Two stores show the same flat line on the revenue chart. One is losing people it already has, at a step it could fix this month. The other is holding on to everyone it gets and has simply run out of demand it can afford to buy. Those are different problems. The first is a leak. The second is a plateau.
The distinction matters because the instinctive fix for a flat line is more traffic, and that helps only one of them. Pour traffic into a leak and the loss scales with the spend. Cut spend on a plateau and revenue falls immediately, because there was never anything wrong with the store.
So before deciding what to do about a flat month, it is worth spending an afternoon deciding which of the two you are looking at. The evidence is usually already in your account, and the test is simpler than it sounds.
A leak shows as a gap between two steps
Leaks are local. They live between two specific points in the buying path, and they show up as a share that does not survive the move from one to the next.
Product page to add to cart. Cart to checkout. Checkout started to order placed. First order to second order. Feed submitted to products actually showing. Each of those is a step with two ends, and a leak is a number at the second end that is smaller than it should be given the first.
The useful property of a leak is that it is asymmetric. It affects one group and not another. Mobile and not desktop. Paid visitors and not email visitors. One collection and not the rest. New customers and not returning ones. That asymmetry is the fingerprint, and it is what makes leaks findable, because you can split the number until the split stops explaining anything.
Leaks also tend to have a date. Something changed: an app was installed, a theme was updated, shipping rates were revised, a payment method stopped appearing on a browser. If you can find a week where the ratio changed and nothing about the traffic did, you are almost certainly looking at a leak rather than a market.
A leak costs you the visitors you have. A plateau costs you the ones you can afford.
A plateau shows as a rising price for the same result
Plateaus are not local. Nothing is broken at any single step, and splitting the data does not turn up a group behaving unusually.
What you see instead is a price. The cost of another order creeps up while conversion rate, average order and repeat rate all sit still. You add budget and the extra spend brings back less than the previous increment did. You add a keyword set or an audience and it performs worse than the one before, which is expected, because the cheapest demand was bought first and what is left costs more.
A plateau is the store meeting the limit of demand it can profitably reach with the current offer, catalogue and economics. It is not a failure. Most healthy stores hit one, and hitting it is evidence that the easy work has been done.
The tell is that the ratios are stable and the costs are not. If conversion is flat, average order is flat, repeat rate is flat, and the only number moving is what you pay to acquire, no amount of page optimisation will change the picture. You are being asked a different question: what would make an expensive visitor affordable.
The test: does adding traffic change the shape
Here is the practical way to tell them apart without a long study.
Take a modest, temporary increase in traffic to a part of the store you can isolate, and watch what happens to the ratios rather than the totals. If revenue rises roughly in line with the visitors and every ratio holds, the store is not leaking. It is converting the new people at the same rate it converts everyone else, and your constraint is the price and supply of visitors. That is a plateau.
If revenue rises by less than the traffic did, and the ratio at some step gets worse as volume grows, something in the path cannot take the load. That is a leak, and often a leak that only appears at scale: a support queue that cannot answer questions fast enough, stock that runs out, a page that slows under load, a payment step that fails more often on the traffic mix you just added.
Run it small and run it short. This is a diagnostic, not a campaign. The question is only whether the shape holds, and the shape answers within a couple of weeks in most stores with reasonable volume.
One of each, worked through
Invented numbers, used only to show the reasoning.
Store A gets 30,000 visitors a month. Desktop converts at 2.4 per cent, mobile at 0.9 per cent, and three quarters of the traffic is mobile. Cost per order has been stable for a year. Nothing about the auction has changed. The gap between devices is the whole story, and it is worth a lot: bringing mobile even part of the way to desktop on that traffic is worth more orders than any budget increase the store could afford. That is a leak, and the work is on phones.
Store B gets 30,000 visitors a month, converts at 2.1 per cent on both devices, and has raised spend three times in six months. Each raise brought fewer orders per dollar than the last. Conversion, average order and repeat rate have not moved through any of it. Nothing is broken. The store has bought most of the demand it can afford at its current order value, and the route forward is to make an order worth more, or to make a customer worth more than one order, so that a pricier visitor becomes affordable.
Same flat chart. Opposite work.
Why the two get mistaken for each other
Mostly because the reporting that reaches the owner is blended, and blending destroys exactly the asymmetry that identifies a leak.
A single conversion rate for the whole store averages a device gap out of existence. A single cost per order across the account hides that one product group is comfortably profitable and another has never paid for itself. A month over month revenue comparison mixes a change in traffic mix with a change in the store's ability to sell, and gives you one number that could mean either.
The other reason is that a plateau is uncomfortable to name. A leak has a villain and a fix with a date. A plateau says the next step is harder: raise order value, build repeat purchases, or find demand somewhere that is not the auction you are already in. It is easier to keep hunting for a broken button.
We would rather say it early. Telling a store that its pages are fine and the constraint is its economics is less satisfying than finding a bug, but it puts the quarter's work in the right place.
The first week of each looks different
If it is a leak, the first week is narrow and fast. Isolate the group that behaves worse, reproduce the loss yourself on the same device or the same path, fix the specific thing, write down the before number with a date, and measure the same ratio a fortnight later. One change at a time, so the result means something. Leaks are the best kind of work: cheap, quick to confirm, and they keep paying on every visitor afterwards.
If it is a plateau, the first week is arithmetic rather than repair. What does an order leave behind after product cost, shipping and fees. What would a second order be worth. Which products could carry a higher cost per order and which never could. What would change if the average order rose by a few dollars through a bundle or a threshold set from your data, or if a share of buyers came back once more a year.
The plateau route is slower and it compounds, because raising order value and repeat rate raises the price you can afford to pay for every visitor after that. That is what turns a ceiling into a higher ceiling, and it is why we would rather spend the first week on the arithmetic than on another round of budget increases into the same auction.