Data & Strategy Checklist

Required evidence input

For the first campaign, complete Client CRO Readiness and the Initial Baseline. For later campaigns, complete Campaign Refresh & Analysis. These inputs are separate from campaign progress and do not change the seven steps below.

How to use this

Start here before any design or build work. Work top to bottom until the approved brief is ready for Design. Progress is saved in this browser on this device and is not shared. Use Reset for a new campaign only after Evaluation & Decision has closed the current campaign.

Tangible objective standard

Give every A/B test a numeric pass or fail objective before Design begins:

Increase or reduce [primary metric] from [baseline] to [numeric target], equal to [absolute change] and [relative change], for [eligible audience], without [guardrail] crossing [harm limit].

State both forms of change. Moving add-to-cart rate from 6.0% to at least 6.6% is +0.6 percentage points and +10% relative. Reducing bounce rate from 48.0% to no more than 43.2% is -4.8 percentage points and -10% relative. Choose the metric closest to the tested funnel step instead of defaulting every test to sitewide CVR.

Use this checklist to open one evidence-led CRO campaign and hand an approved brief to Design.

Reset for a new campaign

1. 1. Frame the campaign

   Define the business question before choosing a tactic Identify the store, audience, page, funnel step, and primary conversion goal Record traffic, timing, promotion, stock, access, tool, and approval constraints Separate traffic-quality problems from on-site conversion problems

   Required output: One specific business question and a clearly bounded campaign scope.

2. 2. Investigate the leak

   Review the recent baseline for sessions, CVR, ATC, checkout reach, AOV, and revenue per session where available Review device and channel splits where available Identify the page or funnel stage with the strongest evidence of leakage Check performance, tracking, stock, merchandising, promotions, outages, and other known noise Record source links and the measurement window

   Required output: A prioritized problem statement supported by current quantitative evidence.

3. 3. Research why it may be happening

   Use at least two evidence types before recommending a change Review relevant heatmaps or session recordings Review customer language from reviews, support tickets, surveys, or ad comments Review technical, merchandising, or competitor evidence when it directly informs the question Separate observed evidence from interpretation

   Required output: A short evidence summary explaining why the leak may exist.

4. 4. Choose the campaign angle

   Name one buyer problem or belief that will anchor the campaign Confirm the angle is grounded in the investigation and research Map the angle across every relevant acquisition and on-site touchpoint Prefer an angle test before isolated element tests unless the angle is already proven

   Required output: One approved campaign angle and its in-scope touchpoints.

5. 5. Define the optimization path

   Write each hypothesis with evidence, change, audience or page, angle, primary metric, success threshold, and guardrail Choose A/B test, direct rollout, or research only for each proposed change Define one primary pass or fail metric tied directly to the hypothesis and tested funnel step Record the baseline, numeric target, direction, absolute change, and relative uplift or reduction for each test Define secondary diagnostic metrics and guardrails Give every guardrail a numeric harm limit or an explicit no-harm decision rule Define the traffic split and mutually exclusive bucketing for experiments Record the recent control baseline

   Required output: An approved hypothesis set with numeric pass or fail criteria and a documented test or rollout decision for every proposed change.

6. 6. Lock the measurement plan

   Calculate the required sample size and expected duration Set a fixed planned end date before launch Plan at least two full weeks, with four weeks preferred Set 95% confidence as the gate for winner claims Set a practical relative lift threshold of 5% to 10% Target at least 100 conversions per variant Require control performance within 10% to 15% of its historical baseline Treat mobile performance as a required ecommerce guardrail Predefine the attributed-revenue source, attribution definition, eligible population, visitor or session denominator, traffic baseline, currency, and 90-day projection method to use if the test produces a valid winner For direct rollouts, define the pre-launch baseline and post-launch comparison window List the external events that must be logged during the campaign

   Required output: A measurement and commercial-impact plan that can be evaluated without redefining success or choosing a favorable formula after launch.

7. 7. Hand off to Design

   Package the question, evidence, angle, hypotheses, surfaces, audience, metrics, constraints, sources, and measurement plan State what is fixed and where Design has room to explore Confirm the brief has an owner and approval Confirm Design does not need to infer the problem, angle, scope, or success criteria

   Handoff gate: Design has one approved source of truth.

Continue to Design