What to Measure in the First 90 Days After a Website Launch

For a rebuild, save the previous site's important URLs, search performance, inquiry definitions, and recent business results. Record the date range, campaigns running, service areas, and unusual conditions. If the business stopped taking one service or increased advertising at the same time as launch, that context belongs beside the numbers.

For a completely new site, the baseline may be zero website history. Do not invent a comparison to an established competitor. Start with an agreed service offer, expected visitor actions, and a functioning measurement plan. Early observations can establish the next baseline even when they cannot establish a trend.

Before publishing, define the few actions that matter: a completed quote request, a connected phone inquiry, an appointment request, or another meaningful step. Label button clicks as clicks. Calling every interaction a lead makes launch reporting look busy without explaining demand.

Define a lead before building the dashboard. For a small service business, a useful working definition might be a unique, genuine inquiry about obtaining a service. That definition excludes spam and internal tests but can still include requests the business cannot accept. Qualification is a later judgment about fit, not a reason to erase inconvenient inquiries from the original count.

Keep the record simple enough that staff can maintain it during a busy week. An inquiry identifier, received date, service, source evidence, status, and reason code may be sufficient. Use a separate field for the date a status changes. Otherwise moving a request into “booked” can make it appear to be a new inquiry in the booking month.

When a person submits a form and calls about the same request, connect those interactions to one inquiry where the business can establish the relationship. Do not force a match when evidence is missing. A documented duplicate policy is more useful than a dashboard that appears precise because it quietly ignores ambiguity.

State the denominator beside each rate. Qualified inquiries divided by genuine inquiries is a fit measure; genuine inquiries divided by measured sessions is a website journey measure with tracking limitations. Calling both a conversion rate can hide a substantial change in what the report means.

Test the live visitor journey on mobile and desktop. A real test includes successful submission, clear error handling, delivery to the correct inbox or CRM, and a named person who sees the request. Check phone links, business hours, navigation, and the next step explanation. Verify that test submissions are labeled so they do not contaminate the business totals.

Review the technical details that can prevent discovery or break existing journeys. A rebuild with changed URLs needs an explicit redirect map, while a site that kept its URLs needs a different change checklist. Google's guidance for moves with URL changes is useful when that situation actually applies. Avoid treating every redesign as a domain migration.

Fix an obvious broken contact route immediately. Do not wait for a monthly report to confirm that losing inquiries is undesirable. A website launch scope should identify who owns these checks and who can make corrections after publication.

Give the office a test schedule and a clear way to recognize test records. If a vendor receives the notification, require the internal owner to see the result too. A successful demonstration inside the developer's preview does not establish that the live site reaches the business's actual destination. Confirm production addresses, routing, and the exact confirmation wording.

Record the minimum evidence needed to repeat a defect: page, time, device or browser, steps, and actual result. Ask for a screen recording only when it helps explain an interaction that a screenshot cannot capture. This keeps the issue log usable without turning ordinary launch support into a documentation project.

For performance, save representative evidence from a service page and the contact journey. The Web Vitals documentation distinguishes loading, interaction responsiveness, and visual stability. Label controlled tests separately from real visitor measurements, and record whether a field result applies to the page or a broader group. A new page may not yet have enough field data; that is missing evidence, not proof that it passes or fails.

Agree on which conditions require immediate intervention. A confirmed lost inquiry, incorrect contact number, or inaccessible primary action deserves prompt repair. An untested idea about a different headline can wait for a considered review. This distinction protects the team's time while keeping operational failures visible.

Review whether important pages are discoverable and whether there are clear indexing or crawling issues to investigate. Search Console can show search activity, but the numbers require context. Its Performance report describes clicks, impressions, click through rate, and position; those are search observations, not a count of customers.

Look at groups of pages with a common job. Are service pages being seen for the kinds of requests the business wants? Are visitors landing on outdated URLs? Are people using the quote route or calling with questions the page should answer? Small samples can reveal a broken journey even when they cannot support a reliable conversion rate conclusion.

Keep a launch issue log with the page, evidence, business effect, owner, and resolution date. Separate a confirmed defect from an improvement idea. “Form fails on this device” and “perhaps the headline should be shorter” deserve different priorities.

Choose a few important page groups and review consistent periods, filters, and devices. A service page can receive more impressions while its click through rate falls because it is appearing for a broader set of searches. Before changing the title, inspect the relevant pages and queries. A sitewide average may conceal useful gains in one service and irrelevant exposure in another.

When AI visibility is part of the launch question, Google's Generative AI performance report for Search provides a specific reporting source for impressions from AI Overviews and AI Mode. Do not use a general Web search export as if it isolated that exposure, or turn an automated “AI health” score into a count of recommendations. Keep the report's scope and available data with any result you present.

Search discovery and inquiry attribution also remain separate. A customer may discover an article in search, return directly, and call from a number saved earlier. Website records may show only part of that sequence. The first month should reveal what the business can observe reliably and which gaps remain, so later comparisons do not rely on an invented complete journey.

By this stage, look beyond whether visitors arrived. Review a manageable sample of actual inquiries with the person who handles them. Were the requested services offered? Was the location within scope? Did the page create a mistaken expectation about price, timing, or project size? These observations can make the next content change more precise.

A page attracting the wrong type of job may need a clearer scope statement rather than more traffic. A page attracting suitable prospects who never receive follow up may have an operational problem. Keep those diagnoses separate so that design changes do not become a substitute for answering inquiries.

Use a source method that matches the evidence available. The small business attribution framework helps distinguish tracked campaign information, customer reported discovery, and unknown sources. Do not credit every sale during the period to the new website merely because the site was recently launched.

Suppose an invented service business receives 18 genuine inquiries in its first 30 days and 24 in the next 30 days. At the day 60 review, the first group has six booked jobs and the second group has four. Those numbers do not establish that the second month performed worse: its inquiries have had less time to reach a booking decision.

For a fairer descriptive comparison, inspect each inquiry at a common age, such as 14 days after receipt. Only include records old enough to have reached that age. The chosen interval should fit the business's normal process; 14 days is an assumption for this example, not a universal sales cycle. Keep later outcomes available for a separate mature review.

Both observed proportions are one in six, but the small samples do not prove equal performance. The table mainly prevents unequal follow up time from being mistaken for a website effect. Review the actual request mix, capacity, and unresolved opportunities before choosing the next change. Keep the 12 newer second-month inquiries outside this particular comparison until they mature.

Also retain later bookings in the business record. Limiting an analytical comparison to 14 days does not mean a job booked on day 20 has no value. It simply means that job answers a different reporting question. An owner should be able to move from the comparable early indicator to the current total without losing the relationship between them.

1. If contact tests fail, repair the live path and verify delivery before interpreting demand.
2. If relevant pages cannot be discovered, inspect access, links, indexing evidence, and migration decisions.
3. If visits arrive but inquiries are unsuitable, compare the page promise with staff reason codes and the actual sources of traffic.
4. If suitable inquiries stall before a conversation, check follow up ownership and the response process.
5. If conversations happen but work is not booked, review fit, proposals, availability, and decision timing with the service team.

These are investigation routes, not automatic explanations. A short page review may resolve a clear coverage misunderstanding; a pricing or service capacity issue needs the business owner. Publishing additional articles does not directly repair every stage of the journey.

Some sales cycles will have progressed far enough to assess appointments, proposals, booked work, or completed jobs. Others will not. Compare cohorts with similar time to mature, and distinguish open opportunities from lost ones. A long proposal process should not be labeled a website failure simply because the reporting month ended.

Here is a hypothetical scorecard structure for an owner review. The entries are examples of what to record, not performance claims.

Use a compact review with four entries: the business question, the strongest available evidence, the important uncertainty, and the proposed next action. For example, a repeated pattern of requests for an unavailable service may justify rewriting a scope section and checking the related campaign. The acceptance check would confirm the new wording and monitor the same mismatch reason, rather than promising a predetermined percentage improvement.

Give one person responsibility for each accepted change and state when enough relevant evidence might exist to review it. Some checks can happen immediately after publishing; inquiry quality needs new records; completed work may take longer. Keep the next review tied to that evidence requirement instead of making every decision wait for another arbitrary 90 day period.

What if the site receives very few inquiries?

Verify the journey, traffic relevance, and measurement first. Use individual cases to identify clear problems, while treating rates based on tiny counts cautiously. Low volume alone does not explain whether the issue is discovery, service fit, or the offer.

Should every launch change be credited for later sales?

No. Record the timing and available journey evidence, then acknowledge other changes such as advertising, seasonality, staffing, and sales activity. A launch date is context, not proof of causation.

What should be achieved by day 90?

A functioning customer journey, maintained business records, and a clear decision about the next improvement are sensible operational goals. Rankings, inquiry volume, and mature sales outcomes depend on circumstances that a calendar cannot guarantee.

Annotate major advertising changes, outages, new service launches, holiday periods, and capacity restrictions. Compare equivalent periods where possible, but do not imply that a year over year comparison removes every external influence. Review absolute counts alongside rates; a large percentage change based on very few inquiries can mislead.

If the launch includes several new domains, use the microsite purpose and maintenance questions before interpreting scattered traffic as progress. More pages and properties also create more places where measurement can break.

At the 90 day review, agree on the next business question and the work needed to answer it. A continuing search visibility program can then use the evidence to prioritize content, technical fixes, and service page improvements. The useful launch outcome is a functioning site and a clearer understanding of how people become suitable customers.