First, separate mix from rate and pin down the definition
Begin with what "value" means in the report. The GA4 e-commerce tracking tutorial follows Google's developer guide to measuring ecommerce in sending the purchase event's value as the item total, with tax, shipping and any coupon carried in parameters of their own, and the GA4 AOV entry explains why that figure seldom matches the shop's own average. A release that moved delivery into or out of the value, or started sending discounted item prices, changes the AOVAverage revenue per order. overnight. Compare the shop platform's average with the analytics average for the same days; if only one moved, the problem is in the records.
Next, split the blended average. The guide to explaining metric changes separates a mix effect, where segments with different baskets change their share of orders, from a rate effect, where baskets inside a segment shrink. Use segments defined the same way in both periods: new against returning customers, channel, and the product or collection that started the order. If returning customers spend what they always did and the fall belongs to a growing group of first orders from one source, the store has gained customers, not lost spend. If baskets shrank inside steady segments, read on.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Single-item orders rise sharply while a promotion runs | Items per order for promoted orders against others; the promotion's minimum basket | A promotion that attracts single-item buyers | Rebuild the offer around a basket the store wants, or end it |
| Orders bunch below a value they used to clear | Distribution of basket values; whether the threshold still shows on product pages and in the basket | A free-delivery threshold removed, moved out of reach or no longer shown | Restore the threshold's visibility or reset it from the distribution |
| The average fell when a campaign launched, led by one cheap product | Item list and promotion data for the campaign slots; top items in campaign orders | Campaign merchandising pushing a cheaper line or entry product | Rebalance the collection so profit and pairing lines sit beside the entry product |
| A trial size, sample pack or starter now opens many orders | Share of orders containing the starter; what those customers order next | Samples, trial sizes or a low-priced starter now a large share of orders | Report starter orders apart and judge them on the full-size orders that follow |
| Returning customers spend as before; first orders or one channel grew | Average by customer status and channel in both periods | A shift in mix towards first orders or a new channel, read as a change in behaviour | Report the average by segment and stop treating the blend as behaviour |
| Shop and analytics averages diverge, or the change lines up with a release | Value, tax, shipping and coupon fields on a recent purchase event; release notes | Order value now recorded with or without tax, delivery or discounts differently | Restore one documented value convention and annotate the break |
Fix, verify and prevent each cause
An offer that rewards buying one thing
A code or flash price that works on any single product invites shoppers to buy exactly that product and nothing else. The e-commerce conversion hub lists a promotion that attracts single-item buyers as a usual cause of order value falling while conversion rises, and the average order value levers guide insists on reading the two together.
Fix: reshape the next promotion so the saving applies to the basket the store wants, such as a set or a second item, or let the current one end on its stated date rather than extending it.
Verify: items per order for promoted orders move back towards the store's norm, and total contribution from the promotion holds or rises.
Prevent: before any offer goes live, write down the basket it is meant to produce and the measure that will show whether it did.
A delivery threshold that stopped working
A threshold persuades only while shoppers can see it and reach it. A redesign that drops the progress bar from the basket, a price rise that pushes typical baskets further below the line, or a threshold quietly removed all leave shoppers with no reason to add the extra item. The levers guide sets thresholds from the distribution of order values and presents them on the product page and in the basket.
Fix: confirm the threshold appears wherever the levers guide puts it, and if it still shows, compare its level with today's distribution rather than last year's.
Verify: the cluster of baskets just below the line thins and the share clearing it returns.
Prevent: add the threshold message to the release checklist for the basket and product templates, so a redesign cannot drop it unseen.
A campaign that led with the cheapest line
Campaign sets usually include entry products meant to make a first order easy. Trouble starts when the entry product takes the hero slot and nothing beside it suggests a second item. The guide to merchandising the store for a campaign gives each line in the set a role and tags every slot with promotion and item list events, which shows which products the campaign actually sold.
Fix: keep the entry product but place profit lines and natural partners beside it in the landing collection and on its product page.
Verify: campaign orders containing more than the entry product grow, read from the slot data.
Prevent: agree the campaign set's roles with whoever plans the offer, so one cheap line is a deliberate choice rather than an accident of sorting.
Starters now opening a large share of orders
A paid trial set or sample pack exists to lower the first step, so it lowers the first basket by design. The guide to samples and trial sizes judges such programmes on cost per full-size customer and on what sample buyers go on to order, read as a cohort, not on the sample order's value.
Fix: report starter orders as their own line and remove them from the headline average used to judge everything else.
Verify: the average for full-size orders, read alone, shows whether baskets changed at all.
Prevent: price the trial size so the full size remains the better buy, as the samples guide advises, which stops regular customers drifting into the small size.
A different set of customers, not smaller baskets
When the decomposition shows returning customers unchanged and a growing share of first orders or of one channel, the average fell because the mix moved. The explaining metric changes guide warns that fixing a mix effect as if it were a rate effect leads to the wrong cure, such as cutting the campaign that was adding orders.
Fix: publish the average by customer status and channel beside the blended figure, and set targets per segment.
Verify: the segment figures stay stable while the blend moves, and the team stops debating the blend.
Prevent: keep the segment split permanent in the store's reporting, with total revenue and conversionShare of visits or clicks that convert. beside it, so a future change in mix is read correctly the week it happens.
A definition that changed under the report
If the shop's average held while analytics fell, or the drop starts on the day of a release, check the purchase event. The tracking tutorial asks teams to decide once, before the first release, whether event values include tax and shipping, and to keep the same product identifier and price format on every template.
Fix: restore the documented convention, then mark the break on every chart that spans it.
Verify: a closed day's revenue and order count reconcile with the shop's order list.
Prevent: record every change to value, tax, shipping or coupon handling in the measurement change log before it ships.
A prevention routine for order value
Order value rarely has an owner, which is why it drifts. Give it one, and read it monthly as a distribution and by segment rather than as a single mean. Each review asks four questions in order. Did anything change in how value is recorded? Did the share of first orders, channels or starter products move? Did any promotion or threshold change in the period? Which products opened the orders that shrank? Answering them in that order stops a team rebuilding its offer when the real change was a tag.
Before any promotion, write down the basket it should produce and the guardrail that will stop it, and check the delivery threshold still shows in the basket after any template release. After each campaign, read the slot data to see which products did the selling. Keep the discount habit apart from this routine: when full-price buying dries up between sales, that is a pattern of its own, diagnosed separately in the e-commerce troubleshooting section. Returns are also separate: a basket that holds its value but comes back is covered by the page on a rising return rate.
Hinkleplum Pantry finds two reasons for smaller baskets
Hinkleplum Pantry, an invented online deli, sees orders rise from 1,840 in March to 2,310 in May while average order value falls from £46 to £38. The analyst first checks the purchase event: value has been the item total since launch, and the shop platform shows the same fall, so the records are consistent. Split by customer status, returning customers' average holds at £49 to £50. First orders grew from 31% to 52% of orders, and 61% of them contain the £15 tasting box promoted on social media in April. That accounts for most of the fall. The rest sits in returning orders between £30 and £39, which rose from 14% to 22% of returning orders: a basket redesign in April removed the line showing how far the shopper is from the £40 free-delivery threshold. The team restores the line, reports tasting-box orders separately, and follows the April tasting-box cohort, 23% of whom place a full-size order within 60 days. By July, returning customers' share of baskets between £30 and £39 is back to 15%, and contribution per visitor is up £0.40 on March.
Common mistakes
Reading the blended average as customer behaviour. Raising the delivery threshold to fix a fall caused by a starter product. Comparing an analytics average with a shop average built on different terms. Ending a campaign that added customers because it lowered the mean. Judging any fix on order value rather than contribution per visitor.
