Assess the intent of messaging leads without copying personal conversations

Judge the expressed task, not the person

Define observable categories connected to the service. An explicit request describes a product, service or booking task. A clarification request asks about a material condition such as availability or included scope. Support concerns an existing order. Unsupported requests describe something the business cannot fulfil. Unknown means the conversation does not yet provide enough information. These are proposed operational labels, not personality scores or platform-defined conversion events.

Separate intent, fit and processing status. Someone can have clear purchase intent for a product you do not sell. Someone who only asks the price may still become a suitable customer. Silence after an unanswered message is not proof of low intent. Keep no response from the business, awaiting customer information and completed assessment as different statuses. Otherwise the advertising review will blame customer quality for an operational queue.

Educational examples use invented tasks rather than real messages. 'Can I book the listed inspection next Tuesday?' is an explicit service request with availability to confirm. 'Does the price include collection?' is a clarification of scope, not automatically a weak lead. 'Where is my existing order?' is support. A greeting alone stays unknown until the team has enough context. The examples teach the rubric; they do not predict which conversation will produce revenue.

Perform assessment in the approved business system and export only aggregate categories needed for the decision. A report can show the number of assessed conversations, compatible requests, unknowns and response delays without names, phone numbers or message bodies. Google's data collection policy places conditions on collection and use; this rubric grants no permission to gather or share additional information. It also does not turn pseudonymous records into inherently anonymous data.

Intent, fit and processing are separate

• Intent field: the task expressed by the customer, with unknown allowed. It helps compare what the offer attracts without inferring sensitive personal attributes.
• Fit field: whether the business can fulfil the task and the objective reason if it cannot. This prevents a high-intent but unavailable request from being mistaken for a sales-handling failure.
• Processing field: whether an assessment and response occurred. It keeps unfinished work visible when enquiry volume is compared with useful conversations.

Produce a useful aggregate review

1. Agree on category definitions and neutral examples with the people handling messages. Use the actual service model rather than a generic numerical lead score.
2. Choose a dated review cohort and state what counts as a unique enquiry. Distinguish repeated clicks or follow-up messages from genuinely new customer tasks.
3. Have two reviewers classify a limited sample in the approved environment. Discuss disagreements and improve the definitions before applying them more widely.
4. Record only the needed categories and outcome status. Avoid exporting conversation details to the advertising workspace, shared memo or external AI client.
5. Compare assessed intent and fit with later qualified outcomes. Show unknown and unprocessed proportions instead of excluding them from the denominator silently.
6. Use the findings to clarify the offer or improve response handling. If one traffic source looks weaker, check language, wait time and assessment coverage before recommending an advertising change.

Unknown is a legitimate result

Sources and further reading

• Google advertising policy: data collection and use