Skip to content

CRO Process

Use this SOP to run CRO as campaign angles, not one-off tests, and to take each idea from an observed need to a tested or shipped improvement.

Before the campaign loop

For the first campaign, complete Client CRO Readiness once, then build the Initial Baseline. If that baseline is current, continue directly to Data & Strategy without a duplicate refresh.

For every later campaign, complete Campaign Refresh & Analysis after Evaluation closes the previous campaign. It updates the evidence pack without rebuilding the full baseline.

Campaign checklists

Run the campaign itself through four connected checklists:

  1. Data & Strategy: investigate the leak, choose the angle, set a tangible numeric objective, lock the measurement plan, and hand an approved brief to Design.
  2. Design: turn the approved brief into reviewed, build-ready assets.
  3. Build & Launch: implement, QA, launch, monitor technical health, and hand the verified launch record to Evaluation.
  4. Evaluation & Decision: validate the result, quantify and responsibly label commercial impact, choose and apply the evidence-supported decision, record the learning, and close the campaign.

After Evaluation & Decision closes the campaign, continue to Campaign Refresh & Analysis for the next one. Each interactive checklist saves progress only in the current browser and can be reset for a new campaign.

The model: campaign angles, not one-off tests

A campaign angle is a strategic theme that a whole body of work reinforces at once: landing pages, ad-to-site continuity, on-site content, and the tests themselves. Running CRO this way does two things a scatter of unrelated tests cannot:

  • Stronger intent per test. Every test answers part of one clear thesis about the buyer, so a result teaches you something about the angle, not just the element.
  • Compounding wins. Changes stack on the same theme, so each win builds on the last instead of standing alone.

Example angle: Buyer's Education, the thesis that shoppers hesitate because they do not fully understand what they are getting, which a set of tests, a matched landing page, and product-page content all reinforce.

Steps 1 to 2 set the angle. Steps 3 to 8 take each test from question to result. Step 9 compounds the wins.

1. Choose the campaign angle

Pick one strategic angle for the round, grounded in evidence rather than preference:

  • Where the funnel leaks (analytics, heatmaps): the stage losing the most.
  • What the customer needs to hear (reviews, support tickets, ad comments): the belief or worry in the way.
  • The angle names a buyer problem, not a page. For example Buyer's Education, Risk Reversal, Speed to Product, or Social Proof.

Run one angle at a time so it becomes the through-line for the round.

2. Build the angle's assets

Line up everything that carries the angle so the experience is continuous:

  • Matched landing pages for paid traffic, so the page lives up to the ad it came from.
  • Ad-to-site continuity: the same message from click to page.
  • On-site modules and content that state the angle where it matters (product page, collection, cart).

Continuity is the point: a visitor should meet the same idea at every step.

3. Define The Question

Write each test as a specific question that serves the angle:

Where are users dropping off, and what change tied to the angle would reduce that friction without weakening the offer or brand?

Capture:

  • Page or funnel step.
  • Primary metric.
  • Current baseline, numeric target, absolute change, and relative change.
  • Secondary guard-rail metrics and their harm limits.
  • Audience or traffic source affected.
  • Known constraints.

4. Gather Evidence

Use at least two evidence types before recommending a change.

Evidence Type What To Look For
Analytics CVR, ATC rate, checkout reach, revenue per session, device split, channel mix
Behavior Rage clicks, dead clicks, scroll depth, missed CTAs, navigation confusion
Customer language Reviews, support tickets, survey answers, ad comments
Technical LCP, INP, CLS, script weight, broken tracking, app conflicts
Merchandising Product hierarchy, stock status, offer clarity, bundle logic

5. Write The Hypothesis

Use this format:

Because [evidence], we believe [change] for [audience/page], as part of [the angle], will improve [primary metric]. We will know it worked if [success threshold] without harming [guard-rail].

Good hypotheses are specific enough to be wrong.

6. Choose Test Or Rollout

Path Use When
A/B test Enough traffic, clear control, reversible change, uncertain impact
Direct rollout Fix is required, traffic is too low, brand update is already approved, or the risk of keeping the current state is higher
Research only Evidence is weak, problem is unclear, or implementation cost is not justified yet

7. QA Before Launch

Minimum checks:

  • Mobile and desktop layouts.
  • Add-to-cart flow.
  • Cart and checkout path.
  • Product variant behavior.
  • Tracking events.
  • Load stability and image dimensions.
  • Accessibility basics: labels, focus, contrast, keyboard path.

8. Evaluate

Record:

  • Date range.
  • Sample size.
  • Primary metric result.
  • Absolute and relative movement against the tangible objective.
  • Guard-rail metrics.
  • Confidence or directional strength.
  • External noise: sale, traffic spike, outage, campaign change, stock issue.
  • Commercial-impact label and 90-day projection where the evidence supports one.
  • Decision and next action.

Use Data & Strategy to lock the tangible objective and impact method before launch, then use Evaluation & Decision to calculate and label the commercial result.

Avoid declaring a winner from a short sale window, a campaign traffic shift, or a partial week unless the caveat is explicit.

9. Compound

Turn each result into fuel for the next, so the program builds rather than resets:

  • Roll wins into the baseline. A shipped winner becomes the new control; the next test builds on it, not on the original.
  • Feed the angle forward. Losses and near-misses sharpen the angle's next test rather than ending it.
  • Move to the next angle only when this one is exhausted, carrying the current best forward so intent and results keep compounding across rounds.