How to Know if Your Conversion Rate Is Actually Bad
A bad conversion rate is one where a slice of your own traffic converts far below the rest of it. The sitewide figure cannot tell you that, because an average hides its worst component. Split your rate by device, by source and by visitor type, and either a gap appears, which is your problem, or none does, which means the number is describing your offer.
- A rate is bad when part of your traffic converts far below the rest, not when it sits below a published average.
- A sitewide figure hides its worst component, so it cannot tell you whether you have a problem.
- The Three Cuts locate it: device, then source, then visitor type, applied in that order.
- A rate falling on a stable traffic mix is a genuine signal, and the one most often missed.
- If no cut shows a gap, the number is describing your offer rather than a storefront fault.
A bad conversion rate is a diagnosis, and almost nobody arrives at it by diagnosis. They arrive at it by comparison: a figure read in an article, a number a founder mentioned, a category average from a report that did not describe their business. The comparison feels like evidence and it cannot be, because it has no access to the two things that set your rate, which are what you sell and who arrives. Last updated: August 2026.
Omniconvert has measured conversion behaviour across the CROBenchmark dataset of 7,000+ websites in 15+ industries, against 248+ audit criteria, over 13 years in eCommerce. The most consistent finding about weak conversion rates is that they are rarely weak evenly. A storefront that looks uniformly mediocre in aggregate usually turns out to be performing acceptably on three quarters of its traffic and badly on the remaining quarter, and the quarter is findable in about twenty minutes.
This piece is the procedure for finding it. Three cuts through data you already have, in a deliberate order, plus how to tell a genuine weakness from an artefact of your traffic mix. If you want the prior question of what the number should be judged against at all, our pillar on a good conversion rate makes that case and this piece assumes it.
What actually counts as a bad conversion rate
Two definitions are in play and only one of them is useful.
The comparative definition says your rate is bad if it is lower than other people's. This is the popular one and it collapses on inspection. A store selling a considered eight hundred pound purchase to cold paid traffic will convert at a small fraction of a store selling a repeat consumable to an email list, and neither store is better run. The comparison is between businesses, not between storefronts.
The diagnostic definition says your rate is bad if some identifiable part of it is much worse than the rest. This one is useful because it is actionable. A gap points at something. An average points at nothing, which is why teams who work from averages tend to respond by redesigning the homepage, an intervention chosen because it was available rather than because anything suggested it.
There is a third case worth naming, which is a rate falling over time on a stable traffic mix. That is a genuine signal and it is the one most often missed, because a slow decline stays inside the range that looks normal week to week. The cuts below catch it, since a decline concentrated in one slice is far easier to see than the same decline smeared across all of them.
The Three Cuts
Each cut takes a few minutes in any analytics tool. Run them in order and stop when one produces a gap large enough to explain the number you were worried about.
Cut one, by device. Mobile, tablet, desktop, over at least four weeks. Mobile converting somewhat below desktop is universal and expected, because phones carry more browsing, more interrupted sessions and more people who will finish the purchase later on a laptop. A very large gap is different in kind. It usually means something on mobile is not merely harder but broken: a payment option that fails, a field that cannot be filled, a delivery selector that needs a hover to operate. Baymard Institute's usability research has catalogued these failures for years, and the striking thing about them is how invisible they are to the team, because the team tests on the two handsets it owns.
Cut two, by traffic source. Branded search and direct traffic on one side, paid prospecting and cold social on the other. These convert at wildly different rates by nature, so a change in the mix moves your sitewide average without anything on the storefront changing at all. A month of heavy prospecting mechanically lowers your rate while revenue rises. Teams who miss this cut regularly conclude that their site got worse during their best growth month, and some of them act on it.
Cut three, by visitor type. New against returning. This one is diagnostic rather than exculpatory. A healthy returning rate carrying a weak new rate says the storefront works and first-time visitors do not trust it yet, which is a content and reassurance problem. A healthy new rate with a weak returning rate is stranger and usually means people are coming back to check something, most often delivery or returns terms, and leaving to find it. Both are specific. Neither is visible in the sitewide number.
Reading the spread instead of the average
The output of the three cuts is not a better number. It is a distribution, and distributions are read differently.
A narrow spread, where every slice sits close to every other slice, is genuinely good news dressed as bad news. It means your storefront is treating all your traffic about equally, so there is no broken segment to find. If the overall level still disappoints you, the constraint is upstream of the storefront: the offer, the price, the delivery promise or the quality of the traffic you are buying.
A wide spread is the opposite. One slice far below the others is a fault, and faults are cheap to fix relative to offers. This is the outcome you are hoping for when you run the cuts, because it converts a diffuse worry into a specific page on a specific device.
The table below is what the three cuts typically surface, and what each pattern means.
| What the cuts show | Most likely cause | Where to look first |
|---|---|---|
| Mobile far below desktop | A broken step, not a harder one | Mobile checkout and payment |
| Rate fell as paid spend rose | Mix change, storefront unchanged | Nowhere: check revenue instead |
| New far below returning | Trust and reassurance gap | Product page facts and proof |
| Returning far below new | People return to check a fact | Delivery and returns visibility |
| One country far below the rest | Payment, currency or delivery gap | Local checkout options |
| Narrow spread, low level | Offer or traffic quality | Price, delivery promise, channel |
The final row is the one teams least want and most often have. A narrow spread at a low level is not a storefront problem, and continuing to optimise the storefront in response to it is the most expensive mistake in this article, because the work is real, the effort is genuine and the ceiling was somewhere else the whole time.
The false alarms, and how to dismiss them
Before acting on any decline, rule these out in order. Each takes minutes and each explains a large share of the drops teams investigate.
The mix changed. Covered in cut two and worth repeating because it is the most common cause by a wide margin. Compare the two periods by channel before comparing them by anything else.
The window is seasonal. Comparing a promotional fortnight against a normal one measures the promotion. Statista's eCommerce reporting has tracked how sharply seasonal buying patterns swing in several categories, and a comparison that straddles one of those swings is measuring the calendar rather than the store.
Automated traffic arrived. Bot and scraper waves add sessions and no orders, which drops the rate immediately and recovers on its own. A sudden change with a flat order count and a jump in sessions from one source is nearly always this.
The definition moved. A tracking change, a new consent banner, a tag deployed slightly differently: any of these can change what counts as a converting session. This is the sneakiest of the four because the number is genuinely different and nothing about the business changed. Check whether your order count agrees with your platform's before believing your rate.
What to do once you have located the gap
- Reproduce the slice. Same device class, same entry point, same country if the cut was geographic. Not a simulator on a desktop browser, an actual phone if the gap is mobile.
- Complete a purchase. All the way through payment, with a real card if you can. Faults concentrate in the last two steps, which is exactly where internal testing usually stops.
- Write down every hesitation. Not just failures. The places you had to think are the places a customer leaves, and you know the store already.
- Fix before testing. If the fault is a breakage, repair it and move on. Running a test to confirm that a broken thing underperforms a working thing spends a month proving something you already know.
- Re-cut in four weeks. The same three cuts. This tells you whether the gap closed, which is a far better measure of your work than the sitewide rate will ever be.
If the cuts sent you to a page rather than to a breakage, the next step is a structured review of that page rather than an improvised one. Our sister guide on how to audit it sets out the criteria to work through, and the underlying method for reading a rate against its own history is covered in this conversion rate analysis guide.
What the cuts cannot tell you
Two limits are worth holding on to.
The first is that a gap is not automatically worth closing. If the underperforming slice is a channel with poor margins or a country where delivery costs eat the order, fixing its conversion rate raises volume on unprofitable business. The cuts see sessions and orders. They do not see what an order is worth, and the difference between those two views is where a great deal of conversion effort quietly goes to waste. Teams who want priority set by profit rather than by volume can see how Nexus by Omniconvert unifies commerce data and prioritises experiments by True Profit, with a human approving what goes live.
The second is that no amount of cutting improves a weak offer. If your price, delivery and returns compare badly against the alternatives a shopper has open in another tab, every slice will be soft and the spread will be narrow, and the diagnosis is correct and unwelcome. Marketing Metrics puts the probability of selling to an existing customer at roughly sixty to seventy percent against five to twenty percent for a new prospect, which is a reminder that a store fighting a hard acquisition rate often has a much easier win sitting in the half of its economics it is not measuring.
Frequently asked questions
How do I know if my conversion rate is bad?
Not by comparing it to an industry average, which cannot see your traffic mix or your price point. Split your own rate by device, by traffic source and by new against returning visitors. If one slice converts far below the others, that gap is your problem and it is specific enough to fix. If every slice sits close together, your rate is a fair description of your offer rather than a fault in your storefront.
What is a normal gap between mobile and desktop conversion?
Mobile converting somewhat below desktop is ordinary across most of eCommerce, because mobile carries more browsing and more interrupted sessions. A very large gap is not ordinary and usually means something is broken rather than merely harder: a payment method failing, a form that cannot be completed on a small screen, or a delivery step that requires a hover. The size of the gap is the signal.
My conversion rate dropped but nothing changed on the site. Why?
Because the denominator changed. A new campaign, a seasonal shift, a change in device mix or a wave of automated traffic all add sessions without adding buyers. Cut the period before and the period after by source and by device, and in most cases the drop resolves into one channel that grew rather than a storefront that got worse.
Should I worry about a low conversion rate on a high-consideration product?
Much less than you would for an impulse purchase. A product people research for weeks produces many non-converting sessions from the same person, which lowers the rate without meaning anything is wrong. Read assisted conversions and returning-visitor behaviour instead, and judge the storefront on whether a decided buyer can complete a purchase easily.
What should I fix first if the cuts show a real problem?
Whatever is closest to the purchase inside the weakest slice. If mobile checkout is the gap, the fix is in mobile checkout, not on the homepage. Locating the slice is most of the work, because it converts a vague concern about a sitewide number into a specific page on a specific device that you can open and test yourself.
Once you know the rate is genuinely below where it should be, the next step is to set a realistic conversion rate target.
The bottom line
Stop asking whether your conversion rate is bad in the abstract, because the question has no answer in that form. Ask instead whether it is bad somewhere, which does have an answer and takes about twenty minutes to find. Cut by device, then by source, then by visitor type, and read the spread rather than the average. A wide spread hands you a specific fault on a specific page, which is the best possible outcome because faults are cheap to fix. A narrow spread tells you something harder and more valuable: the storefront is doing its job and the constraint is your offer or your traffic. Either way you leave with a next action, which is more than any industry average has ever given anybody.