Online retail admin: where the stock, returns and reviews hours actually go

The order goes out. The confirmation email fires. From the customer's side, that's the whole job done. From yours, it's the thirty-second bit at the end of a much longer chain - the reconciling, chasing, replying and re-listing behind it is where the actual week goes, and in the last year that week has picked up a new line item: reviews and pricing compliance under UK consumer law.

The Digital Markets, Competition and Consumers Act 2024 gave the Competition and Markets Authority new powers over fake and incentivised reviews, and a parallel ban on "drip pricing" - hiding mandatory fees until checkout instead of showing the full price upfront - both came into force together on 6 April 2025. Paying for reviews, suppressing genuine negative ones, or publishing reviews you haven't taken reasonable steps to verify are now enforcement matters, not just a Trustpilot policy breach, and the same is true of a checkout that doesn't show the full price from the start. None of this is optional admin you can leave for a quiet week. It sits on top of the stock counts, supplier chasing and returns processing that were already eating the day, and it lands hardest on shops selling across more than one channel, where the same product needs a compliant listing on Shopify, on Amazon and on eBay, each with its own review policy to satisfy.

Where the hours actually go

Ask most online retailers where their time goes and you'll get "orders and stock." Sit down and map an actual week and it splits out into four distinct jobs, each individually small enough to shrug off and collectively large enough to explain why growth work never quite gets started.

Stock reconciliation. Checking what the system says you hold against what's actually on the shelf or in the unit, then chasing down every discrepancy - a mis-scanned barcode, a return that never got logged back in, a supplier short-ship nobody flagged. Typically two to four hours a week even in a tidy operation, more once you're running multiple sales channels against one stock pool, because a sale on one channel has to update stock everywhere else before someone oversells a line that's actually gone.

Supplier chasing. Confirming orders, chasing late deliveries, sorting credit notes for damaged stock - fifteen to thirty minutes per supplier per week sounds small until you have a supplier list running into double figures, at which point it's a standing block of hours that mostly lives in your inbox rather than a system, and the same three or four questions get typed out fresh each time.

Returns processing. Read the email, check the order against stock, issue the label, log the return, update inventory, process the refund. Twelve to twenty minutes each, done properly, which adds up to half a day gone in a week where returns run high - not unusual after a promotional push, a size-heavy range, or the run-up to a peak trading period, when return volumes typically climb well above the rest of the year and the same team is trying to process them alongside a busier-than-usual order book.

Listings and reviews. Writing descriptions, titles and search-friendly copy for new stock at twenty to forty minutes a product, plus the newer job of checking your review flow is genuinely compliant - no incentivised five-star requests, no quietly buried complaints, no checkout pricing that doesn't match what the DMCCA now expects shown upfront. A thirty-SKU range used to be a two-day job on its own; the compliance layer adds a further, ongoing check nobody built into the original process.

Add it up across a retailer's typical week and the figure UK online retailers we've assessed land on is close to nine hours per person, per week, which at a blended rate works out to somewhere around £12,400 a year in lost admin time for each person on the team - use the calculator to run your own numbers against your own rates rather than ours, since a shop on 40 SKUs and a shop on 400 will land in very different places.

What's safe to hand to AI, and what still needs a person

The reviews rules make this section more important than it used to be, not less. AI can draft a product description, draft a customer reply, and flag a stock discrepancy for you to check. It cannot decide, on your behalf, that a batch of five-star reviews is genuine, or that a total price shown at checkout is compliant. Those are judgement calls with legal weight behind them now, and they stay with a person.

What tends to be safe: drafting listings and SEO copy from your existing brand voice and product data, so a human edits and approves rather than writes from scratch; drafting saved replies for the "where's my order" and "can I swap the size" questions that make up most of a day's customer service volume; and flagging stock discrepancies and overdue supplier orders for a human to act on rather than resolving them unsupervised. Shopify Magic already does a version of the drafting work for stores on that platform, and most shops use only a fraction of what they're already paying for it.

What isn't safe to fully automate: anything that touches whether a review is genuine, whether a price is shown correctly under the new rules, or a customer complaint that needs actual judgement rather than a template. Customer data also needs a business-account plan that keeps it private and doesn't train on it - worth checking before any tool goes near your order history. You decide what goes near a tool and what stays manual; the assessment's job is to draw that line clearly for your operation, not to hand the whole workflow to a script and hope.

A returns queue, before and after

Take one task on its own: processing a single return. Before any change, a typical sequence looks like this - open the email, search the order in the system, check it's within the returns window, generate a label, send it, wait, log the item back into stock when it arrives, process the refund, close the ticket. Call it fifteen minutes when nothing goes wrong, longer when the order number doesn't match or the size doesn't tally with what's on file. Across a week with forty returns, that's roughly ten hours of work - more than a full working day for one person.

After mapping the workflow and adding a structured AI layer at the points that genuinely repeat - drafting the return confirmation and label instructions automatically, matching the return to the order without a manual search, flagging the refund for a one-click approval rather than a full re-entry - the same forty returns typically come down to three or four hours of hands-on time. The steps that need a person - checking a genuinely disputed return, deciding on an exception outside policy - stay with a person. The steps that were pure repetition don't.

That's the pattern across most of these tasks. The time doesn't disappear because the AI is clever. It disappears because someone mapped the actual steps first and only then decided which ones no longer needed a human doing them by hand, and which ones did.

Find your own number

Averages are a starting point, not your answer. A shop running heavy multi-channel stock will lose more time to reconciliation than one selling direct; a shop with a generous returns policy will lose more to processing than one with a tighter window; a shop mid-way through bringing its reviews and pricing in line with the DMCCA will have a chunk of one-off work the averages don't capture at all. The only way to know your own figure is to look at your own week.

A HoursBack Assessment is a 60-minute conversation that maps exactly where your hours go - stock, suppliers, returns, listings, reviews compliance and customer queries - and hands back a prioritised, plain-English report within two working days, built around named tools and step-by-step recipes for your operation, not a generic list. See where your shop's hours are going. It's £799, and for a shop running more than 100 SKUs it typically pays back in under a month; if the report doesn't show at least five hours a week worth reclaiming, that comes straight back.

Before you're ready to book, our piece on why most AI automation fails is worth a read - it covers why automating a messy returns or listings process just produces the same mess faster, and what to fix first. And if you want a two-minute read on where the friction sits before committing to anything, the free AI Readiness Quiz gives you a rough picture at no cost.