Merchandising

Merchandising, in a digital retail context, is the practice of deciding which products are shown, in what order, and with what prominence across a site. It covers category page ordering, search result promotion, homepage and landing page selections, recommendation placements, and the seasonal and promotional curation that shapes what customers encounter.

The discipline exists because a catalogue of any size cannot be presented neutrally. Something must appear first, and that position substantially determines what sells, since most visitors examine only the initial results. Merchandising is the deliberate exercise of that influence, in service of both commercial objectives and customer relevance, and the tension between those two is the central problem of the discipline.

Purely algorithmic ordering optimizes for observed behaviour and has systematic blind spots. Ranking by popularity entrenches existing bestsellers and buries new products that have not yet accumulated data, which prevents the range from refreshing. Ranking by conversion rate favours low-priced items that convert easily over higher-value ones that generate more margin. Algorithms also cannot know that a line is being discontinued, that stock is constrained, or that a supplier relationship has changed.

Purely manual curation has the opposite failure. Human merchandisers apply commercial knowledge that no algorithm has, but they cannot maintain thousands of categories, they order by preference and habit, and their decisions become stale as inventory and demand shift. The workable arrangement combines an algorithmic base with manual override for the specific cases where commercial judgment adds something, and requires tooling that makes the override straightforward.

Business rules bridge the two and handle the cases that recur. Automatically demoting out-of-stock items rather than showing them, boosting products with strong margin within a relevance band, surfacing new arrivals for a defined period, and suppressing items that cannot be delivered to the visitor's location are all rules that apply consistently without individual attention. Encoding these removes most of the routine merchandising workload.

The constraint that keeps the discipline honest is that customers notice when they are being managed. Results ordered by commercial preference rather than by what the customer asked for produce visible irrelevance, and a search that returns promoted items unrelated to the query trains people to distrust the results and eventually to stop using them. The commercial value of a prominent position depends on the surrounding results being credible.

Seasonal and calendar-driven curation is where manual judgment earns its cost most clearly, because the relevant signals precede the data. A range becomes relevant before customers begin searching for it, and an algorithm ranking on observed behaviour will surface it only after the peak has begun. Planning these transitions ahead of demand, and reverting them promptly once the period passes, is work that no automated system currently does well.

Measurement should extend beyond the placement being optimized. Merchandising decisions move revenue between products rather than only adding it, so a promoted item selling more may simply be displacing another. Assessing at category and basket level, including margin rather than only revenue, and watching return rates for pushed items reveals whether a change created value or relocated it.

Because effective merchandising requires commercial knowledge, catalogue data, and control over presentation, it needs tooling that puts adjustment in the hands of people who understand the range. In practice the interfaces and recommendation placements are designed through product design, the impact is measured with data analytics at category rather than item level, and the day-to-day curation typically sits within an e-commerce department where inventory and commercial responsibility already reside.