When launching search ads on a marketplace, you can manage the flow of buyers in two ways. The first is strict filtering at entry: you control every phrase and pay only for people searching for the item word for word. The second is broad reach: algorithms assemble a pool of queries around your listing on their own, while profitability rests on capping cost per click and using a visual filter on the cover image.
The trouble starts when you copy a familiar setup from one platform to the other. Copy Wildberries query cleaning onto Ozon, or let a WB auto-campaign run loose—and your budget burns up with zero orders. Let's look at why these two platforms' search engines work in opposite ways and how to configure each one without paying for empty clicks.
Why the Two Search Auctions Work in Opposite Directions
On the storefront, the Wildberries and Ozon search bars look identical, but inside the dashboard I see two different models. On Wildberries, auto-campaign billing is built around cost per thousand impressions (CPM)—you pay for the listing's mere appearance on screen, like renting a shelf near the store entrance. So the main task on WB is not showing the product to random passersby.
On Ozon, in "Трафареты" and search, billing is per click (CPC) or per order (CPO). It pays for the platform to show the listing to a broad audience to find those who will click through. In the "Фишечки" section, experts from the supplier channel "Альянс" summed up this difference accurately: on Wildberries we manage relevance through surgical phrase selection, whereas on Ozon the algorithm attaches anything close in meaning to the listing without letting you manually cross out individual keywords. I wrote about how adjacent formats work in my breakdown of Wildberries and Ozon recommendation shelves.
Surgical Filtering on Wildberries: Plus-Phrases and the Cluster Trap
Take a simple example: you sell a set of plastic jars for dry goods. A buyer types "jars for grains," yet one person is looking for glass containers and another for tin ones. If you run ads across all queries indiscriminately on Wildberries, glass seekers will scroll past your listing, click-through rate (CTR) will sag, and money will be drained for nothing.
That is why a WB campaign is tuned surgically. A seller gathers phrase statistics and—via the dashboard or a bidder, a bid-management tool like the "Комета" service from the "Альянс" breakdown—turns on filtering. You lock in plus-phrases (a whitelist of queries, such as "plastic jars for grains") and enable auto-minus to cut out "glass" and "tin." As a result, the advertising cost-to-revenue ratio (ACoS / DRR) stays within 8%.
When cleaning manually, keep WB's search mechanics in mind: the platform groups synonyms and typos into shared folders—search clusters. By moving a single phrase to negatives, you turn off impressions for the entire cluster, so alongside a typo you can easily block your main keyword. You can only add a word to exclusions after at least one impression, and the cap is up to 1,000 negative phrases per campaign.
Broad Pools on Ozon: Sterile SEO, Click Control, and a Cover Filter
Now let's move those same plastic jars to Ozon. In "Трафареты," there is no manual negative-keyword filtering: the neural network scans the title, attributes, annotation, and text on photos, stitching the listing to a wide pool of associations. In the "Мои товары" section, the authors of the "Альянс" channel examined a telling case: a listing racked up 1,600 irrelevant impressions for the query "Галстук и токей," even though those words weren't in the title, description, or attributes. Removing such a keyword manually is problematic, so on Ozon you protect your budget differently:
- Sterile search engine optimization (SEO). Don't write "replacement for glass jars" in the description—the algorithm will start showing plastic to glass seekers. Keep only direct keywords from top competitors.
- Holding the average cost per click. Inside an indivisible cluster, the algorithm's appetite is capped by a per-click bid matched to your margin, as I wrote in the article on ad economics and auto-bid control on Ozon and Wildberries.
- A visual barrier on the cover. So irrelevant buyers don't drain budget with clicks, put a large badge reading "Impact-resistant food-grade plastic" on the main jar photo.
Before starting any search campaign, ask yourself: "Where is my barrier against wasted spend on this platform—in a list of excluded phrases before the impression, or in a clean description and cover image before the click?" Once you have built a plus-phrase perimeter with cluster checks for Wildberries, and stripped extra associations from the text while putting key product traits on the first photo for Ozon, go ahead and launch promotion—the ads will bring orders rather than burning working capital.


