An executive website review should answer a decision, not merely display a scorecard. Traffic may be falling because demand changed, a page lost visibility, a tracking property broke, or the offer no longer fits the buyer. A slow template may matter, but a faster page with the wrong promise will not repair a commercial path. Diagnose the evidence and the first bottleneck before approving a large roadmap.
Start with the decision to be made
Write the decision in one sentence: approve a technical repair, improve a conversion path, consolidate content, test a new offer, or defer investment while evidence is repaired. Add the time horizon, affected audience, budget boundary, owner, and acceptable risk.
Do not begin with a list of requested features. “Rebuild the site” describes a delivery method, not a business problem. The review must show why the proposed work is expected to change a defined outcome and what evidence would show that the assumption was wrong.
Separate the five evidence layers
Use the Executive Website Decision Memo to keep five layers distinct:
| Layer | Questions | Typical evidence | | — | — | — | | Demand | Are relevant people looking? | query, impression, referral, market context | | Visibility | Can the intended pages be found? | indexed URL, query-page match, canonical, search result | | Experience | Can people use the page? | device, template, load and interaction observations | | Action | Do suitable visitors take the next step? | event, form, call, booking, qualified inquiry | | Commercial | Does the action support the business? | accepted lead, opportunity, revenue, retention |
Search Console’s Performance report guidance describes impressions, clicks, queries, pages, and filters that can be used to inspect search visibility. These are important observations, but they are not a substitute for a CRM or finance record. Keep a search decline separate from a sales decline until the intervening evidence is checked.
Diagnose the shape of the change
Compare a consistent period with a prior period and segment by page, query class, country, device, template, and brand status. Look for the first layer that changed materially:
- Demand changed while rankings and page experience stayed stable.
- Visibility changed for a page set while broader demand stayed stable.
- Experience degraded for a device or template after a release.
- Actions fell while qualified traffic stayed similar.
- Actions held but accepted opportunities or revenue changed downstream.
Do not infer causality from two lines moving together. Record the alternative explanations: tracking changes, consent or privacy configuration, seasonality, sales capacity, offer changes, pricing, outages, and a change in how the CRM classifies an inquiry.
Test technical experience in context
Google’s Core Web Vitals report helps identify groups of URLs with user-experience issues. Treat a report as a diagnostic input, not as a universal redesign mandate. Confirm which templates, devices, markets, and journeys are affected, then connect the issue to a user task or business outcome.
For each suspected issue, record the URL group, evidence date, affected journey, reproducibility, release history, owner, and rollback option. A laboratory score or a single slow page may be useful for prioritisation, but it should not be presented as proof that all conversion loss came from performance.
Validate the action measurement path
List the events that represent meaningful progress: view of a relevant offer, completed qualification question, submitted inquiry, booked meeting, or other agreed action. Google’s GA4 event documentation describes events and parameters; the business still has to define what each event means, when it fires, and how it is reconciled with a real record.
Sample the path from page to action. Check whether the event fires once, whether consent and exclusions change the count, whether forms produce an owned record, and whether the sales or service team can confirm quality. If the measurement path cannot reproduce a small set of journeys, label the result as uncertain and repair the evidence before making a large investment.
Build a bottleneck hypothesis
Use a table that forces a decision:
| Observation | Leading hypothesis | Cheap falsification test | Investment implication | | — | — | — | — | | relevant impressions down | demand or visibility changed | compare query and page cohorts | investigate discovery before redesign | | impressions stable, clicks down | result promise or intent mismatch | review query, title, and page fit | test message before template rebuild | | clicks stable, actions down | page, offer, form, or trust break | replay a representative journey | repair the narrow path first | | actions stable, qualified records down | fit or qualification changed | sample accepted and rejected records | align offer and sales gate | | experience issue on one template | technical bottleneck | isolate template and device | fix the smallest affected surface |
Choose the smallest test that can distinguish the hypothesis from its strongest alternative. A roadmap should sequence evidence repair, reversible experiment, implementation, and follow-up review rather than bundle every possible improvement into one irreversible project.
Review the portfolio and the owner
An executive review should show which page groups deserve attention and who can act on each finding. Group URLs by offer, audience, template, lifecycle stage, and commercial owner. Mark pages that are intentionally informational, supporting, transactional, or retired. This prevents a high-traffic page from consuming the roadmap simply because it is visible, while a low-volume page that serves a strategic account is ignored.
For each proposed item, record the dependency on product, sales, legal, engineering, content, or data teams. State what happens if the dependency is not available. A technically correct recommendation that no owner can implement is not a funded roadmap item; it is an unresolved assumption.
Write the executive decision memo
The memo should contain: the requested decision; scope and time window; evidence sources and freshness; material change; first suspected bottleneck; alternative explanations; proposed test; cost and capacity; owner; success and stop criteria; and rollback path. Put unknowns beside the claim they limit. A clean page is not proof of a clean measurement system.
Approve the roadmap only when the first phase has a bounded question and a way to learn. If the review cannot distinguish visibility, experience, action, and commercial evidence, approve an evidence-repair phase instead of pretending that a large build is already justified.
How did this article land?
Choose one reaction. You can change it anytime.
