Your traffic is fine. Here is where the money leaks.

The four gates

Site-wide conversion rate is one number describing four completely different problems. Separating them is the single most useful hour you can spend on your store, because it turns "sales are down" into a specific fault with a specific fix.

The point of this exercise is not to get all four numbers up. It is to find which one is anomalous, because effort spent on the other three is close to wasted.

Measuring them without rebuilding your analytics

You need five raw counts, over the same period. Use ninety days if you can, so the numbers are not dominated by one good week.

1. Sessions. From your analytics.
2. Sessions that viewed at least one product. From your analytics, as a page or event count.
3. Add to cart events. From your analytics, or WooCommerce if you track it.
4. Checkouts started. Sessions that reached the checkout page. Analytics will give you this as a pageview count on your checkout URL.
5. Orders. From WooCommerce, not analytics. This is the one number you should take from the shop itself, because analytics undercounts orders for reasons covered below.

Then divide each step by the one before it. That is the whole method. Five numbers, four divisions.

How to read the pattern

Now compare your four percentages against each other rather than against any published benchmark. You are looking for the one that is dramatically worse than its neighbours.

Leaking at gate 1: they arrive and never see a product

This is a traffic or navigation problem and nothing else. Either you are buying the wrong visitors, or people cannot find what they came for. Check your top landing pages and top search terms. If your paid traffic leaks here and your organic traffic does not, that is a media buying problem, and no amount of work on your store will fix it.

Leaking at gate 2: they look and do not add

The product page is not making the case. In rough order of how often it is the cause: delivery cost or timing is unclear, the photography is inadequate, the price is wrong for the market, there are no reviews, stock or variant availability is confusing, or the description does not answer the obvious question.

The cheapest test here is to state your delivery cost and timing on the product page and watch this gate for a fortnight. A surprising share of gate 2 leaks are actually gate 3 leaks that the shopper anticipated.

Leaking at gate 3: they add and never start checkout

This is almost always cost. The shopper reached a total and did not like it. The Baymard Institute's abandonment research puts extra costs at 40% of abandonments, well ahead of anything else, and a gate 3 leak is that number showing up in your own data.

Also worth checking: a minimum order value, a shipping method that does not cover their country, or a coupon field that made them leave to go hunting for a code. Read should you offer free shipping, the maths not the opinion before you react to this one, because the obvious fix is not always the profitable one.

Leaking at gate 4: they start checkout and do not finish

This is the gate you control most directly, and it is the one where defects hide. The shopper has seen the total, accepted it, and started typing. Something then stopped them.

Candidates, in the order worth checking: the form is broken or unusable on their phone, a required field is failing validation in a way they cannot see, the payment method they wanted is not there, their card was declined with an unhelpful message, or the checkout is simply long enough to lose people. Baymard's benchmarking found the average checkout carries 11.3 form fields when most stores need only 8.

Three traps that will hand you a wrong answer

Each of these produces a confident number that is not true, and all three are common.

1. Analytics undercounts orders

Ad blockers, consent banners, tracking that fires on a page some buyers never reach, and gateways that redirect away and back all cost you recorded conversions. It is normal for analytics to show meaningfully fewer orders than WooCommerce. Take your order count from WooCommerce and everything upstream from analytics, and accept that the final ratio is approximate.

2. The thank you page double counts

The opposite failure, and it inflates rather than deflates. WooCommerce's order received page can be loaded more than once by the same buyer, refreshing or returning from an email. If your conversion tag fires on page load rather than once per order, your recorded conversions are too high and your reported cost per acquisition is too low. The fix is in how to customise the WooCommerce thank you page.

3. Bots and your own traffic

Unfiltered crawler traffic lands almost entirely at gate 1, which makes gate 1 look catastrophic and every downstream gate look better than it is. Your own visits and your developer's do the same thing at smaller scale. Filter internal IPs and check whether your analytics is excluding known bots before you conclude anything about gate 1.

Watch the same four numbers, forever

The one-off diagnosis is worth doing. The recurring version is worth much more, because it turns a silent decline into an alarm.

When revenue falls, the untrained response is to look at traffic. The useful response is to look at which gate moved. A store whose gate 4 quietly dropped four points after a plugin update has a fixable, specific fault that a traffic report will never surface. A store whose gate 1 fell because a campaign changed has a media problem that no amount of checkout work will touch.

• Record the four numbers monthly, split by device. A spreadsheet is enough.
• Note every significant change you make next to the month it happened in. Plugin updates and theme changes count.
• Investigate a gate that moves several points, and ignore a gate that wobbles by a fraction. Which is which depends on your volume, and the arithmetic for that is in how many orders before your numbers mean anything.

Fixing the worst gate

Two rules, and they save a lot of wasted work.

Fix defects before you optimise anything. A broken thing is not a low conversion rate. If your gate 4 is bad on mobile because the pay button is unreachable when the keyboard is open, that is not a design question and no amount of A/B testing will find it. Walk your own checkout to the payment step on a real phone with a realistic cart before you theorise.

Then work only on the anomalous gate. The temptation after doing this exercise is to improve all four. Resist it. Three of your gates are probably normal for your business, and effort there produces fractions of a percent while the anomalous one is losing you real orders.