---
title: "B2B Data Cleansing: The 6-Step Loop (2026)"
description: "B2B data cleansing that only deletes shrinks your file. The 6-step loop, the merge rule, what to verify, and the real cost per 1,000 rows. Start free."
canonical: "https://derrick-app.com/data-quality-management/b2b-data-cleansing"
category: "Data Quality"
updated: "2026-07-22"
---

# B2B Data Cleansing: How to Clean a List Without Shrinking It

> B2B data cleansing is the recurring loop that keeps a business database accurate, consistently formatted, unique and complete enough to act on. It runs in six steps: profile, standardize, deduplicate by merging, verify what is expensive to get wrong, refill the holes, and stamp what you did. The mistake that costs the most is treating it as a subtraction, because a bounced email in B2B is almost always a job change rather than a dead person, so the row should be refilled and not deleted. Judge a pass on the usable record rate, not on how many records you removed.

*Canonical: https://derrick-app.com/data-quality-management/b2b-data-cleansing* · *Data Quality*

---

A B2B database never breaks all at once. It breaks one field at a time, quietly, while nobody is looking at it. A job change here, a rebrand there, a duplicate created by a form fill, a phone number that was formatted by hand in 2023. Each one is trivial. Together they are the reason a campaign that should have worked did not.

B2B data cleansing is the recurring operation that pushes back against that drift. Most guides describe it as a subtraction: find the bad rows, remove them, feel cleaner. That framing is where teams lose. A file you only subtract from gets smaller every quarter, and a smaller file is not a better one. The number that decides whether a cleansing pass was worth doing is not how many records you deleted. It is how many records you can still contact afterwards.

This guide covers what a cleansing pass actually touches, the four defects hiding in every list, why deletion alone makes the file worse, the six-step loop that works, how to merge duplicates without losing the good half, which fields deserve paid verification, what the whole thing costs on a list of 1,000 rows, and the cadence that keeps you off the treadmill.

## What B2B data cleansing actually is, and what it is not

B2B data cleansing is the process of making every record in a business database accurate, consistently formatted, unique, and complete enough to act on. It covers four distinct operations that are usually lumped together: standardizing formats, resolving duplicates, verifying that values are still true, and refilling the fields that are empty or have gone stale.

What it is not: a one-time project, a purge, or a synonym for deleting old contacts. Those are the three misreadings that produce the classic outcome, a CRM that got noticeably smaller and works no better than before.

The distinction that matters most is between **cleansing** and **auditing**. An audit tells you what is wrong and how much of it there is. Cleansing changes the records. Running the second without the first is how teams end up standardizing a field nobody uses while the field that actually drives routing stays broken. If you have never measured your file, start with the [20-point data audit checklist](https://derrick-app.com/data-quality-management/audit) and come back here with numbers.

The second distinction is between cleansing and *enrichment*. Cleansing corrects and removes. Enrichment adds. In practice they are the same loop, because every correction and every removal opens a hole, and a hole is just a missing field waiting for an enrichment call. Treating them as two separate projects owned by two separate people is why so many databases are simultaneously clean and useless.

![four defects in a B2B data cleansing list](/illustrations/articles/b2b-data-cleansing/four-defects.webp)Each defect has its own fix, cost and detection method.

## The four defects hiding in a B2B list

Every dirty record fails in one of four ways, and each failure mode has its own fix, its own cost, and its own detection method. Sorting your problem into the right bucket before you touch anything is the difference between a two-hour pass and a two-week one.

| Defect | What it looks like | How you detect it | What fixes it |
| --- | --- | --- | --- |
| Inconsistent | Phone written five ways, country as FR and France and Fr | Value counts on each column | Format normalization, free and instant |
| Duplicate | The same person on three rows, each holding a different half of the truth | Key matching on email and domain | Merge with a precedence rule, never blind delete |
| Stale | Title, company or email that was correct last year | Verification, or a timestamp older than your refresh window | Re-verify, then re-enrich the row |
| Incomplete | Empty cells in the fields your segmentation depends on | Fill rate per column | Enrichment from a source that has the field |

Notice that only one of the four is fixed by removing anything. Inconsistency is a formatting job. Incompleteness is an addition. Staleness is a refresh. Duplication is the only one that legitimately reduces your row count, and even there the goal is to end up with one complete record rather than two half-empty ones.

The reason lists rot in this specific order is documented in our [2026 B2B data decay benchmark](https://derrick-app.com/data-quality-management/b2b-data-decay-benchmark-2026): contact data degrades at roughly 2.1 percent per month, which compounds to somewhere between a fifth and a third of a file within a year, and person-level fields such as title and current company move far faster than firmographic ones.

![B2B data cleansing delete row versus refresh](/illustrations/articles/b2b-data-cleansing/delete-vs-refresh.webp)Refreshing keeps the identity and finds the person's new company.

## Why deleting bad rows makes the file worse

Here is the trap. You run a verification pass on 10,000 contacts. 1,400 emails come back invalid. You delete them. Your bounce rate drops, your dashboard turns green, and you now have 8,600 contacts instead of 10,000.

But an invalid email is almost never an invalid *person*. In B2B the overwhelming cause of an email going bad is that the human behind it changed employer. That person still exists, still buys, and is frequently a warmer prospect at their new company than they were at the old one. Deleting the row throws away the name, the title history, the company they came from, and the fact that someone on your team once talked to them. You paid to acquire that record and you just paid again to destroy it.

The alternative costs almost nothing: keep the row, mark the email as dead, and send the identity back through enrichment to find where the person is now. What was a deletion becomes a refresh, and the record comes back more valuable than it was, because a recent job change is one of the better reasons to reach out.

This is why the metric to track across a cleansing pass is the **usable record rate**: the share of rows that carry a valid, current, reachable contact point plus the fields your targeting needs. Records removed is a vanity number that always goes up. Usable records is the one that can go down while your dashboard says everything improved. The related trap, deleting rows because a field is empty rather than filling it, is covered in the guide on [handling errors and missing data](https://derrick-app.com/data-quality-management/missing-data).

![B2B data cleansing loop six steps](/illustrations/articles/b2b-data-cleansing/six-step-loop.webp)The six steps only work in this order.

## The B2B data cleansing loop, step by step

The loop has six steps and they only work in this order. Running verification before deduplication means paying to verify the same email twice. Running enrichment before standardization means enriching on a malformed key and getting nothing back.

**1. Profile the file.** Before changing a single cell, get the fill rate of every column and the distinct value count of every categorical field. Ten minutes with a pivot table tells you which of the four defects you actually have. Most teams discover their real problem is not the one they came to fix.

**2. Standardize formats.** Trim whitespace, unify casing on names, normalize country and state values to one spelling, put phone numbers into a single international format, and strip tracking parameters and trailing slashes from URLs so that one company is one domain. This step is free, reversible, and it is the one that makes deduplication work, because two rows that look different to a matcher are usually just two spellings of the same thing.

**3. Deduplicate by merging.** Detailed in the next section. The output is one row per real entity, holding the best available value for each field.

**4. Verify what is expensive to get wrong.** Emails before a send, phones before a calling block. Not everything, and not on a calendar. Details in section 06.

**5. Refill the holes.** Every row that lost a value in steps 3 and 4 goes back through enrichment. This is the step that separates a cleansing pass that pays for itself from one that just shrinks the file.

**6. Write down what you did.** Stamp each row with the date it was last verified and the date it was last enriched. Without those two columns your next pass has to redo everything, because you cannot tell what is fresh. Two extra columns save you the entire cost of the next cycle.

![B2B data cleansing deduplication merge partial rows](/illustrations/articles/b2b-data-cleansing/merge-dedup.webp)Merge partial rows into one complete record, per field.

## Deduplication: merge, never delete

Duplicates in B2B lists are rarely identical rows. They are partial rows. One came from a webinar and has the phone number. One came from a form fill and has the correct job title. One came from a list import and has the company domain. Delete two of the three and you keep whichever fragment happened to survive, which is not a decision, it is a coin toss.

Merging correctly needs three things settled in advance.

**A matching key, in order of reliability.** Professional email is close to unique and is the strongest key you have. Company domain plus last name is the next best. Full name plus company name is the weakest, because people share names and companies have several legal spellings, and it should require a human look before merging. Detection itself is free in Google Sheets: [Find Duplicates](https://derrick-app.com/features/find-duplicates) is an unlimited feature on the free plan, so flagging the collisions costs nothing and you spend your judgment on the merge rule instead.

**A precedence rule, per field and not per row.** The right question is never which row wins. It is which value wins for each column: most recent for job title and company, non-empty over empty for phone and email, longest for description fields, and the value carrying the higher confidence score where you have one. Scoring per field is exactly what the [confidence scoring guide](https://derrick-app.com/data-quality-management/confidence-score) is for.

**A record of what was merged.** Keep the losing values in an archive column or a separate tab for one cycle. Every team that skips this discovers within a month that one of its merge rules was wrong, and without the archive there is nothing to roll back to.

One more rule that saves painful mornings: never auto-merge across companies. Two people with the same name at two different domains are two people, and a matcher that is loose enough to combine them will eventually put a prospect's phone number on their namesake at a competitor.

## Verification: which fields are worth paying to check

Verification is the only step in the loop with a variable cost, so it is the only one where a rule is worth writing down. The rule is simple: verify what is expensive to get wrong, at the moment it matters, and not before.

An email address is worth verifying immediately before a send, because the cost of getting it wrong is not one wasted message, it is the deliverability damage a bounce does to every message that follows. [Email Verification](https://derrick-app.com/features/email-verification) is available on the paid plans at 1 credit per email, billed on the result. Verifying 2,000 addresses the day before a campaign is a small, predictable line item against the alternative.

A phone number is worth verifying before a calling block, for the same reason at a smaller scale: a rep working through a list of dead numbers loses an hour and some morale.

What is usually not worth paying to verify: firmographic fields you can re-enrich for less than the verification would cost, and anything on a segment you have no plan to contact this quarter. Verifying a dormant segment on a schedule is the most common way to spend a budget on a number that will be stale again before anyone dials it.

Two nuances the SERP tends to skip. First, a verification result is a probability, not a fact, and the correct handling of an ambiguous verdict is a separate branch of your workflow, not a rounding decision. Second, verification tells you an address is deliverable, not that the person still works there, which is why a clean verification pass on a stale file can hand you a green dashboard on a file full of catch-all domains. The difference between an accurate value and a merely present one is its own question, and it is worth settling before you trust a green dashboard.

![B2B data cleansing refill in Google Sheets](/illustrations/articles/b2b-data-cleansing/sidebar-refill.webp)Choose a feature, map input to output, run row by row.

## Refill: the step that turns cleansing into a net gain

Refilling is where the file grows back, and it is the step most cleansing guides mention in one line before moving on. It deserves more, because it is the only reason to run the other five.

Work from the identity outwards. A row that lost its email still has a name and a company, and a name plus a company is a perfectly good input key. A row that lost its company still has a person, and a person can be resolved back to their current employer. The identity is the durable part of a B2B record. Contact points are the perishable part sitting on top of it, and you should treat them as replaceable rather than as the record itself.

Three refills cover most of what a cleansing pass destroys. Company fields, headcount, industry, location, come back from the company entity with [Enrich Companies](https://derrick-app.com/features/enrich-companies), a free-plan feature at 1 credit per company. Person fields, current title, seniority, current employer, come back with [Enrich Leads](https://derrick-app.com/features/enrich-leads), also on the free plan at 1 credit per profile. And the cheapest refill of all is the one people do by hand: splitting an email into a first name, a last name and a domain, which [Find Names and Domains by Email Addresses](https://derrick-app.com/features/find-names---domains-by-email-addresses) does unlimited and free.

Measure the pass on the usable record rate before and after. A good cleansing cycle ends with fewer rows than it started with and *more* contactable ones. If your row count fell and your contactable count fell with it, you ran a purge, not a cleansing pass.

![B2B data cleansing cost on 1000 rows](/illustrations/articles/b2b-data-cleansing/cost-1000-rows.webp)Most of the cost sits in verification, billed per result.

## What a B2B data cleansing pass costs on 1,000 rows

Concrete arithmetic, because the category is full of pricing that only exists on request. Take a list of 1,000 B2B contacts that has not been touched in a year.

Profiling and standardization: free, spreadsheet formulas, roughly an hour of attention the first time and ten minutes every time after, once you have saved the formulas.

Duplicate detection: free and unlimited. Say the pass surfaces a hundred collisions on a list of that age, which merge down into fifty rows, leaving you at 950.

Email verification on the 950 remaining: 950 credits, billed per result. Take 15 percent invalid for the arithmetic, a deliberately conservative read against the annual decay range in the benchmark above, leaving roughly 810 valid and 140 rows that need a new contact point.

Refill on those 140 rows plus the empty columns you found during profiling: 1 credit per profile or per company, so a few hundred credits depending on how many holes the audit exposed.

Total order of magnitude: around 1,200 to 1,400 credits for a full pass on 1,000 rows, most of it in verification. On the free plan you get 100 credits per month, which is enough to run this loop properly on a segment of about 80 rows, or to test the whole workflow end to end before committing. Paid plans start at 9 euros per month, and the per-credit cost falls as volume rises, which is what makes the same loop viable at 1,000 rows and at 100,000.

The comparison that matters is not against doing nothing. It is against the cost of the campaign you would have run on the uncleaned file: the wasted sends, the deliverability damage, and the rep hours spent on rows that were never going to answer. That arithmetic is laid out in the guide on [the real cost of missing B2B data](https://derrick-app.com/data-quality-management/cost-of-missing-data).

## Cadence: cleanse at the point of use, not on the calendar

The last decision is when to run the loop, and the standard answer, quarterly, is the wrong shape. Decay is continuous but your need for accuracy is not: a record only has to be correct at the moment you use it. A contact you will not touch for six months does not need to be verified today, and one you are emailing tomorrow needs to be verified tomorrow regardless of when it was last checked.

That gives a two-tier cadence. Verification runs **at the point of use**, immediately before a send or a calling block, on the segment being activated. Standardization, deduplication and refill run **on the file**, on a slower rhythm, quarterly for person-level fields and semi-annually for firmographics, because those decay slower. Choosing the interval per field type rather than for the whole database is the subject of the [data freshness guide](https://derrick-app.com/data-quality-management/freshness).

Then there is the prevention half, which costs nothing and removes most of the work. Validate formats at the point of entry so malformed values never enter. Enforce one canonical spelling for country, industry and company name at import. Stamp every row with a verification date. A file with entry validation needs a fraction of the cleansing that an open one does, and the teams that never seem to have a data problem are not cleaning harder, they are letting less dirt in.

Run it once end to end on a single segment before you try to fix the whole database. Two hundred rows, all six steps, and measure the usable record rate before and after. That number tells you what the full pass is worth, and it is a far better argument for the time than any benchmark someone else published.

## Use Derrick in Claude (MCP) or via API

Derrick isn't only a Google Sheets add-on. The same B2B data enrichment runs as an MCP server (use it directly inside Claude and other AI agents) and as a REST API:

- **Claude / AI agents (MCP)**: connect the Derrick MCP server, then enrich from chat. Setup: /mcp
- **REST API**: call the same enrichment endpoints from your own stack (Standard plan and up). Docs: https://app1.derrick-app.com/api/v1/docs/

## FAQ

### What is B2B data cleansing?
B2B data cleansing is the recurring process of making every record in a business database accurate, consistently formatted, unique and complete enough to act on. It covers four operations: standardizing formats, resolving duplicates, verifying that values are still true, and refilling fields that are empty or stale. It is not a purge and it is not a one-time project, because the underlying data keeps decaying whether you run it or not.

### What is the difference between data cleansing and data enrichment?
Cleansing corrects and removes, enrichment adds. In practice they belong to the same loop: every correction and every deletion opens a hole in a record, and that hole is a missing field waiting for an enrichment call. Running cleansing without a refill step is what produces a database that is technically clean and commercially useless, because the row count fell and the number of contactable people fell with it.

### Should I delete contacts whose email bounces?
Usually no. In B2B the main reason an email stops working is that the person changed employer, and that person is often a warmer prospect at their new company than they were at the old one. Keep the row, flag the email as dead, and send the identity back through enrichment to find where they are now. Deleting the row throws away the name, the title, the history and the acquisition cost of that record.

### How often should I clean a B2B database?
Use two tiers rather than one calendar. Verification runs at the point of use, immediately before a send or a calling block, on the segment being activated, because a record only has to be correct at the moment you use it. Standardization, deduplication and refill run on the whole file more slowly, roughly quarterly for person-level fields such as title and employer, and semi-annually for firmographic fields, which decay more slowly.

### How do I remove duplicates without losing data?
Merge instead of deleting. Pick a matching key in order of reliability, professional email first, then company domain plus last name, then name plus company name for human review only. Then set a precedence rule per field rather than per row: most recent wins for title and employer, non-empty beats empty for phone and email, and the higher confidence score wins where you have one. Keep the losing values for one cycle so a bad merge rule can be rolled back.

### What does a cleansing pass cost on 1,000 contacts?
Profiling, standardization and duplicate detection are free, since Find Duplicates is unlimited on the free plan. The variable cost is verification at 1 credit per email on the paid plans, plus refill at 1 credit per profile or per company. On a 1,000 row list that has not been touched in a year, expect an order of magnitude of 1,200 to 1,400 credits for a full pass, most of it in verification.

### Can I clean a B2B list in Google Sheets?
Yes, and for most B2B teams the spreadsheet is the right place for it, because the file you clean is usually the file you are about to work from. Derrick runs as a sidebar inside Google Sheets, so duplicate detection, email verification and re-enrichment happen on the rows in front of you rather than in an export and re-import cycle. The free plan includes 100 credits per month, and the same workflow is available through the API and MCP from the Standard plan.

### Which metric tells me whether a cleansing pass worked?
The usable record rate: the share of rows carrying a valid, current, reachable contact point plus the fields your targeting needs. Records removed always goes up and proves nothing. A good pass ends with fewer rows than it started with and more contactable ones. If both numbers fell together, you ran a purge rather than a cleansing pass.

## Related

- [The B2B Data Decay Benchmark 2026: How Fast Your Data Goes Stale (and the Refresh Cadence)](https://derrick-app.com/data-quality-management/b2b-data-decay-benchmark-2026)
- [How to Measure the Quality of Your Enriched Data: The Complete Guide](https://derrick-app.com/data-quality-management/accuracy)
- [Data Freshness: How Often Should You Refresh Your B2B Data?](https://derrick-app.com/data-quality-management/freshness)
- [Coverage Rate vs Accuracy Rate: Understanding B2B Enrichment Metrics](https://derrick-app.com/data-quality-management/coverage-vs-accuracy)
- [Confidence Scoring: How to Evaluate and Weight the Reliability of Your B2B Data](https://derrick-app.com/data-quality-management/confidence-score)
- [False Positives in Data Enrichment: How to Detect and Fix Them](https://derrick-app.com/data-quality-management/false-positives)
- [Handling Errors and Missing Data in Google Sheets Data Enrichment](https://derrick-app.com/data-quality-management/missing-data)
- [Error Handling and Retry Logic: how to make your B2B enrichment workflows bulletproof](https://derrick-app.com/data-quality-management/error-handling)
- [B2B Data Audit: The 20-Point Checklist (2026)](https://derrick-app.com/data-quality-management/audit)
- [The Real Cost of Missing B2B Data (And How to Fix It)](https://derrick-app.com/data-quality-management/cost-of-missing-data)
- [Data Enrichment Rate by Industry: B2B Benchmarks to Know in 2026](https://derrick-app.com/data-quality-management/enrichment-rate-by-industry)
- [The State of B2B Data Quality 2026: Decay, Cost and Benchmarks](https://derrick-app.com/data-quality-management/b2b-data-quality-benchmarks-2026)