Small-business outsourcing research

What share of customer records contain a stale contact detail?

A protocol for measuring stale emails, phone numbers, and addresses, with the source of the change and the customer impact.

What share of customer records contain a stale contact detail?

Research question and scope

This protocol asks what share of customer records contain at least one stale contact detail. The unit is one customer record in the business's system of record, sampled at a defined date. The study measures stale email addresses, phone numbers, and postal addresses, the source of the last update, and whether a recent message failed. It does not judge the customer or the data-entry clerk.

Set the scope to one business, one customer system, and one definition of active customer. Record the fields in scope, the systems that can update them, and the date of the sample. If the business merges two systems during the window, keep the earlier definition for the full sample or analyze the change as a separate period. A new sync can look like an improvement when it is only a different source.

Define the population as every active customer record that has at least one contact field and has interacted with the business in the last twelve months. Exclude prospects and records with no contact detail, and state that exclusion. A record that is a duplicate of another record is a separate unit linked to the canonical record. Test the definition on two ordinary records before collection.

Definitions that make the measure reproducible

Define a stale email as one that has hard-bounced or has not been confirmed in the last twelve months. Define a stale phone as one that the carrier reports as disconnected or that has failed two contact attempts. Define a stale address as one returned by the postal service or contradicted by a recent delivery. Each definition needs a positive and a boundary example.

Do not count an unsubscribed email as stale. An unsubscribe is a preference, not a bad record. Do not count a phone that was not answered as disconnected. A failed contact attempt is not the same as a failed number. The definitions should separate a data-quality problem from a customer-choice problem.

Record the customer type, the acquisition channel, the last interaction date, and the system that owns the record. These strata matter because a record created at a pop-up event and one created through a long-term contract have different update paths. Include records from every channel. Use one category list for the whole sample.

Evidence standard and data dictionary

For each record, capture the customer identifier, the field, the current value, the last update date, the source of the update, and the evidence of staleness. Add whether a recent message failed and whether the customer corrected the detail. Keep the bounce or return record with the sample. A second reviewer should be able to repeat the classification.

Also capture whether the record has a second contact field that still works. A stale email is less severe if a valid phone exists. Record the fallback channel and whether the customer has consented to it. That field turns the rate into a contactability measure.

If the business uses a marketing system and a billing system, record which one is the system of record. A value can be current in one system and stale in the other. The protocol should not count the same customer twice without a link. A canonical identifier is required for every sample row.

Sampling and collection

Draw the sample from the system of record rather than from a list of bounced messages. Record the export date, the filters, the row count, and the selection rule. Include records from every acquisition channel and every region so the result is not dominated by one campaign. If the population is ordered by customer id, take a fixed interval across the list.

Collect the record as it existed at the sample date, not after a later cleanup. If a customer updated a detail after the sample, record the update and analyze the original state separately. Do not infer staleness from a missing field alone. Note the collection date separately from the sample date.

A useful additional measure is the share of records updated in the last ninety days. A high update rate may indicate a healthy process or a recent migration. Record the source of the update and whether it was customer-initiated. That view separates a maintained record from a recently rewritten one.

Analysis and reporting

Report the number of records, the share with at least one stale field, and the share by field type. Show the counts by customer type, channel, and system. Present the share of records with no working contact channel and the share with a recent failed message. Use the same denominator for every rate.

Inspect the stale records and group the causes without naming the customer. A cluster from one acquisition channel may point to an intake form that does not validate the field. A cluster in one system may point to a sync that overwrites a newer value. Report duplicates separately because they distort both the rate and the contactability picture.

A useful additional view is the time since the last successful contact. A record with a stale email but a recent phone conversation is still contactable. Show the last successful contact date beside the stale flag. That view helps the owner prioritize the records that actually affect service.

Limitations and sources of error

The record cannot show whether a customer moved, changed a number, or simply stopped using an email. It can show only that a message failed or a detail was not confirmed. A high stale rate is not proof of poor data entry, because customers change details without telling the business. The protocol also cannot capture a customer who uses a different name or a shared family account.

Bounce data can be delayed, and a soft bounce may be recorded as a hard bounce by mistake. A phone may be temporarily disconnected rather than permanently stale. Record those cases and their treatment instead of dropping them. If the business has no permission to contact a customer by a fallback channel, note the consent limitation.

A useful check is whether a new assistant can update a contact detail and see the change in every system that uses it. Ask them to test one record and record where the change did not propagate. That test measures the integration, not the assistant. Fix the sync before asking for more careful entry.

Review cadence and next action

Give the result to the owner and the customer-operations lead and ask them to choose one observable change. The change might be a validation rule at intake, a quarterly confirmation message, a sync fix, or a fallback contact policy. Record the owner, the effective date, and the measure that will show whether the change worked. A change without a measure becomes an opinion at the next review. Avoid changing the form and the sync at the same time.

Repeat the protocol with the same definitions after the change and compare the same fields. If the stale rate fell but failed messages did not, the records may be updated too late to help. If failed messages fell but the stale rate rose, the business may be suppressing bad records instead of fixing them. Close the review when the owner accepts the finding and the next sample is scheduled.

A useful additional view is the share of records with a working fallback channel. A stale email is a smaller problem when a valid phone number is available and the customer has agreed to be contacted there. Record the fallback and the consent status. That view helps the team prioritize the records that are truly unreachable.

A useful additional view is the share of stale records that were updated by the customer rather than by staff. A customer-initiated update is stronger evidence than an internal guess. Record the source and the date. That view helps the owner decide whether to invest in a self-service update path.

Ask the support team which stale record caused a real service failure in the last quarter. A concrete case is more persuasive than a percentage and often reveals a missing fallback channel. Record the case and the fix. Use it to set the priority for the next data-quality project.

Sources

Frequently asked questions

Does a favorable sample prove the routine caused the result?

No. The design is observational and supports a bounded operating decision, not a universal causal claim.

What should an owner review first?

Start with high-consequence defects, missing denominators, exceptions, and cases that required the owner to reconstruct the evidence.

Related research