A jewellery brand I worked with had six months of beautiful conversion-rate charts. February to August, the line went up and to the right almost without a wobble. The team put it in every review deck. Somebody used the phrase "compounding gains."
Revenue over the same six months was flat.
Not down. Flat, to within a couple of percent. And nobody could explain it, because every metric anyone was looking at said the site was getting better at selling.
What was actually happening
Conversion rate is a fraction. Orders divided by sessions. When a fraction goes up, one of two things happened: the top got bigger, or the bottom got smaller.
Everyone assumes it's the first one. It was the second.
Paid reach had been quietly collapsing since March — budget had been trimmed twice, and a campaign restructure had killed the prospecting audiences. What was left arriving on the site was almost entirely people who already knew the brand: returning visitors, direct traffic, branded search. People who were coming to buy something they'd already decided on.
Of course they converted well. They always did. The site hadn't got better at persuading strangers — it had simply stopped receiving any.
| Month | Sessions | Conversion rate | Revenue |
|---|---|---|---|
| February | 100,000 | 0.60% | ₹13.2 L |
| May | 52,000 | 1.15% | ₹13.1 L |
| August | 29,000 | 2.05% | ₹13.1 L |
Conversion rate up 240%. Sessions down 71%. Revenue: unchanged.
Read that table again, because it is the whole lesson. Every number in the middle column is an improvement. The business behind them was dying, slowly, and the improving metric was the thing hiding it.
Why this fools experienced people
Because a rising conversion rate is one of the few metrics that feels unambiguously good. Nobody interrogates it. If your cost per click goes down, you check whether traffic quality dropped. If your average order value jumps, you check whether one wholesale order skewed it. But when conversion rate rises, everyone just nods.
It's also the metric that gets reported without its denominator. "CVR is up 40% month on month" is a complete sentence in most marketing reviews. It shouldn't be. It's half a sentence.
A ratio can improve because the thing you want went up, or because the thing you measured it against went down. Those are opposite situations and the ratio looks identical in both.
The check that takes ninety seconds
Whenever a rate improves, look at the two raw numbers that made it. Not the rate. The numerator and the denominator, side by side, in absolute terms.
For conversion rate, that's orders and sessions. Put them in three columns with revenue and you can immediately see which of the four situations you're in:
- Sessions up, orders up faster — genuinely working. Rare and worth understanding.
- Sessions flat, orders up — the site or the offer got better. This is what CRO is supposed to produce.
- Sessions down, orders flat — you're harvesting existing demand. Feels fine for a quarter. Isn't.
- Sessions down, orders down, rate up — the situation above, further along. Usually noticed only when the quarter closes badly.
The third and fourth cases both produce a rising conversion rate. Both are bad. Neither is visible if you only look at the percentage.
The second check: new versus returning
The faster diagnostic is the share of sessions from new users. In the case above, new-user share had gone from 68% to 31% while conversion rate climbed. That single number would have caught it in March instead of August.
It works because it separates the two businesses you're actually running. Acquiring strangers is one business. Converting people who already know you is a different one, with different economics and a hard ceiling — you can only sell to your existing audience so many times. When the second business starts to dominate your traffic mix, your blended metrics improve and your growth stops. Both at once.
So put new-user share next to conversion rate on whatever weekly sheet you keep. If conversion rate is rising while new-user share is falling, you don't have a conversion win. You have a demand problem wearing a conversion win as a costume.
What we did about it
Nothing clever, in the end. We rebuilt the prospecting layer that had been switched off, accepted that conversion rate would fall — it went from 2.05% back to roughly 0.9% — and watched revenue grow 60% over the following two quarters.
That was the hardest part of the whole engagement, incidentally: getting agreement to deliberately make a headline metric worse. It looks like failure on a dashboard for about six weeks. There's no way around that except explaining the arithmetic to whoever reads the dashboard, before you start, in writing.
If you take one thing from this
Ratios are conclusions, not evidence. Conversion rate, ROAS, click-through rate, cost per acquisition — every one of them can move for a reason that's the opposite of what it appears to mean.
Before you celebrate any of them, look at both numbers that made it. It takes ninety seconds and it is the single highest-value habit I know of in this job.