What CRO actually covers, and who can own it

2026-11-09

CRO keeps turning up in strange places on the org chart. Under digital marketing. Inside trading. With one manager who owns a testing tool and none of the things being tested.

• Ecommerce
• CRO
• Director

I keep finding CRO in odd places on the org chart. Sometimes it's a line in the digital marketing team's objectives. Sometimes it has been handed to trading, on the reasoning that trading already looks at the numbers every Monday. Occasionally it belongs to one CRO manager with a testing tool, a backlog and no authority over anything the backlog touches.

Each of those arrangements gives one function responsibility for a number it can only partly move. Conversion rate is a fraction. Digital marketing decides much of the bottom half, because it decides who arrives. The site, the range, the price, the stock and the delivery promise decide the top half. And the customer's previous order, which none of those teams was watching when it reached the doorstep, decides whether they come back to be counted at all.

I wrote about part of this in September, when I argued that CRO shouldn't live inside marketing. This is the longer version: what CRO is made of, piece by piece, how much of it comes down to the plain quality of the website, and why I think only one role can hold the whole thing.

Conversion starts in the ad

By the time someone's even got to the product page, a lot of the conversion rate has already been determined by the media plan. Broaden the keywords, push reach on social, run some prospecting into a cold audience and you've changed who's in the denominator.

Contentsquare's 2026 benchmark (99 bn sessions across ~6,000 sites) has returning visitors converting at 2.9% and new visitors at 1.7% with 52.8% of traffic being return traffic. Those are cross-industry averages and a conversion in that data doesn't necessarily equate to a purchase, so take the following as pure arithmetic. At that mix, the blended rate is 2.33%. If you move the return traffic to 45% it drops to 2.24% (a 4% drop). Nobody touched the site. In a weekly trading meeting, that 4% is invariably pinned on whatever it was that launched most recently.

This is mix shift and it's behind more confused conversations than it should be. The team that sees only the blended figure goes off and starts to fix pages that were working perfectly well. The team that's changed the traffic reports a lower cost per click and a good month. Each of them is looking at their own number and reading it correctly.

The ad also writes the first line of the customer's expectation. Pirolli and Card's information foraging theory (1999) is still the theory that Nielsen Norman Group and Baymard use (the language they employ). It's about people following the cues that tell them they're on the right track. Those cues are called 'information scent'. An ad for a navy parka at a stated price that lands on a general outerwear grid breaks the scent and the user goes back to look elsewhere. The site will record a bounce for a loss that originated in the ad account and the revenue per visitor will be hit wherever the blame ends up.

The quality of the site does most of the work

Ask a room what makes a website convert and someone will say speed within 30 seconds. Speed matters. A slow response to a tap makes people doubt it registered, so they tap again or give up, and the damage falls hardest on mobile. It's also the one part of site quality that comes with a score attached, which is why it gets so much of the attention. Everything else has no dashboard and tends to belong to nobody.

Information hierarchy is the textbook example. People scan to save effort. The F-pattern was first reported by the Nielsen Norman Group in 2006 and they confirmed it in both desktop and mobile environments in 2017. When there are no strong visual cues on a page, people's attention is going to be in the top left and they're going to miss everything else. That's what they call interaction cost. Put the delivery date, the fit note or the returns info in a section that the eye doesn't go to and you've got the information on the page, but the customer doesn't have it. Analytics won't pick it up. The user simply leaves slightly less certain than they needed to be and on something they can't touch before they pay, that's often enough to stop them.

I've left the checkout out of this on purpose, because it already gets most of the airtime. The pages before it get far less, and Baymard Institute's benchmarks suggest they're in poor shape. Baymard scores leading US and European retail sites by hand against its own usability guidelines. In its 2026 product page benchmark of more than 155 sites, 52% of desktop sites, 62% of mobile sites and 64% of apps rate mediocre or worse. For homepage and category navigation, also 2026 and more than 180 sites, it's 61% of desktop and 65% of mobile. For product lists and filtering, from 2025 and more than 170 sites, 58% of desktop and 78% of mobile. These are large retailers with proper budgets.

Underneath the headlines, the failures are specific enough to be embarrassing. 67% of sites give no estimate of total order cost near the buy section. 57% don't use buttons for size selection, which hides what's out of stock until the shopper has already got interested. 44% don't show or link the returns policy from the main product content, though 60% of shoppers in Baymard's survey work look for it there. 37% have no in-scale image, while 42% of users in testing try to judge size from the photos. 80% show fewer than three thumbnails per item in product lists. 51% are missing at least one of five basic filters: price, rating, colour, size and brand.

Follow one of those through to the money. Baymard finds 79% of sites use categories where filters should do the job, so slim and skinny jeans live in separate places. A shopper who'd consider both can't see them in one list, has to hold the first set in memory while looking at the second, and at some point stops. Baymard's own conclusion is that some users will never find a suitable product. The stock exists and has been paid for. Trading sees a slow line, reads it as weak demand and marks it down, when part of the problem was that the customer never got to it. That's sell-through and margin, caused by a taxonomy decision made years ago by somebody who has probably left.

A caution on all of this. Baymard's ratings measure conformity to its guidelines. They don't report what each failure costs a given site in conversion, and Baymard sells audits. I read the benchmarks as a map of where to look first.

The full list is longer than most job descriptions

When people say CRO, they tend to mean A/B testing. Testing is the measuring instrument. This is what's being measured, as I'd write it up on a wall.

Technical performance. Load time, response to taps, server latency and the sheer weight of every script that was added once and then just completely forgotten about.

Information hierarchy. What comes first on each template, what is pushed down the page and what the eye is led to in terms of the information that determines the decision.

Navigation and taxonomy. Category names that are the ones the customers would actually use, menus that don't just plain overwhelm (Baymard's participants were beginning to struggle with more than about 10 subcategories), and a very clear sense of where you are.

Product lists, filtering and sorting. Which products appear first, which filters exist, and how much a listing tells you without a click.

Product content. Images at scale and on a body, copy that answers the real question, and size selection that shows what's available.

Product data. Attributes that are complete and correct, since filters, feeds and increasingly AI assistants are only as good as what's in them.

Interaction design. The number of actions a task takes, accordions and tabs that hide what people came for, error states, and whether saving an item demands an account first.

Mobile ergonomics. Thumb reach, tap targets, what a sticky bar covers and how much of the first screen is visible.

Accessibility. WCAG conformance and, for many businesses selling into the EU, a legal requirement under the European Accessibility Act from June 2025.

Price and promotion presentation. Reference prices, unit pricing, the total the customer will really pay, and how often the sale banner has cried wolf.

Trust. Reviews and how you answer the bad ones (89% of sites in Baymard's benchmark don't), who you are, how to reach you.

Stock and availability. Whether what the site shows is really in the building, and what happens to a page when it isn't.

Delivery promise. Date, cost, cut-off and collection options, shown early enough to count.

Returns. The policy, how easy it is to use, and how fast the refund arrives.

Customer service. Whether a question asked before purchase gets answered before the customer goes elsewhere.

Range and merchandising. Whether the right products are there at price points the traffic can afford.

Localisation. Language, currency, landed cost and sizing conventions in every market you claim to serve.

Stability and QA. Device testing, release discipline, third-party apps, and how many banners and pop-ups stack up on a phone before the product appears.

Measurement. Tracking that survives consent laws, analytics people trust, and tests with the amount of traffic to mean anything.

Post-purchase communication and retention. Order updates, the first email after delivery, CRM, and a reason to place a second order.

Count the owners. In most of the businesses I've seen, it's a fairly standard list: digital marketing, trading, design, IT or an agency, merchandising, operations, customer service, legal and finance. UX is the one that gets the most interesting treatment in that it's the one that's most often just given to whoever is available. That could be brand, a freelance designer, IT, the person who picked the theme. It's the one area that very rarely makes it to the Ecommerce Director's desk, despite the fact that the layout is absolutely fundamental to what the customer sees of everything else on that list.

Operations converts too

A delivery date on a product page is a double-edged sword. On the one hand, it reduces risk and so increases conversion. On the other hand, it becomes the benchmark against which the purchase will be evaluated. That's expectancy disconfirmation. Your level of satisfaction is going to be a function of how the outcome measures up against the expectation. If you miss the date, then the promise that helped conversion this month is gone in the next period.

The best evidence I can find at the moment is from Amorim, Eng-Larsson and Rooderkerk in the Journal of Retailing (2025) using data from a major omnichannel grocer. A failed fulfilment resulted in a 7.22% average delay to the next order and substitutions (which did recover the sale) were often worse in terms of the delay. That's in the online grocery space, which is a high-frequency category, so I wouldn't take that number into fashion at all. The direction is consistent with the older work though (Rao, Griffis and Goldsby in a 2011 study which is now 15 years old) who looked at an online retailer of printed matter and found that a delayed order reduced the frequency with which previously loyal customers ordered and the amount they spent.

Stock accuracy works the same way. A product that shows as available and can't be picked produces an order, a conversion on the dashboard and, a day later, a cancellation email. The conversion rate went up that day. A CRO programme measured only on the session would count it as a good result.

The order is where the next session starts

As acquisition gets harder, the people who have already bought are going to be a bigger part of your rate. Because returning visitors convert better, every customer you keep improves the mix you'll be measured on later. The Contentsquare 2026 data bears this out: new visitor traffic was down 6% versus 2% for return visitors and new visitor conversion was down 7.9% versus 4.1% for return visitors. As acquisition gets harder, the existing customer base is going to be increasingly important.

What brings them back is largely decided after the sale: whether the parcel arrived when you said, whether the item matched the page, how the return went, what you sent them next. Conversion, as usually reported, can't see any of it. It also rewards the wrong behaviour. A team measured on session conversion has every reason to discount its way to a first order from someone who will only ever buy at that price. Goodhart's law applies neatly: once the rate becomes the target, it stops telling you much about the health of the business.

I'd put three measures in front of whoever owns this: revenue per visitor, contribution per session and the percentage of new customers who go on to buy again in the 12 months following their first purchase. None of those can be improved for any length of time by one department on their own.

Why marketing and trading can't hold it

I have a lot of respect for both functions. A good digital marketing lead makes the bottom of the funnel cheaper and better qualified. A good trading manager keeps the right product in front of the customer at the right price this week. Their scope stops where it stops by design and to ask either of them to be responsible for the whole rate is to hold them to account for things they simply can't touch.

The clash is very evident on the page. Baymard found that in their testing, intermediary category pages performed very badly when promotional content was used to push the subcategory navigation off the page. 38% of sites simply don't have subcategories as the primary content on those pages. Trading wants that space for this weekend's offer, usability wants it for getting people to the right products and both are absolutely right from their own perspective. Someone has to make the call and that person is going to need to look at revenue per visitor for both of those outcomes and be able to account for the result.

CRO specialists still matter, and I'd want one. Testing discipline is hard and most businesses underinvest in it. The specialist simply needs a backlog drawn from the whole journey, and a sponsor who can make operations, IT and marketing act on what the tests find.

Why it ends with the Ecommerce Director or Head of Ecommerce

The title depends on the size of the business. It's a Head of Ecommerce in a smaller company and an Ecommerce Director in a larger one. What counts is the remit: under the right structure, that person answers for the customer's full path, from the acquisition spend that shapes who arrives, through the site and its UX, the trading plan and the delivery and returns promise, to the second order. No other role in the business has responsibility that runs the same length as the funnel.

The condition is that phrase, the right structure. A Head of Ecommerce who has a revenue target in place, whilst Marketing has the budget, IT has the roadmap and operations report elsewhere has the accountability and none of the levers. If that's your org chart, then putting the CRO onto their objectives makes very little difference until the levers move with it.

It also asks a lot of the person. They need deep practice in several of those disciplines and working fluency in the rest, enough to tell whether a dip is traffic mix, a hole in the stock file or a template release before anyone commissions a test to find out.