Why did conversions fall after traffic shifted to different landing pages?
Short answer
Your blended conversion rate is a weighted average, so it moves when the weights move even if no page changed. Recompute this period's conversion rate using last period's traffic distribution. If the recomputed number matches your old rate, the mix explains the entire decline and nothing on your site got worse.
Work the arithmetic once
This is easier to see with numbers than to argue about.
Last period. 10,000 sessions on pricing at 5%, 10,000 on blog at 0.5%. That's 500 plus 50, so 550 conversions from 20,000 sessions. Blended rate 2.75%.
This period. Traffic grew to 24,000, but the growth all landed on the blog. Pricing still gets 10,000 at 5%, blog now gets 14,000 at 0.5%. That's 500 plus 70, so 570 conversions from 24,000 sessions. Blended rate 2.38%.
Conversions went up. The rate fell 13%. Neither page changed by a single basis point.
If your reporting leads with conversion rate, this reads as a site that got worse. If it leads with conversions, it reads as modest growth. Both are computed correctly from the same data.
Run the mix-adjusted calculation
The decomposition takes fifteen minutes and settles the question.
1. Pull sessions and conversions per landing page for both periods. Top 50 pages covers most sites.
2. Compute conversion rate per page for each period.
3. Recompute this period's blended rate using last period's traffic weights. Apply the old distribution to the new per-page rates.
4. Compare three numbers:
| Number | What it tells you |
|---|---|
| Last period's actual blended rate | Your starting point |
| This period's actual blended rate | Where you ended |
| This period's rate at last period's weights | What the rate would be if mix hadn't moved |
If the third number is close to the first, mix explains everything. If it's still well below, some pages genuinely got worse and the gap between the second and third numbers is how much of the decline is real.
That single calculation is what turns a month of debate into one sentence.
Why the mix moves without anyone deciding
A blog post took off. One piece ranks well, brings 8,000 low-intent sessions, and dilutes the average.
A commercial page lost position. Its share of the mix shrinks and the high-converting weight goes with it.
Paid was paused. Paid traffic usually lands on high-converting pages, so pausing it changes the organic mix too.
Seasonal intent rotation. Some months skew informational in most categories.
AI answers absorbed the top-funnel clicks unevenly. This one is newer and worth checking specifically.
The uneven absorption that reshapes your mix
AI answers don't remove clicks uniformly. They take informational clicks heavily and commercial clicks barely.
Pew Research found that when an AI summary appears, users click a source link in only about 1% of visits, and our own state of AI search research, covering more than 22,000 answers, found Google AI Mode returns a citation-rich answer in 97.4% of matched B2B queries.
The counterintuitive result is that this can push your blended conversion rate up while total conversions fall, because the low-converting pages lose traffic first. A rising conversion rate alongside falling conversions is the signature, and it's frequently misread as a CRO win.
Check both metrics together or you'll congratulate yourself for a traffic loss.
What to report instead
Lead with conversions, not conversion rate. The absolute number is what the business cares about and it doesn't move on mix alone.
Report conversion rate per intent bucket, never blended. Four numbers instead of one, and each is interpretable.
Track traffic mix as its own metric. Share of sessions per bucket, monthly. It's the leading indicator for everything above.
Automate the decomposition
The reason this analysis gets skipped is that it needs per-page rates for two periods and a weighted recomputation, which is a pivot table nobody rebuilds monthly.
The funnel-coverage recipe classifies pages by funnel stage automatically, so the bucketing is a node rather than a manual pass. GSC Page-Keyword Breakdown supplies the queries behind each page so the classification reflects real search intent rather than your original intent when publishing.
A monthly agent:
Start (schedule, 1st of month) → funnel-coverage recipe → GA4 AI Traffic Overview and AI Landing Pages → GSC Page-Keyword Breakdown (top 50 pages) → HubSpot Get CRM Objects (conversions by landing page, both periods) → Code (per-page rates for both periods, then recompute current rate at prior weights, and output the three-number comparison) → Prompt LLM (state what share of the change is mix versus per-page performance, and name the pages driving each) → Excel export → Send Email to CMO.
The three-number output is the deliverable. Blended rate fell from 2.75% to 2.38%. At last period's traffic mix it would be 2.71%. Mix explains 92% of the decline, and these three pages account for the rest. That sentence ends the conversation instead of starting it.
FAQ
Related answers
- Why did my leads decline while rankings and traffic stayed stable?
- Why is my SEO traffic growing while revenue stays flat?
- Is bot traffic masking a real conversion problem?
- How do I find the most likely cause of a traffic anomaly?
Want the mix-adjusted conversion number every month? Start a free Analyze AI trial and connect GA4, GSC, and your CRM.