Why Did My SaaS Conversion Rate Drop? A Practical Investigation Order

A conversion-rate drop is a ratio problem, not a diagnosis. The numerator may have fallen, the denominator may have grown with lower-intent traffic, or the tracking event may have stopped firing correctly. Changing the product before separating those cases can make the problem worse.

1. Verify tracking health first

Confirm that the primary conversion event still arrives, the tracker is active and the relevant pages are being measured. A broken signup event can look exactly like a conversion crisis. Analytics should prove the measurement system is healthy before it recommends product work.

2. Check whether traffic quality changed

If sessions rose while conversions stayed flat, conversion rate falls even though the product itself may be unchanged. Compare source and campaign mix with the previous period. A viral post, broad paid campaign or low-intent referral can change the denominator quickly.

3. Compare landing pages

A new source often enters through a different page. Look for landing pages with meaningful traffic and unusually low primary-goal conversion. That tells you whether message-to-intent mismatch is concentrated at the top of the funnel.

4. Find the largest funnel drop

Build an ordered funnel from the events you already collect. Compare the step conversion rates. If the biggest deterioration happens before checkout_started, the investigation is different from a drop between checkout_started and purchase.

5. Segment only after finding the weak step

Device, country and source segmentation are powerful when you already know where the issue lives. Starting with dozens of segments creates false leads. First locate the broken step, then ask whether one segment explains most of the change.

How Pulse approaches the problem

Pulse's Decision Feed is designed to surface material traffic and conversion changes before you manually compare every report. Funnel Lab then lets you inspect ordered event drop-off without creating a saved funnel first.