How to measure advertising effectiveness in agentic commerce: we break down verified data and event context, plus actionable insights for brands, agencies, and the ad market.
TikTok Shop generated $11 bln ₽ in revenue in the first quarter 2026 — representing 1% of all US retail sales and 3% of e-commerce, according to Circana data. Shopping no longer requires leaving the retailer's website: social platforms and AI agents compress the funnel to a single step, breaking traditional ad attribution in the process. Brands lose the ability to track clicks, impressions, or cookies — the agent completes the transaction without these signals, and classic closed-loop retail media models stop working.
How to measure advertising effectiveness in agentic commerce?
The traditional retail media model was built on retailers controlling the customer journey: ad impression → click → add to cart → checkout. Each step was tracked, and brands saw the complete picture. Now consumers discover a product on TikTok, buy it there through the built-in Shop, and never visit the retailer's website. The platform becomes the first party — purchase data stays with it, and the retailer loses exclusivity over transaction information.
AI agents are dealing another blow to attribution. According to Circana, 70% of respondents use them to search for products, 48% to get recommendations, and nearly 50% delegate routine purchases to agents. An agent compares prices, selects a retailer, and places an order—all without traditional attribution markers: no banner impression, no click, no user session that can be tied to an advertising campaign. An attribution system built on impression IDs and cookies simply doesn't see this transaction.
What's Happening with Retailer Data
Retail media networks built their value on first-party data: a customer visited the site, viewed a product, made a purchase—and the retailer knew the entire journey, selling this information to brands. Social commerce breaks that monopoly: part of demand is now generated and fulfilled on TikTok, Instagram (owned by Meta), or other platforms. The retailer only sees the final delivery, but doesn't know which content, which blogger, or which ad drove the order.
«Retailers are losing exclusivity over commercial data — they used to own it completely, but now part of transactions are moving to social platforms,» notes Kiara Barrett, Vice President at Circana.
For the Russian market, the challenge is compounded by regulatory restrictions: Instagram and Facebook (owned by Meta) are blocked, TikTok isn't developing Shop in Russia, and local marketplaces — Wildberries, Ozon, Yandex Market — still maintain control over their data. However, the trend is inevitable: Telegram, VKontakte and other platforms are testing built-in commerce, and once it scales, Russian brands will face the same attribution challenges.
How to measure campaign impact when signals disappear
Traditional attribution models either count the last click or distribute weight across touchpoints. Both require a user identifier — a cookie, device ID, or email. Agent commerce removes these signals: an agent leaves no trace that can be connected to an ad impression from three days ago. The solution is to shift toward incremental measurement, which compares the behavior of control and test groups rather than tracking individual customer journeys.
Incrementality shows the net sales lift that a campaign delivered. A brand divides its audience into two groups: one sees the ad, the other doesn't. The difference in purchases is the increment. This model works without personal identifiers, but it requires purchase data across the entire market — not just from a single retailer. If a customer is counted as a new client on Ozon but previously purchased the brand on Wildberries, the first-party data from one marketplace will show a conversion, while the actual increment is zero.
Third-party firms like Circana, Nielsen, and local panels gather purchase data through receipts, household panels, and bank transactions. This costs more than built-in marketplace analytics, but provides the full picture: a brand can see whether its market share grew, whether customers switched from competitors, or if the campaign simply redistributed sales across channels.
Marketer's checklist: how to adapt your measurement approach
- Request from your retail media network not just conversion data on their platform, but rest-of-market metrics: did the customer buy your brand elsewhere before or after the campaign.
- Include control group tests in your media plan: allocate 10–15% of your budget to incrementality measurement, not just last-click ROAS.
- Connect third-party panels (Circana, Nielsen, Romir and Data Insight in Russia) if you work across multiple marketplaces: one platform's first-party data won't show you sales cannibalization.
- Reassess your KPIs: if an algorithm selects products by price and rating, display advertising works at the top of the funnel—track awareness and consideration growth, not just direct sales.
- Negotiate data access at SKU level with the retailer: if an algorithm is buying a specific item, it matters to understand which attributes (price, reviews, availability) influenced the choice, so you can optimize the product card.
How AI Agents Are Shifting Brand Priorities
An AI agent doesn't view banners — it analyzes product descriptions, ratings, prices, inventory availability, and delivery speed. If a brand has invested in display advertising on a marketplace but its product card underperforms competitors on these parameters, the agent will choose a different product. The classic funnel of "awareness → consideration → purchase" transforms into instant attribute comparison, and the winner is whoever is optimized for the agent's criteria.
For the Russian market, this means budget reallocation: some funds from performance advertising on marketplaces will need to shift toward content marketing (so agents find brand mentions), product card optimization, review and rating management. If an agent is trained on public data — articles, reviews, blogger posts — a brand must be present in that context, or the agent simply won't include it in its recommendation list.
Purchase Data as the New Standard for Measurement
When impressions and clicks stop being reliable signals, the only source of truth is the purchase itself. Purchase data becomes the foundation for all calculations: incrementality, market share, brand switching. In the US, retailers and measurement companies are already building unified measurement frameworks where transaction data is collected from all channels — online, offline, social commerce — and consolidated into a single view.
In Russia, the main barrier is data fragmentation. Wildberries, Ozon, and Yandex Market don't share information with each other, banks are restricted by personal data protection laws, and household panels cover 5–10% of the market. Brands are forced to purchase data from each marketplace separately and manually consolidate it. The solution is to require retailers to integrate with third parties (Data Insight, Romir) and include contractual access to aggregated category data, not just your own sales figures.
Frequently Asked Questions
How to measure the effectiveness of influencer advertising when a customer uses an AI agent
An AI agent considers brand mentions in public sources—articles, reviews, blogger posts—when generating recommendations. Measure not direct link clicks, but growth in brand mentions in search results and increased consideration in surveys: if more people view your product as a purchase option after partnering with a blogger, the agent is more likely to include it in their shortlist. Additionally, track sales growth through rest-of-market data: compare your market share before and after the campaign across all channels, not just one marketplace.
Which metrics matter for retail media when traditional attribution doesn't work
Shift from last-click ROAS to incremental lift: compare sales between your test group (exposed to ads) and control group (not exposed). Layer in upper-funnel metrics—brand awareness, ad recall, consideration—because display advertising now influences agent recommendations indirectly through brand recognition. Monitor share of voice in your category: if an agent is trained on public data, it matters that your brand gets mentioned more often than competitors. Use purchase data to calculate market share and cross-channel cannibalization.
Should Russian brands prepare for agentic commerce now
Yes, although widespread adoption of AI shopping agents in Russia lags the US by 1–2 years. Telegram, VKontakte, and Yandex are testing built-in assistants to help customers choose products, and marketplaces are rolling out AI-powered recommendations in search and catalogs. Start optimizing product cards for machine readability right now: structured attributes, target keywords in descriptions, strong ratings, and current reviews. Test incremental measurement approaches to understand each channel's true contribution while last-click attribution still works—this baseline will let you compare results when signal loss begins.
* Instagram and Facebook are owned by Meta, an organization recognized as extremist, with its activities banned in the Russian Federation.
In brief
- TikTok Shop generated $11 billion in Q1 2026, capturing 1% of all retail sales and 3% of e-commerce from traditional retailers — social commerce is no longer just a discovery channel, but a full-fledged transaction platform.
- 70% users leverage AI for product discovery, 48% rely on it for recommendations, and nearly 50% delegate purchases to agents that leave no traditional attribution signals — cookies, clicks, impressions.
- Closed-loop measurement in retail media loses accuracy: when an agent completes a transaction without identifiers that link the purchase to the ad, brands cannot see the campaign's impact.
- Incrementality measurements with control groups and rest-of-market data are becoming the standard: they demonstrate net sales lift and account for channel switching that a single retailer's first-party data cannot capture.
- Purchase data — the only source of truth when impressions and clicks disappear: actual purchase facts allow you to calculate incrementality, market share, and campaign effectiveness without personal identifiers.
- Russian brands should prepare now: optimize product cards for AI agents, request rest-of-market data from marketplaces, reallocate budget from performance advertising to content and reputation management — these are the factors that agents will rely on when making selections.
ETC helps brands build a measurement strategy for influencer marketing across multi-channel scenarios and signal loss—from metric definition to incrementality interpretation.
CEO comment
Leonid Naumtsev CEO, ETC AGENCY