When this scope is useful
The journey works technically, traffic is relevant and there is enough observation to distinguish signal from noise.
With low traffic, start with usability research and clearly labeled improvements instead of promising statistical certainty.
Questions that distinguish the team
- What is the primary metric and unit of analysis?
- What sample and stopping rule are planned?
- Which guardrails prevent a misleading win?
Inputs before work starts
Journey data, conversion definition, traffic volumes, device mix, change history and research access.
Three ways to scope the work
These are buyer-side scope models, not an agency price list. Use the inputs, client responsibilities and exclusions below for each model.
01 / Journey research
- Work
- Research the journey and identify plausible friction.
- Handover
- Evidence-based hypothesis backlog.
02 / Planned experiment
- Work
- Plan and implement a bounded experiment with QA.
- Handover
- Experiment plan, variants and event validation.
03 / Learning program
- Work
- Review outcomes against the agreed rule and retain learning, including inconclusive tests.
- Handover
- Decision record with uncertainty and next action.
Acceptance and shared responsibility
The metric, sample assumptions and stopping rule are written before launch; variants receive correct tracking and no unexplained concurrent changes.
Client responsibilities
Provide access, approve variants and maintain stable operating conditions where possible.
What needs separate agreement
Guaranteed uplift, post-hoc success rules and claims of significance without sufficient data are excluded.
Ask for a cost estimate based on the actual access, production volume, integrations and review cadence. A comparable price requires the same scope.
Take this into the brief
Service: Conversion rate optimization Inputs: Journey data, conversion definition, traffic volumes, device mix, change history and research access. Acceptance: The metric, sample assumptions and stopping rule are written before launch; variants receive correct tracking and no unexplained concurrent changes. Questions: What is the primary metric and unit of analysis? What sample and stopping rule are planned? Which guardrails prevent a misleading win?
Common buying questions
Is every increase a test win?
No. Apply the planned decision rule and consider uncertainty.
What if the result is inconclusive?
Record it and review sample, implementation and the value of another test.
Evidence to examine
Evidence review
One Click Contractor: traffic and MQLs move differently
Different growth rates can be useful for diagnosing the journey, but they do not by themselves prove that a redesign caused the change. Inspect audience mix and qualification rules.
Evidence review
Norse Atlantic: keep the sales growth baseline visible
A new airline changes distribution, route availability and awareness while campaigns run. An increase during launch is not a reusable forecast for a mature business.
Sources and review date
Sources read on: 2026-09-11. A source review is not an audit of the agency’s systems.