The standard CRO engagement goes like this: an agency or in-house team runs a funnel analysis, identifies where shoppers drop off, and runs A/B tests on those steps. Exit pages get redesign proposals. Cart abandonment triggers a new email sequence. Checkout fields get reduced. Buy buttons get bigger. Shipping cost display gets moved earlier. Each intervention addresses a real friction point and some of them move the metric they target.
The conversion rate overall moves modestly, if at all. And the team feels like they have been thorough - they looked at every step in the funnel. They missed the one that matters most, because it happens before the funnel starts.
The Discovery Step Is Outside the Standard Funnel
Conversion funnel analysis typically begins at the landing page. A visitor arrives from an ad or an email or organic search, and the funnel tracks what happens next. The implicit assumption is that the question "did the visitor find something they wanted to buy?" was already answered by the acquisition channel. The job of CRO is to reduce friction between that implicit yes and the completed purchase.
That assumption fails for a large fraction of sessions. A shopper who arrives from a brand awareness ad or a top-of-funnel blog post may have genuine purchase intent for something in your catalog, but they have not yet found the specific product. They land on the homepage or a category page, and what they encounter is the products your merchandising team decided to feature - which may or may not align with what this specific visitor is actually looking for.
If they find a product that matches their intent, they enter the funnel you have been optimizing. If they do not, they leave - and they register as a bounce or a short session in your analytics, not as a conversion funnel drop-off. The discovery failure is invisible to your standard funnel because it happens before the funnel's first tracked event.
Discovery Rate as a Metric
The metric that captures the upstream problem is discovery rate: the percentage of sessions in which a visitor views at least one product detail page. This is distinct from the add-to-cart rate, which measures what happens after discovery, and from the homepage conversion rate, which blends discovery and purchase intent in a way that obscures the distinction.
Discovery rate tells you whether shoppers are finding products worth looking at. A low discovery rate means your surfacing logic is not matching shoppers to relevant products early in their sessions. A high discovery rate with a low add-to-cart rate suggests the discovery problem is partially solved but the product-shopper fit is still off - they are finding products, but not the right ones. These are different problems with different solutions.
For most DTC catalogs, tracking discovery rate separately from conversion rate reveals something uncomfortable: a meaningful percentage of sessions never generate a single PDP view. Those shoppers came, looked at what was on the surface, did not recognize anything relevant to them, and left. No CRO intervention at the checkout or cart stage can recover them, because they never got there.
Why the Upstream Lever Is Larger
The size of the conversion opportunity in product discovery is related to funnel position. At each step downstream, you are working with a smaller population. The pool of shoppers who added to cart is smaller than the pool who viewed a PDP, which is smaller than the pool who landed on any page. Improving a downstream conversion rate by a few percentage points has a numerically smaller effect than improving the discovery rate by the same amount, because you are working with a larger base.
More importantly, improving discovery quality - not just discovery rate, but the quality of the match between what a shopper finds and what they were actually looking for - compounds through the funnel. A shopper who found the right product is more likely to add to cart, more likely to complete checkout, and more likely to return. The downstream friction reductions you have been testing with A/B experiments are easier to realize when the shopper arrived at the product stage with genuine intent, not a resigned "well, this is the closest thing I could find."
What Category Page Performance Really Shows
Category pages are the main discovery surface for shoppers who cannot find what they want from the homepage. A well-merchandised category page surfaces the right products for each shopper through the category, which can be a large space. If your "home goods" category has 4,000 products and the page shows 24 per load, you are surfacing 0.6% of the category per page view.
Default sort orders - typically bestseller or recency - are aggregate signals that do not account for what the specific visiting shopper is looking for within the category. A shopper who has been browsing a specific price tier and aesthetic register within the category is shown the same ranking as a shopper with different signals. The category page optimization question is not usually "how do we get people to scroll more" - it is "what should we be showing them when they first arrive?"
Search Is a Symptom and a Signal
High search usage is often treated as a positive engagement metric. It is also a signal that browsing surfaces failed discovery. A shopper who uses the search bar is often doing so because they could not find what they were looking for through the standard surfacing. They had enough intent to stay and try an explicit query rather than leaving, which is valuable, but the fact that they needed to search to find what they wanted is worth reading as a failure of the upstream surfaces.
Correlating search usage with prior session behavior reveals a pattern in most catalogs: shoppers who use search within the first two minutes of a session have lower discovery rates from browsing surfaces than shoppers who do not. They went to search because browsing was not working. This is useful information about where the surfacing gaps are.
How to Reorient the CRO Roadmap
The practical shift is to add discovery metrics to the top of the optimization framework before the funnel begins. Track homepage engagement quality (not just click rate, but whether clicks lead to PDP views and whether those PDP views lead to engagement), category page discovery rate by entry path, and the correlation between discovery quality and downstream conversion. These metrics will reveal where the largest leverage is.
Then run experiments upstream. Test different homepage configurations for different traffic sources and see whether discovery rates change. Test different category sort logic for shoppers who have revealed preference signals versus new visitors. Measure whether personalized discovery surfaces produce better discovery rates than static ones, and whether that improvement carries through to conversion.
The checkout UX work is not wasted. Friction reduction at every stage matters. But if you have not measured discovery rate and do not know what fraction of your sessions fail before they enter the funnel, you are optimizing the middle of a problem you have not fully diagnosed. Start at the top.