External traffic seems like an easy way to bypass expensive marketplace advertising: place a link with a blogger, launch ads on VK, or buy a Telegram post. But the marketplace receives that traffic in its own environment, where price, delivery, ratings, listing content, alternatives, and promotions affect the decision. An advertising channel buys a visit, not a guaranteed order.

So “how much did a click cost?” is almost useless without answers to the next two questions: how many visitors completed a purchase, and how much money remained after all variable costs? External media buying begins with unit economics, before the media plan.

Calculate margin before advertising

Take the actual selling price after discounts. Subtract cost of goods, platform commission, logistics, storage, return processing, taxes, and expected losses from uncollected orders. This gives the contribution of one completed purchase before marketing. It is the absolute ceiling for acquisition spending, rather than a target bid: the business still needs profit and a buffer for fluctuations.

Calculate by product or a narrow group with similar economics. A store's average margin can hide one SKU funding another's unprofitable promotion. Count repeat purchases separately only if you can connect them to the customer journey, rather than assuming loyalty.

The listing first, scaling second

If a listing converts relevant internal traffic poorly, external traffic is more likely to increase costs. Check relevance to the query, price relative to the category, delivery speed, stock, rating, reviews, and the listing's first screens. Promotion amplifies what is already happening after a visit.

Platforms have their own reports on impressions, clicks, carts, and orders. Wildberries describes the funnel as a progression from impression to click, cart, and order; separate tools allow sources to be compared. Use them alongside advertising-channel tags, but distinguish an order from a completed purchase and an attributed order from an incremental one.

Attribution: what happened because of advertising

Some customers will see an ad and later buy through marketplace search. Others already planned to buy and simply use the link. The attribution window credits both sales to the channel. To estimate incrementality, leave a comparable group without external promotion or alternate periods with and without ads while keeping prices and stock stable.

  • One source, one SKU or a homogeneous group, one promise.
  • Fixed price, listing, and logistics for the duration of the test.
  • A separate link or an available tool for tagging the source.
  • A control period or group without paid promotion.
  • Results measured in completed purchases and margin after returns, rather than clicks.

The minimum budget is a volume of data

A test budget is determined by the number of conversions needed for a decision, rather than an amount that feels comfortable. If a listing converts one visit in fifty, a hundred clicks will produce too much noise. Estimate the baseline conversion first, then set a volume at which differences between channels no longer depend on a couple of random orders.

Do not scale a campaign simply because the first day looks good. Wait for the full delivery and returns cycle. For seasonal products, compare similar weekdays and account for the platform's other promotions. Marketplace economics change faster than a media plan, so keep the calculation date and fees alongside the result.

When external traffic is actually useful

It works well for launching an easily understood product with a strong visual explanation, supporting branded demand, reconnecting with an existing audience, and testing new segments. It works less well for a product without reviews or an advantage, a complex category with a long decision cycle, or an SKU whose margin disappears after logistics.

The right outcome of a test is a decision: which source at which price generates additional margin, and how much traffic the listing can handle without conversion deteriorating. “The channel delivered sales” and everything else are just an attractive revenue report.

Example: revenue grows, profit does not

A product sells for 2,400 rubles. After cost of goods, commission, logistics, storage, the average discount, and a returns reserve, 520 rubles of contribution margin remain before advertising. A campaign brings 300 clicks at 32 rubles and 24 attributed orders. Spending is 9,600 rubles, or 400 rubles per order. The report shows a profitable channel: 120 rubles remain per order. But after delivery, only 18 units are actually purchased, raising the cost per completed purchase to 533 rubles — the economics are already negative.

Even this calculation does not prove the effect. If four completed purchases were expected in the control group without advertising, the campaign added fourteen. The incremental cost per completed purchase is around 686 rubles. Such a test may be useful for gathering reviews or launching a product, but that is a separate investment objective that cannot be called profitable scaling.

What to save in the experiment record

  • Dates, source, segment, creative, and the exact tagged link.
  • SKU, price, discount, stock, and delivery time throughout the test.
  • Clicks, orders, cancellations, completed purchases, and returns within one observation window.
  • Contribution margin before advertising and after advertising spend.
  • Control baseline, test limitations, and the decision on the next budget.

Set scaling rules in advance. For example: increase the budget by 25% if incremental margin is positive after the full purchase-completion window and stock covers demand; stop if the cost of an additional completed purchase exceeds contribution margin for two consecutive periods. These boundaries guard against decisions based on an attractive chart from the first few days.