The monthly report says the site converts at 8%. The sales team says they got eleven inquiries. Both numbers are correct, and they are describing different things, because somewhere in the analytics account a newsletter signup, a PDF download, and a click on a phone number were all marked as conversions alongside the contact form.
This is the most common way a conversion rate stops being useful, and it never announces itself. The number keeps arriving every month, it keeps going up, and it stops predicting anything.
What counts as a micro-conversion?
Any action that shows interest without committing to anything: a newsletter signup, a guide download, a pricing page visit, a video watched to the end, an account created but never used.
They are real signals. Somebody did something deliberate. What they are not is the thing the business is trying to make happen, and the distance between those two ideas is where reporting goes wrong.
Should micro-conversions be in the headline rate?
No. One rate should track one outcome, and the headline rate should track the one the business is paid for. Everything else belongs in the report, just not in that number.
The reason is arithmetic rather than principle. A conversion rate built from several event types is a weighted average, and the weighting is set by whichever event happens most — not by whichever event matters most. A download that fires ten times as often as a form submission will dominate the combined rate no matter how the business feels about downloads.
So the combined number moves when downloads move. A change in the thing you actually care about is a rounding error inside it.
What breaks when you blend them?
Four things, and each one costs a decision.
- Direction stops being readable. Sales down and downloads up produces a flat line, which reports as "no change" in the month something significant happened.
- Channel comparison collapses. Traffic sources differ enormously in how far down the funnel they arrive; a blended rate ranks them by how many free things people took.
- Tests optimize the wrong thing. A page change that trades a form submission for a download wins on the blended rate and loses on revenue.
- The number inflates quietly. Every new tracked event raises the rate without anything improving, which makes year-over-year comparison meaningless unless somebody kept notes.
A blended conversion rate is weighted by frequency, not by value. Adding a cheap high-volume event to the count does not make the metric more complete — it hands control of the number to the least important thing in it.
Are micro-conversions worth tracking at all?
Yes, and the case for them is strong enough that deleting them is the wrong correction. They answer questions the primary rate cannot.
They tell you where a funnel loses people, which page types do the persuading, and whether traffic arriving from organic search is engaging at all before it leaves. On a site with a long sales cycle, a micro-conversion is often the only signal that arrives inside a reporting month — the deal that closes in March was sourced by something that happened in November.
The correction is to separate them, not to remove them. Track them, name them, report them on their own line, and leave the primary rate alone.
Which events belong where?
Sort each tracked event by what it predicts rather than by how it feels.
| Event | What it predicts | Where it belongs |
|---|---|---|
| Form submission, checkout, booking | Revenue directly | Primary conversion rate |
| Phone call over 30 seconds | Revenue directly | Primary conversion rate |
| Quote or demo request | Revenue directly | Primary conversion rate |
| Newsletter signup | Future contact, weakly | Secondary line |
| Guide or PDF download | Research stage interest | Secondary line |
| Pricing page visit | Intent, on some sites strongly | Secondary line |
| Video completion, scroll depth | Engagement with one page | Page diagnostics only |
| Time on page, pages per session | Very little on its own | Page diagnostics only |
The third column is the whole exercise. Nothing gets deleted; things get filed.
When is a micro-conversion the right primary metric?
When the real conversion is too rare to measure. A page that produces two sales a month cannot be optimized against sales, because two is not a number you can compare to three.
In that situation, promoting a frequent upstream action to the primary metric is the only way to get a readable signal — but it comes with an obligation. You have to check, at least quarterly, that the upstream action still correlates with the outcome. A stand-in metric that has drifted loose from what it stands in for is worse than no metric, because it is trusted.
The check is not complicated: look at what share of the people who took the upstream action went on to convert, and watch whether that share holds. If it falls while the micro-conversion count rises, the page got better at giving things away.
Adding a new tracked event mid-year and comparing the rate to last year's. The line jumps, somebody credits the work that shipped that month, and the change is entirely definitional.
How do you report both without confusing anyone?
Name each rate after what it counts, and never use the bare phrase "conversion rate" in a document where two of them appear.
A lead rate, a subscriber rate, and a download rate on three lines are readable by anyone. A single conversion rate with a footnote explaining what it includes is not, because nobody reads the footnote and the number ends up in a slide without it.
Two habits keep this stable. Write the event list into the report itself, so the definition travels with the number. And when an event is added or removed, annotate the date on the chart — a definitional change that is not marked looks exactly like a performance change, and six months later nobody can tell which it was.
What about engagement metrics?
Keep them out of the conversion count entirely. Scroll depth, session duration, and bounce rate describe behavior on a page rather than a decision about the business.
They are useful in exactly one place: diagnosing a specific page that is getting traffic and not converting. That is a real job, and it is a different job from reporting. The same applies to click-through rate — it measures whether a result got picked in the first place, which happens before any of this.
Open your analytics account and list every event currently marked as a conversion. For each one, ask whether the business would notice its absence in a revenue report. The ones that fail that question are not worthless — they are just not what the headline number should be counting, and moving them to their own line is a thirty-minute change that makes every report after it mean something.