneously?
In enterprise programmes, we run between 12 and 24 simultaneous tests. Each hypothesis has a defined owner, window and acceptance criteria before starting.
What testing platforms do you use?
VWO, Optimizely and GrowthBook. Platform choice depends on client infrastructure and scale. We also have an internal stack for Bayesian analysis.
What is the average test cycle?
18 days per hypothesis on average. from build to roll-out decision. With Bayesian frameworks, we can read results faster than the classical t-test without losing statistical rigour.
How is financial impact measured?
Each test has a primary revenue metric defined before launch. Quarterly review consolidates accumulated win-rate and attributed financial impact, presented to C-suite.
How does design variation approval work?
We work within the client design system. For variations that require new components, we create the spec in Figma and go through design team approval before implementation. We do not compromise brand guidelines just to gain testing speed.