Your ads work but your site does not convert: how to tell the difference

Sales are flat. Traffic looks fine. Interest looks fine. The obvious response is to buy more traffic, which pushes more people down the same blocked path and makes the loss bigger. This is the most expensive mistake in Indian D2C, because nothing in the standard reporting flags it clearly.

Here is how to tell a conversion problem from a media problem in an afternoon, and what to do first once you know.

The test that separates the two

Split your metrics into click-side and site-side and compare each against your own trailing baseline, not against an industry benchmark.

• Click-side: CPM, CTR, cost per click, cost per landing page view.
• Site-side: landing page view to product view, product view to add to cart, add to cart to checkout initiated, checkout initiated to purchase.

If click-side metrics are stable and one or more site-side rates have fallen, you have a conversion problem and buying more traffic will not fix it. If CPMs and CPCs have risen while every site-side rate held, you have a media problem. If both moved, you usually have a targeting change that brought in a different, worse-fitting audience, and the fix is in the ad account even though the symptom is on the site.

Ignore industry conversion benchmarks entirely for this. Published averages span store maturity, category, price point and traffic mix so widely that they cannot tell you whether your number is good. Your own number from eight weeks ago can.

Instrument the steps before you change anything

Overall conversion rate is a single number covering five or six independent failures. It tells you that money is leaking, never where. Instrument each step so the drop between them is visible:

• Landing page view
• Product page view
• Add to cart
• Checkout initiated
• Payment attempted
• Purchase, and separately, order delivered

Then segment by device and by traffic source. In Indian D2C the overwhelming majority of paid traffic is on a phone, so a desktop-healthy funnel with a broken mobile step will read as a mild overall decline while being a catastrophe where the money actually is. Always read mobile first.

The step-to-step rates are the diagnosis. The largest single drop is the only one worth touching first, because it is the one where a fix is large enough to be measurable against the traffic you already have.

Fix instability before you test anything

This is the step most teams skip and it invalidates everything after it. Optimisation testing on an unstable site measures the instability rather than the change. If pages fail intermittently, if a step errors for some devices, if an element renders wrongly on a phone, your A/B results are noise wearing a confidence interval.

Butterfly Ayurveda is the case we point to here. Demand was never the missing piece: the brand had technical and conversion roadblocks on-site that were capping sales despite real demand. That is a worse problem than weak demand, because the reporting does not flag it. Traffic looks fine, interest looks fine, revenue simply does not follow, and the usual response is to buy more traffic. The sequence that worked was strict: the technical foundation had to be sound before anything else was worth doing, and only then was the conversion path worth touching. The result was structural rather than numerical: the site stopped capping the demand it was already receiving, and the funnel was cleared end to end rather than patched at a single step.

What "stable" means in practice: no JavaScript errors on the critical path, checkout completing on the three or four device and browser combinations that make up most of your traffic, and page loading behaviour good enough that people do not leave before seeing anything. Google's Core Web Vitals are a reasonable, public definition of the loading, interactivity and visual stability thresholds to aim at, and Google Search Central documents how they are measured on real users rather than in a lab.

Which step to touch first, and why one change at a time

Once the site is stable and instrumented, work the biggest drop. Each step loses people for a different reason, so the fix is different at each one:

Product page

Loses people who are not yet convinced. It needs proof and answers: what it is, why it is different, what it costs to try, what happens if it is wrong. For food and supplements, ingredient detail, sourcing, and honest usage guidance do more than lifestyle photography.

Cart

Loses people who have just met a cost they did not expect. Shipping charges that appear late are the classic. So is a free-shipping threshold set just above the natural basket size, which reads as a penalty rather than an incentive.

Checkout

Loses people to friction, error and forms that ask for more than they need. This is also where payment method matters: UPI-first ordering, saved addresses, and a COD option that is available where you actually want it.

The step immediately before payment

Underrated and expensive. Traffic that reaches the final step has already been paid for in full, so every drop-off there is a total loss on a cost already incurred, unlike a weak impression which costs almost nothing. On Kroslo, the highest-value finding of the whole engagement came out of the funnel rather than the ad account: the step immediately before payment was where the money was going. Fixing that exact step, rather than running a general redesign, was part of doubling ROAS in 25 days on the same budget.

Ship one change at a time. A redesign changes many variables at once and takes far longer to read, which is why it feels productive and teaches you nothing.

What the ceiling actually is

Two numbers from our own work, because "CRO can help" is not a claim anyone should accept without them. Barosi moved on-site conversion from 0.8% to 6.8% over three months alongside a repositioning and a rebuilt site, with monthly revenue going from ₹25k to ₹21 lakhs. Soothys went from 1.3% to 6% in three months, with revenue up 208% and order volume doubled, and its testing surfaced two hero products that now drive over 60% of sales.

Both are large moves and neither came from a single button colour. They came from fixing the reason people were leaving, which was different at each step, on stores that had first been made stable enough to test on.

Halve your funnel leaks and you halve effective CAC. That is the cheapest media buy available to you, and it does not go through an auction.

When it really is a media problem

To be fair to the ad account, some signs genuinely point there: CPM up sharply with site-side rates unchanged; frequency climbing while CTR falls, which is creative fatigue; a sudden CTR jump paired with a conversion-rate collapse, which usually means a new creative is attracting the wrong people; or a targeting change that widened the audience just before the decline. Those are ad-account problems and no amount of checkout work will fix them.

Most of the time, though, the honest answer is that both need work and only one of them is being looked at. The site work is what our CRO and web practice does, usually in the same sprint as the media, and the category-specific version for consumables is on our food and beverage page.