Your Conversion Rate Is Below Benchmark: What to Do

Finding that your conversion rate is below benchmark tells you to start looking, and almost nothing about where to look. Omniconvert has measured conversion across the CROBenchmark dataset of 7,000+ websites in 15+ industries, against 248+ audit criteria, over 13 years in eCommerce, and the stores that recover a real gap quickly almost all follow the same order: counting, then composition, then location, then research, then a controlled change. The stores that spend a quarter and find nothing almost all start at the end of that list. Last updated: October 2026.

For where the level ought to sit in the first place, see a good conversion rate. For why two stores reporting the same figure are rarely in the same position, see why two stores with the same CVR are not equal.

What a conversion rate below benchmark actually tells you

The first thing to establish is what the comparison is made of, because that determines how much weight it can carry.

A category benchmark is an average across stores that measure differently. One filters bots aggressively, another barely at all. One fires its conversion event at payment authorisation, another at order confirmation. Those choices move a store's reported rate by meaningful amounts, and the benchmark contains all of them mixed together.

It also blends composition. A category average includes established brands harvesting people who searched for them by name alongside newer stores buying cold demand, and the first group converts at multiples of the second. Being below the average may simply mean you have less brand demand than the brands in the sample.

None of that makes the benchmark useless. It makes it a single-use instrument: look once, decide whether the gap is large enough to investigate, then put it away and work against your own data. Teams that keep the benchmark on a dashboard end up arguing about the comparison for a quarter.

Step one and two: counting, then composition

Taking these in order is the whole discipline, so it is worth being specific about what each check involves.

Bot and crawler filtering. This moves in one direction: unfiltered traffic inflates sessions and depresses the rate. Compare your filtered and unfiltered session counts, and if the gap is large, your reported rate is understated against stores that filter properly.

Session definition. Timeout settings decide whether one person's two visits in an afternoon are one session or two. A shorter timeout inflates the denominator.

Consent handling. Where analytics records only visitors who accepted tracking, both halves of the ratio change, and stores in markets with different consent behaviour are not comparable on this metric at all.

Where the conversion event fires. At payment or at confirmation makes a difference, and so does whether failed payments are excluded. Check this in the implementation rather than in the documentation.

Then composition, which is the same work described in the companion article and worth repeating as a list: branded traffic share, device mix, returning-customer share, average order value and purchase frequency. Each one moves the blended rate without the store changing. The correction is to compare like with like, which in practice means non-brand mobile against non-brand mobile.

Shopify's analytics documentation is the practical reference for where these settings live on that platform, and it is worth reading before accepting any reported figure as comparable.

Step three: narrow the gap to one segment and one step

This is the step that converts a number into something a team can work on, and it is mostly patience.

Start with the two splits that explain the most variance: device and source. Build the funnel separately for mobile and desktop, and within each for brand and non-brand. Four cells. In most stores the gap concentrates visibly in one or two of them, which is the first genuinely useful output of the whole exercise.

Then split by step. Product view to add-to-cart, add-to-cart to checkout entry, checkout entry to order. A gap concentrated at one transition points somewhere specific. A gap spread evenly across all three is usually telling you something about traffic intent rather than about the pages.

Be sceptical of a diffuse result. If after the counting and composition checks the gap is still everywhere, the most likely remaining explanations are the mix of demand you buy and a measurement difference you have not found yet. Neither is fixed by a page change.

The table sets out the diagnostic order with what each step costs and what a negative result rules out.

The last row is in the table because it is the most common path taken. It is not a step, it is what happens when the first four are skipped.

Step four and five: research the cell, then change one thing

By this point the work is narrow, which is what makes it fast.

Scoped research is a different activity from general research. Watching twenty recordings of non-brand mobile sessions that reached checkout and abandoned is a morning, and it is usually conclusive in a way that watching two hundred unfiltered sessions never is. Exit surveys filtered to that same cell are similarly direct.

What you are looking for is a reason, stated by a visitor or visible in their behaviour, rather than a theory. Baymard Institute's checkout research is a reasonable prior for which problems are common at that step across stores, and it is a prior rather than a diagnosis: confirm the specific failure exists in your own recordings before building anything.

Then change one thing. Omniconvert Explore is the CRO platform for that, with A/B and multivariate testing, on-site surveys, heatmaps and session recordings in one place, and Explore averages a 23.2% conversion uplift across 70,000+ experiments. The reason to test rather than ship is specific to this situation: everything you know so far came from a benchmark comparison and some recordings, which is enough to justify a test and not enough to justify a permanent change.

Bain and Company's work with Reichheld on retention economics, which found that a 5% retention lift can raise profits by 25 to 95 percent, is worth holding alongside all of this, because a store that is genuinely converting as well as its traffic allows often has more to gain from the second purchase than from the first.

What a store owner should do this week

• Run the four counting checks. Bot filtering, session timeout, consent handling, conversion event timing. One hour, in the analytics implementation rather than the documentation.
• Build the four-cell split. Mobile and desktop, brand and non-brand. Then the three funnel transitions inside each cell.
• Write down where the gap sits. One sentence naming the segment and the step. If you cannot write it, go back to step one.
• Get a free audit of that step. CROBenchmark scores your store against the 248+ criteria behind the dataset, which tells you which known failures are present at the place you have localised.
• Validate any change against a control. Explore runs it as a test. Where order, session and margin data sit apart and have to be assembled before the gap can even be located, Nexus by Omniconvert unifies that data and ranks the next actions by True Profit, with a person approving what goes live.

FAQ: a conversion rate below benchmark

My conversion rate is below benchmark. Is my site the problem?

Possibly, and it is the third thing to check rather than the first. A published benchmark is a blended figure from stores with different measurement settings, different brand demand and different device mixes, so a gap against it carries at least three explanations before site quality. Rule out counting differences, then composition, then narrow the gap to a single segment and step. Teams that start by redesigning a page spend a quarter and often find the gap was in the analytics configuration.

What is the first thing to check?

How you count sessions and conversions. Bot and crawler filtering, session timeout settings, consent-banner effects on measurement, and whether the conversion event fires at payment or at order confirmation all move the figure materially. These checks take about an hour, they need no development work, and they resolve a meaningful share of benchmark gaps outright. Doing them last is the most common sequencing mistake in this situation.

How much of a gap is worth acting on?

Any gap that survives the measurement and composition checks and then localises to one segment and one step. A gap that remains evenly spread across every segment after those checks is usually still an artefact rather than a finding. Size matters less than location: a modest gap that sits entirely in mobile checkout is far more actionable than a large one spread evenly across the whole funnel.

Should I compare against the benchmark or my own history?

Your own history for decisions, the benchmark for orientation. Your history holds composition roughly constant, which makes a movement in it meaningful, while a published benchmark tells you only whether you are in an unusual position for your category. Use the benchmark once to decide whether to investigate, then stop looking at it and work against your own baseline, because that is the comparison where a change you make shows up honestly.

What if the gap is real and the site looks fine?

Then the problem is usually upstream of the site, in the mix of demand you are buying. A store acquiring broad, early-stage traffic converts lower than one harvesting brand demand, and no page change closes that gap because the visitors arrived with a different intent. The honest response is to look at revenue per session and at channel-level conversion rather than to keep testing a page that is working as well as its traffic allows.