In short: To measure the revenue a chatbot drives, track four events: product view, add to cart, checkout and purchase. Then match the purchased products to the products the bot recommended, using the SKU (stock keeping unit). Puhera records these events, calculates chatbot-attributed revenue with a SKU-based attribution model, and shows it in summary, funnel and top-products reports.
Why is chatbot revenue hard to measure?
On an online store, a buying decision rarely happens in a single moment. A shopper looks at a product, asks the chatbot about sizing or shipping, browses a few other pages and completes the order later. Asking "did they open the chat?" tells you very little about the bot's share of that journey.
The core problem is that most chatbot metrics measure activity, not outcomes. Conversation counts, message volume and satisfaction scores show how busy the bot is. They do not show how much it earns. Yet the question a store owner or leadership team actually asks is simple: how much did this bot sell? Answering it requires a setup that connects the conversation, the product and the order.
Which events should you track?
An event is a single recorded action a visitor takes on your site. Four core events are enough to measure chatbot revenue:
- Product view: the visitor opened a product page. If the bot recommended a product and the visitor went to it, this is the first concrete sign of interest.
- Add to cart: the visitor put the product in their cart. Intent gets stronger here; the product is no longer just a curiosity.
- Checkout: the visitor moved to the payment step. This middle step is what lets you understand cart abandonment.
- Purchase: the order was completed. Revenue is calculated from the products and amounts in this event.
Knowing which product each event relates to is essential. That is why each event should carry the product's SKU, the unique stock code your store assigns to it. Without the SKU you cannot draw the line from "the bot recommended product X" to "the visitor bought product X."
What is SKU-based attribution and how does it work?
Attribution is the rule that decides which source gets credit for a sale. Marketing teams use many attribution models, such as last click or first click. For a chatbot, the clearest approach works at the product level: SKU-based attribution.
The logic is straightforward:
- During the conversation, the chatbot recommends one or more products. The SKU of each recommended product is recorded.
- The visitor keeps shopping, generating view, add-to-cart, checkout and purchase events.
- The products in the purchase event are compared with the SKUs the bot recommended.
- The value of the matching products is reported as chatbot-attributed revenue.
The strength of this model is that it does not credit the bot with the whole basket. Say a clothing store's bot recommends a jacket and the shopper buys the jacket plus a pair of socks. Only the jacket counts toward chatbot revenue. The report shows the bot's real contribution without inflating it, and that conservative approach makes the numbers easier to trust.
How do you read a conversion funnel?
A funnel lines the events up in order and shows how many visitors remain at each step. Views form the widest step; purchases form the narrowest. When you read a funnel, the drop between steps matters more than any single number.
- Recommendations happen but views are low: the bot may be describing the product poorly, or the suggestion may not match what the visitor asked. Review your product data and the bot's directive.
- Views are high but add-to-cart is low: price, stock or size information may not be convincing. Make sure the bot reads prices and availability from an up-to-date source.
- Add-to-cart is high but checkout is low: the friction is usually shipping cost, payment options or return terms. Answer those questions clearly in your knowledge base.
- Checkout is high but purchases are low: there may be a technical issue or a trust gap at payment. A handoff to a live agent can help at this point.
Funnels are most useful over time. When you add a new FAQ to the knowledge base or change how the bot recommends products, watching which step improves turns guesses into decisions.
What are the common measurement mistakes?
- Crediting every order to the bot: counting every order from a visitor who opened the chat inflates the figure. Product-level matching prevents this.
- Inconsistent SKUs: if the code in your catalog differs from the one in your order system, nothing will match. Check that variants such as size and color use consistent codes.
- A stale catalog: if the bot recommends products that are discontinued or out of stock, you lose the sale and the measurement loses its meaning.
- Looking only at revenue: revenue matters, but funnel steps and top products tell you what to improve.
- Deciding on a few days of data: short windows can reflect a campaign or a seasonal spike. Compare similar periods when you evaluate a change.
Measurement readiness checklist
- Confirm your product catalog is connected to the bot and kept up to date.
- Verify that every product and variant has a unique SKU.
- Test that view, add-to-cart, checkout and purchase events are being recorded.
- Define recommendation behavior in the bot's directive: when should it suggest products, and how many should it show?
- Save your first weeks of funnel and top-products reports as a baseline.
How does Puhera report chatbot revenue?
Puhera's conversion and revenue tracking records product view, add-to-cart, checkout and purchase events. Its SKU-based attribution model measures purchases of products the AI recommended and shows chatbot-attributed revenue directly. Reporting comes in three views: a summary report for the big picture, a funnel report that shows where visitors drop off, and a top-products report that lists what sells best.
Attribution only works when recommendations come from a current catalog. Puhera can sync your product catalog from a Google Merchant feed or from T-Soft; our Google Merchant feed guide covers feed preparation. The recommendation side is explained on the product recommendation chatbot page. If you want the bigger picture first, start with what is an e-commerce chatbot?
To try it with your own data, you can use Puhera free for 7 days with no credit card: start your free trial.
Frequently asked questions
What is the difference between chatbot-attributed revenue and total revenue?
Total revenue is the value of every order on your site. Chatbot-attributed revenue is only the value of products the bot recommended whose SKU matches a purchase event, so it is always a subset of total revenue.
What happens if a shopper buys other products along with the one the bot recommended?
With SKU-based attribution, only the recommended products are credited to the chatbot. The rest of the basket is not added to chatbot revenue, so the report does not overstate the bot's contribution.
Do I need to connect a product catalog to measure revenue?
In practice, yes. The bot can only recommend catalog products by their SKU, and without a SKU there is no way to match a recommendation to a purchase.