Product Interest vs Purchase Intent: How to Read Customer Messages

Product interest means a customer has expressed curiosity or preference. Purchase intent is stronger evidence that they are considering a transaction under particular conditions. A completed order is a separate event. Keep these distinctions when turning support messages into a demand report.

The practical value is better interpretation, not perfect prediction. A message can reveal a blocked purchase condition without telling you whether the customer will eventually buy.

Use an evidence ladder

This is an editorial coding framework, not a validated probability model or a built-in Outlined scoring system. Do not attach conversion percentages to these categories without a suitable dataset and method.

Read the conversation before classifying it

“Do you have this in black?” might concern a product in a photo, an already purchased item, or an alternative the agent suggested. Identify the referent first. If the product cannot be established, record an unmatched request rather than guessing.

Negation matters. “I wanted blue, but I bought black” contains historical interest and a transaction involving another colour. “I no longer need medium” should not remain a current request. Preserve changes in intent over time rather than accumulating every mention as new demand.

Also distinguish the customer from the agent. A product name in a suggested reply is not evidence that the customer independently requested it.

Decide what one request means

For a buying review, a useful starting unit is one verified requester, product or variant, request type, and reporting window. If the same person follows up three times about one restock, that can represent one request with repeated contact, not three independent shoppers.

Cross-channel deduplication should rely on verified identity. Without that confidence, report channel-level counts and the possibility of overlap. A transparent upper-bound count is better than an unjustified precise number.

Retain the date and availability context. Interest recorded while a variant was sold out should not be compared casually with interest during a heavily promoted in-stock period.

Build a report that preserves uncertainty

Use columns for product, variant, evidence category, distinct-requester basis, source count, unresolved matches, and relevant conditions. Include short anonymised paraphrases that capture the decision the customer was trying to make.

A fictional report might say: “Eight included conversations asked about the green variant; five explicitly required a particular delivery date. Identity overlap across channels has not been resolved.” It should not say “Eight guaranteed orders lost.”

If you connect messages to later orders, document the matching method and time window. An order after a conversation is an association. A customer could have purchased because of an advertisement, a discount, or prior plans.

Use interest to choose the next investigation

Merchandising can compare requests with stock position, sales history, margin, supplier minimums, and seasonality. Marketing can inspect whether a campaign is attracting questions about an unavailable item. CX can identify the factual answer needed to help the shopper proceed.

Do not automatically contact everyone who expressed interest. Follow channel rules, consent requirements, and your approved communication process. Tracking a request does not establish permission to send a future marketing message.

For external context, Shopify documents product-interest topic labels in Inbox. A label is useful for organisation; your team still needs to interpret the underlying evidence for its decision.

Start with the product request tracking worksheet. It records evidence and buying conditions without turning uncertain interest into a sales forecast.