How to Diagnose a Drop in Website Conversions

Start by not changing anything

A drop in conversions creates pressure to act. Leadership notices, someone suggests the ads have stopped working, someone else blames the new homepage, and within a week three things have been changed at once. If conversions recover, nobody knows why. If they do not, the original cause is now buried under the fixes.

The diagnostic order below starts with the cheapest and most common explanations and moves toward the expensive ones. The most common causes live in the first two steps. Write down what you check as you go; the record is useful the next time this happens, and it will happen again.

Before starting, define the drop precisely. Which conversion, from which date, by how much against what comparison? “Leads are down” is a feeling. “Quote form submissions recorded in GA4 fell sharply from the week of the 12th, compared with the same weeks last year, while phone calls held steady” is a starting point.

Step one: is the drop real, or is it measurement?

Many conversion drops turn out to be tracking failures. The business is getting the same inquiries; the analytics simply stopped counting them. Check this first because it is quick, and because every other step depends on trusting the data.

• Compare against a second source. Count form submissions in the form tool, the inbox or the CRM for the same period. If the CRM shows no drop and GA4 does, you have a measurement problem, not a marketing one.
• Look for a sudden step. Real behaviour changes tend to be gradual or follow an event. A cliff on a specific day usually means something technical changed that day.
• List recent changes. Website releases, form plugin updates, a new booking engine, a redesigned thank-you page, a Tag Manager publish, a CMS update. Ask the developer and the agency, not just the marketing team.
• Check the consent banner. A new or reconfigured banner, or a change in consent mode, can sharply reduce what analytics records without changing anything about real inquiries.
• Check the key event definitions. Was an event renamed, a key event unmarked, or a trigger changed? Did someone fix a duplicate that was previously inflating the numbers? A drop caused by correcting double-counting is good news.
• Walk the conversion yourself. Complete the form on desktop and on a phone with GA4’s debugging tools open and confirm the event fires once, on success.

Our GA4 setup and audit checklist covers each of these checks in more detail, and Tag Manager vs direct tags explains why tracking so often breaks after a redesign.

Step two: did the traffic change?

If the drop is real, the next question is whether fewer people arrived, different people arrived, or the same people stopped converting.

• Volume by channel. Did organic, paid, referral or email traffic fall? A drop concentrated in one channel narrows the search immediately.
• Paid campaigns. Was a campaign paused, a budget cut, a bid strategy changed, a payment method declined? Did the ad platform shift spend toward a cheaper audience or placement that produces clicks but not inquiries?
• Organic search. Check Search Console for lost rankings or impressions on commercial queries. A core algorithm update, a site migration with missing redirects or a competitor’s new content can all move this. If you recently relaunched the site, compare against our website migration SEO checklist.
• Traffic mix. Conversion rate can fall because of a surge of low-intent traffic: a blog post that went viral, a press mention, a bot spike. Total conversions held, but the rate looks worse. Check both counts and rates.
• Geography and device. A shift toward mobile, or toward visitors from outside your service area, changes conversion behaviour even if nothing on the site broke.

Step three: do the forms and pages still work?

If traffic is steady but conversions fell, something between arrival and inquiry has broken or become harder. Test it the way a customer would.

Forms and booking

• Submit every important form on several devices and browsers. Do the submissions arrive where they should, promptly?
• Check spam filtering. An overzealous spam setting, captcha change or email deliverability problem can silently discard real inquiries.
• Check third-party booking, payment and scheduling tools for outages, changed embed codes or new steps added by the vendor.
• Check that confirmation emails still send. A visitor who receives nothing may submit again, or give up and call a competitor.

Pages and experience

• Compare conversion rate by landing page and by device. A drop isolated to mobile Safari, for example, points to a rendering or script problem.
• Check page speed on the key landing pages. A new script, a heavy image or a chat widget can slow pages enough to matter.
• Look for content changes: a removed phone number, a new pop-up covering the form, a price or offer change, pricing that is no longer visible, a call to action that moved below the fold.
• Session recordings and heatmaps (with appropriate consent) can show where people hesitate or abandon.

The conversion rate optimization playbook covers the page-level side in depth.

Step four: is it seasonal, or the market?

Some drops are not problems at all. Compare against the same period last year and the year before, not just the previous month.

• Seasonality. For tourism, recreation, hospitality, trades and education across Atlantic Canada, demand follows the calendar more than any campaign. The seasonal marketing playbook describes how those cycles shape planning.
• Calendar shifts. A long weekend, school break or holiday falling in a different week can move demand between months.
• Weather and events. A mild winter, a storm, a road closure or a local event can change behaviour for weeks.
• Market and competition. Interest rates, a competitor’s promotion, a new entrant in a small market, or a change in public funding cycles for organizations that depend on them.
• Your own operations. Did prices change, a service pause, a location close or a waitlist open? Customers notice.

If the drop matches last year’s pattern, the answer may be to adjust expectations and targets rather than tactics.

Step five: fix one thing, and test it properly

Once you have a likely cause, change it deliberately. If tracking broke, fix it and annotate the gap so nobody later reads it as a performance drop. If a form broke, fix it and check whether inquiries recover. If the cause is less certain, such as a page change or a shift in audience, treat the fix as an experiment.

Designing a test that tells you something

• One change at a time, or at least one change per page or campaign, so the result can be attributed.
• A written hypothesis. “Restoring the phone number above the form will recover mobile inquiries on the service pages.”
• A defined measure and period, decided before the test starts, long enough to cover normal weekly variation.
• A comparison. A true A/B test splits traffic between versions. Where traffic is too low for that, which is common for smaller sites in small markets, compare against a matched period or use a holdout (a region, a campaign, a set of pages left unchanged).
• Humility about small numbers. With a few dozen conversions a month, small differences are noise. Look for clear, sustained changes rather than declaring a winner after a week.

Prevent the next one

Most of this diagnosis becomes faster with a few standing habits:

1. Annotate site releases, campaign launches, consent changes and tracking fixes in your reporting, so changes line up with dates.
2. Set up an alert for an unusual drop in primary conversions, so you hear about it in days rather than at month end.
3. Reconcile analytics against the CRM or form tool monthly.
4. Test primary conversions after every website release.
5. Keep an executive dashboard that compares against the same period last year, so seasonal dips are not mistaken for crises.

When we are asked to look at a sudden drop as part of our analytics and measurement work, we follow this order every time. It is not exciting, but it usually finds the answer before anyone has to rewrite a single ad.