First: is this even outside normal variance?

Pull the last 52 weeks of AOV. Is a 12% week-over-week move unusual? If standard deviation of weekly AOV is 8%, a 12% move is ~1.5σ — not rare. If SD is 2%, 12% is a fire.

If it is real, the decomposition:

AOV = total_revenue / order_count. Something moved in one or both.

Revenue flat, orders up → more smaller orders. Could be a new segment of low-value users, a coupon / discount, or a promo driving small-basket buying.
Revenue down, orders flat → same volume, cheaper mix. Maybe a high-ticket item went out of stock, or a campaign promoted lower-tier items.
Revenue down, orders down → total business contraction. AOV movement is a side effect, not the issue.
Both up, but revenue up less → new volume coming from a lower-AOV segment.

Segmentation to isolate the driver:

By product category. Is the drop in one category dragging the mean? A single category where a flagship item sold out will move AOV.
By user type. New vs. returning. If a marketing campaign brought in a flood of new users who tend to buy smaller baskets, AOV drops while order count rises — that's success, not failure.
By channel. Mobile vs. desktop (mobile often has lower AOV); organic vs. paid (paid often has lower AOV); domestic vs. international.
By promotion / coupon. Did a coupon launch this week? Coupons drop AOV by design.
By payment method. Sometimes new payment options attract smaller-basket buyers.

A likely finding:

"AOV dropped 12% week-over-week. Drill: order count is up 25%, revenue up 10%. The AOV dip is explained by a promotional email campaign that drove a spike of new first-time buyers, who average $35 per order vs. the site-wide $42. Total revenue is up — this is a healthy acquisition event, not a degradation. Expect AOV to recover as cohort mix normalizes."

What to report to the PM:

The decomposition (revenue vs. orders).
The segment driving the drop.
Whether it's a problem or a natural mix-shift consequence of something good.
The specific next thing you'd measure to confirm the hypothesis.

Red flags to check before concluding:

Data pipeline change (did 'revenue' suddenly start excluding taxes / shipping / tips)?
Definition drift (is 'orders' including refunded orders that used to be excluded)?

What senior candidates add: AOV by itself is often not the decision-useful metric. 'Revenue per acquired user in the next 90 days' captures the tradeoff between 'acquired cheaply' and 'has lower AOV forever.' Focus analysis on what actually predicts business outcome.

AOV dropped 12% this week. Walk through your investigation.

Draft your answer

Stuck? Peek at a hint

More product metrics questions