Tenbound Insights
Data strategydata cleansingCRM data quality

Prioritising Data Cleansing: Fix What Blocks Work

CRM data is never fully clean, so cleansing has to be prioritised. The four-way split that decides what to fix first, and the fields that are not worth touching.

Tenbound Editorial / / 3 min read /9 sections

A note on this page. It replaces a URL that carried a named Tenbound framework. That framework's original definition is not available to reproduce accurately, so it is not restated here under its old name. What follows is the prioritisation logic the problem actually requires, written from scratch.

Nobody cleans everything

CRM data decays continuously: people change jobs, companies restructure, numbers get reassigned. A cleansing project scoped as "make the database accurate" fails, because the target moves faster than the work.

The only workable framing is triage. Which records, which fields, in what order.

The two questions

Does this field block work today? A missing direct dial stops a call. A missing industry code stops nothing, unless routing depends on it.

Does fixing it stay fixed? Company name is durable. Job title decays within a year. Direct dial decays faster.

Those two produce four groups, and they take very different treatment.

Stays fixedDecays
Blocks workFix first, once, properlyFix continuously, at the point of use
Does not blockFix opportunisticallyStop collecting it

Blocking and durable: fix first

Account identity, domain, parent-child relationships, and the ICP fields your segmentation depends on.

Worth a one-time project because the result persists. This is also the group where duplicates live, and duplicates are the most expensive data problem a sales team has: two reps working one account, split history, and a handoff that loses half the context.

Blocking and decaying: fix at the point of use

Direct dials, mobile numbers, job titles, email addresses.

The instinct is a bulk refresh, and it is wrong: you refresh 40,000 records, of which reps will touch 3,000 this quarter, and by the time they reach the rest it has decayed again.

Verify at the moment of use instead. It costs less and the accuracy is higher where it matters. Where enrichment is automated, trigger it on entry into a working state rather than on a schedule.

Non-blocking and durable: opportunistic

Firmographics, technographics, headcount bands. Useful for segmentation, and nothing stops if a record is missing one.

Fill on enrichment, never as a project.

Non-blocking and decaying: stop collecting

The group nobody wants to name. Fields added for a report that ran twice, required on a form, maintained by nobody, and trusted by no one.

Audit them and delete. A field that is 30% populated and never queried is worse than absent, because it looks like data.

Start with the reps

The cheapest possible diagnostic: ask three reps what stopped them working this morning. The answers are the blocking list, and they are usually not what a data-quality dashboard shows, because dashboards measure completeness and reps experience blockage.

Completeness is the wrong first metric. A field at 95% completeness that is 40% wrong is more damaging than one at 60% completeness that is right, because reps trust the first.

Measure trust, not fill rate

Two better metrics:

  • Bounce and wrong-number rate on the records actually worked. Accuracy

where it counts.

  • Whether reps check elsewhere before acting. If they verify on LinkedIn

before every call, the CRM is not trusted, and no fill-rate number changes that.

Where this sits

Data quality underpins Market and Signal in the Tenbound Pipeline Architecture Standard. It is the layer beneath the tech stack: every tool above it amplifies whatever the data says, which is why buying engagement tooling on bad data reliably produces more of the wrong activity.

Primary sources

  1. What Is a Sales Development Representative? — Salesforce; accessed 2026-08-26.
  2. Sales Tech Stack — Salesforce; accessed 2026-08-26.