
Traffic dropped after your website launch: what to check in the first 48 hours
Separate a tracking problem from a search problem, identify the affected pages and give your team a clear recovery brief.
Written by Rob KnottA new website should be a step forward. When your traffic chart falls after launch, you need to know whether customers have stopped finding you, your measurement has changed, or both. Changing a dozen things at once makes that answer harder to find.
Your first 48 hours are for establishing the incident, containing confirmed faults and giving the right people evidence they can act on. They are not a deadline for rankings to recover. This sequence is a practical investigation plan, with a fictional example you can adapt to your own website.
First two hours: establish what actually fell
Write down the exact launch time, the first observed change and the systems showing it. “Traffic is down” could mean fewer Google Search clicks, fewer measured sessions, fewer enquiries or a lower total in a dashboard that now uses different filters. Those are different incidents.
Compare complete periods with matching weekdays, markets, devices and channels. A partial day, a holiday or delayed data can create an apparent collapse. Keep a copy of the original exports and filters. Search Console and analytics measure different things, so their totals will not match exactly; look for a coherent pattern rather than forcing equality.
Ask the developer for the release notes and the analytics owner whether tags, consent settings, domains, events or reporting definitions changed. If all measured channels drop abruptly but orders remain steady, investigate tracking promptly. That pattern is a clue, not proof that search is unaffected.
| Observation | Start here | Do not conclude yet |
|---|---|---|
| Analytics sessions fall; Search Console clicks look steady | Check tagging, consent, attribution and complete reporting periods | That nothing else broke |
| Search clicks fall mainly on one directory | Compare old/new URLs and templates in that directory | That the whole domain has been penalised |
| Clicks look steady; enquiries fall | Test forms, checkout, tracking and the landing-page journey | That SEO visibility caused the loss |
| Search declines before launch | Extend the timeline and check demand, updates and earlier releases | That the redesign explains every change |
If the decline predates launch, use the wider traffic-drop investigation to assess other explanations.
Hours 2–6: find the smallest affected group
Break the change into page groups you can investigate: products, categories, services, articles and country folders. Within each group, separate pages that kept their URLs from pages that moved. Compare branded and non-branded queries where the data supports it.
Build two small URL samples. One contains valuable pages with a clear deterioration; the other contains similar pages that appear stable. Include different templates and countries if relevant. A comparison group helps you avoid calling every difference a fault.
Record old URL, intended new URL, current response, landing-page content and business importance. Start with pages associated with meaningful enquiries or sales, while recognising attribution gaps. Use the old sitemap, crawl, analytics exports and redirect map to reconstruct what existed. If you have no baseline, state that limitation instead of inventing one.
Hours 6–12: check the live journey
Open the old address and follow its actual redirect chain. Does it reach the intended replacement, or a generic category, homepage, error or loop? Check that the final destination works without your staff login. Test the public version rather than accepting a clean staging screenshot.
For important pages, inspect the HTTP response, indexing directives, canonical URL, visible main content and internal links. A canonical is a signal about the preferred URL, not a command that guarantees Google's choice. Compare the declared canonical with inspection evidence where available. Check whether navigation or template changes have made the page harder to discover.
Look at both the response HTML and rendered page where JavaScript is involved. A successful browser view on your computer does not establish that every important user agent receives the same usable content. Ask the team to investigate access rules, CDN behaviour and server errors with appropriate logs.
- Confirm the intended destination and test the full old-to-new route.
- Check accidental noindex directives and crawler access restrictions separately.
- Compare meaningful content, headings, navigation and links with the previous version.
- Inspect a stable comparison page before assuming a template difference explains the loss.
- Save time-stamped evidence before changing the affected configuration.
For the URL-level testing method, test your redirect map.
Worked example: a category migration with two separate faults
Imagine a fictional retailer that launches on Monday. Its analytics report shows a large session decline. At the same time, the team receives complaints about an old category link. Those observations justify two investigations rather than one universal explanation.
The analytics owner finds that a consent change has altered measurement. Separately, an old category URL now redirects to the homepage even though a relevant replacement category exists. Several URLs share that faulty rule. The team corrects the mapping, tests every URL governed by the rule, checks the new category and updates its internal links.
The next day's cleaner analytics chart cannot be used as evidence of recovered search visibility: measurement changed during the investigation. The incident log keeps the tracking repair and redirect repair separate. Search recovery is assessed later using consistent complete periods and the affected page group.
Hours 12–24: agree containment and ownership
Prioritise confirmed failures that stop customers or search systems reaching important content. Give each issue an owner, a proposed correction, a test and an agreed review time. Keep speculative opportunities in a separate backlog so they do not obscure the incident.
A rollback is a business and engineering decision. Before proposing one, establish whether the previous deployment is safe to restore, whether orders or content would be lost, and whether a targeted correction is faster and less disruptive. Do not repeatedly switch between releases in response to a live traffic chart.
Freeze unrelated changes where practical. Keep necessary security and operational work moving, but record it. You need a timeline that explains which changes happened and which evidence supports them.
| Field | What to capture |
|---|---|
| Observed problem | Metric, page group, period and comparison |
| Evidence | Export or test, timestamp and exact URL |
| Working explanation | What the evidence suggests and what would disprove it |
| Correction | Specific change, owner and dependencies |
| Acceptance test | Expected response, destination and visible content |
| Follow-up | Who rechecks technical behaviour and who reviews later performance |
Hours 24–48: verify fixes and set the next review
Repeat the same tests against the public deployment. Check that the specific fault is resolved and that adjacent pages still work. Save both results. “The redirect now reaches the correct page” is a defensible technical outcome; “SEO has recovered” requires performance evidence over a suitable period.
Agree a review cadence that fits the scale of the move and your data volumes. Google's site-move guidance makes clear that processing takes time. You can resolve a technical error quickly while its search effects take longer to assess. Monitor important URL groups, discovery and indexing evidence, search performance and genuine business outcomes.
Finish the first 48 hours with a short statement of what is confirmed, what was fixed, what remains uncertain and what happens next. If you cannot explain the pattern, widen the investigation rather than presenting the nearest plausible issue as the cause.
When independent help is useful
Bring in support when the affected URLs span several systems, ownership is disputed, important revenue is exposed or the team needs a second opinion before a major intervention. You do not need to hand over every aspect of SEO to get a focused diagnosis.
For an initial discussion, prepare your launch date, a description of what changed, several affected URLs and the reports showing the decline. Access can be arranged securely once the scope is agreed. A useful engagement should leave your team with a supported diagnosis, prioritised actions and acceptance tests, not just another list of warnings.
Key takeaways
- Establish which metric changed before choosing a fix.
- Compare affected and stable page groups, then test the public journey.
- Separate tracking repairs from search repairs in your evidence.
- Verify implementation first and assess performance over a suitable period.
