Upsells at checkout and after it, without wrecking the order

Average order value is the least glamorous growth lever and frequently the cheapest. The traffic is already there, the customer has already decided, the payment is already happening. A few per cent more per order goes almost entirely to the bottom line.

It also has the worst risk profile of any optimisation, because the intervention sits directly on top of a customer who was about to give you money.

Two moments, two different rules

At checkout, an offer competes with completing the order. Anything that adds a decision costs conversions, and a lost order is worth far more than an accessory. So the offer has to be genuinely relevant and dismissible without thought — batteries for the toy, a case for the phone. The kind of thing where “oh, yes” takes no consideration at all.

After purchase, the risk profile inverts. The order is placed and paid; nothing you show now can lose it. This is the safest place to make an offer and the most under-used, and it is where a more considered upsell belongs.

Relevance is the whole thing

An offer that does not follow from what is in the basket is noise, and noise at checkout is expensive. Offers are attached to products rather than shown globally, because the point is the companion nobody thought of, not another product.

The test is simple: would a good shop assistant mention this? They would mention batteries. They would not mention an unrelated item that happens to be on promotion, and neither should the checkout.

Record the no

Responses are recorded, and this matters more than it sounds. Showing the same customer the same declined offer repeatedly is how a shop starts to feel like an advert — and the second showing converts far worse than the first while annoying more people.

Knowing what was declined also tells you which offers are wrong. An offer nobody accepts is not a pricing problem, it is a relevance problem, and it is costing you conversions on every order it appears on.

Measure the whole order, not the upsell

Every upsell tool reports what it sold, which is always a positive number and always makes the tool look good. It is the wrong measurement.

The number that matters is revenue per session across everyone who saw it, including the people who abandoned. A checkout upsell that adds £4 to one order in twenty and loses one order in fifty is losing money, and its own report will show it winning.

If you cannot measure it that way, be conservative — and if you have to choose, put the offer after the purchase, where it cannot cost you anything.