Turning a 190,000-Record CRM into a Cleaner Marketing Asset

A CRM that has been collecting records for many years accumulates a lot of history. Old contacts, duplicates, addresses that bounce, people who changed jobs, and records that were never supposed to receive marketing in the first place. Ours had around 190,000 records and nobody could say with confidence how many of them were real, current and marketable.

That uncertainty had costs. Sender reputation had fallen to 32, low enough that deliverability was a problem in its own right. Segmentation was hard to trust. And the marketing team could not increase send frequency without making deliverability worse.

I led an audit of the full database. Every record was validated, classified and either kept, corrected or removed from the marketable pool. I worked across business units to settle who owned which records and what "marketable" meant for each segment, and with external vendors for verification and platform support.

• Validated addresses through Kickbox and isolated 120,000 verified marketable contacts from the 190,000-record base.
• Classified records by market segment so that academic, public library, international schools, corporate, government and nonprofit audiences could be addressed separately.
• Rebuilt segmentation, automated flows and the send strategy inside Klaviyo.
• Set the operating process for how new records enter the marketable pool and how bad ones leave it.

Deliverability was the visible symptom. The actual work was deciding what belonged in the database and building a process that keeps it that way.

I owned the audit and the segmentation rebuild, worked directly with the verification and platform vendors, and worked with teams on adopting the new process. Work spanned multiple business units and required agreement from people who each had a stake in the records.

Sender reputation went from 32 to 71 in six months. Send frequency went from monthly to several times per week while blended open rates rose from 12% to 36%, with top campaigns above 70%. The cleaner database also made the later work possible: segmentation, account-level attribution and the AI-assisted enrichment workflow all depend on it.

No contact records, email addresses or customer data appear on this page. The figures shown are aggregate results from the resume. The process visual is illustrative.