70.22% of carts are abandoned, and 40% of the shoppers who leave blame extra costs. Those two numbers come from Baymard's review of fifty studies, and as of November 10, 2026 they still describe the most common failure in e-commerce: a product page that promises one thing and a checkout that charges another. Slow delivery is cited by 20 per cent of leavers, account creation by 18 per cent, site errors by 17 per cent and the returns policy by 13 per cent. Every one of those is a promise made or broken on the page.
Complaints follow the promise, not the product
Treated as a consumer trend rather than a service metric, the pattern is consistent: people do not complain that a product is bad, they complain that it was not what the page said. The pouch says a size, the listing says a pack count, the checkout adds a fee — and the review that follows reads like a product complaint but is really a copy failure.
This is where sentiment work earns its keep. Reading complaints as product feedback leads to reformulation nobody asked for; reading them as listing feedback leads to edits that can ship the same week. In a year when discretionary turnover grew just 1.7 per cent, the cheaper fix is the one worth trying first. Treat it as review analysis with a narrow output: every consumer review theme must map to a page element a team can edit, and every edit must be re-checked against the next month of sentiment analysis on the same archive.
Four themes that repeat
What the pack actually contains
Pack count, net weight and serving size generate the largest volume of disappointed reviews in grocery and beauty, because the page disagrees with the parcel. A listing that shows the unit price alongside the pack price removes most of the ambiguity, and it is the single edit that most reliably reduces one-star reviews about value.
What it costs at checkout
Extra costs are the top stated reason for abandonment and a frequent review theme. Shipping thresholds, handling fees and minimum order values belong on the product page, not three screens later. When the number appears only at payment, the shopper has already formed a price expectation that the checkout breaks. The returns line follows the same logic: a shopper who discovers the policy after delivery writes a review about being misled, while a shopper who reads it before ordering writes a review about the product. Running review analysis on those two groups separately is what separates a copy problem from a product problem.
| Review theme | Page element at fault | Edit that fixes it |
|---|---|---|
| Smaller than expected | Pack count and net weight | Unit price and scale reference |
| Hidden fees | Shipping threshold | Delivered price on the page |
| Hard to return | Returns policy link | Return window in the buy box |
Turning review text into listing edits
The mechanism is unglamorous. Take the last two hundred reviews below three stars, tag each against a page element rather than a product attribute, and count the tags. The top three tags are the next three edits, and the count before and after is the proof that the edit worked.
Tagging against elements rather than attributes is the step that most teams skip, and it is the reason review programmes produce slide decks instead of changes. A tag such as 'size' cannot be assigned to anyone; a tag such as 'net weight not shown' can be fixed by one person in an afternoon.
Measure twice before the peak
Mobile is where this lands hardest. More than half of peak online sales come from mobile, and mobile web converts at roughly half the desktop rate. A page that buries the pack size below three scrolls is a page that will collect exactly the reviews described above, because the shopper never saw the number that would have set expectations.
Before the busiest weeks, check three things on a phone: whether the pack size is visible before the first scroll, whether the delivered price appears before checkout, and whether the return window is stated where the buy button is. Each is a small edit and each maps directly to a review theme that already exists in the archive.
FAQ
Why treat reviews as listing feedback rather than product feedback?
Because most low-star complaints describe a gap between the page and the parcel. Fixing the page is faster and cheaper than reformulating, and it addresses the complaint the shopper actually made.
What is a realistic sample for tagging?
Two hundred recent reviews below three stars is enough to rank themes. The value comes from re-counting the same tags later, not from a larger first sample.
Does showing the delivered price on the page reduce conversion?
It reduces add-to-cart for shoppers who would have abandoned at payment anyway. Net orders usually hold, and complaints about hidden fees fall.
How does the ACCC sweep change this?
It raises the cost of unsubstantiated claims. Every claim on a page now needs a source the team can produce, which is the same discipline review tagging creates.
