A customer database you can finally trust
Junk emails removed, phone numbers fixed, every record classified, scored and deduplicated, automatically, against live sources.
The problem
How it works by hand
Every business over a few years old has the same database: half the emails bounce, phone and mobile share one field, the same company appears four times with three spellings, and nobody knows which records are worth anything. So campaigns go to dead addresses, reps waste hours on duds, and every report built on top of it is quietly wrong.
A worked example
What a working version looks like
Records flow through an enrichment chain that checks each one against live sources: the company website scraped for current contact details, search results where no website exists, Facebook pages for the businesses that live there, email verification before anything is trusted, and geocoding so every record maps cleanly. An AI pass classifies what each organisation actually does against a controlled vocabulary, not whatever was typed in five years ago. Then the cleanup runs: junk emails and dead websites stripped, phone and mobile separated, quality scores assigned, duplicates merged into one master record with the evidence retained. Out the other side: one clean, classified, scored database.
The exact tools change per business. The shape does not.
What it needs
Honest inputs, nothing exotic
- 01Your existing database, in whatever state it is in (CRM export, spreadsheet, anything)
- 02Agreement on what a duplicate is for your business
- 03The categories you want records classified into (or we derive them)
The payoff
What you get back
Campaigns stop bouncing, reps stop dialling dead numbers, and every report downstream gets more honest. The cleanse also tells you what your database is actually worth: how many real, reachable, relevant contacts you own.
Do it yourself
How you would build this yourself
No course, no upsell. This is the order we would build it in, with the tools named, and a prompt to start from.
- 1
Export everything to CSV and take a copy before you touch a single row. You will be glad of the backup.
- 2
Run email verification first (ZeroBounce, NeverBounce or similar). It is cheap and it tells you within an hour how much of the database is actually dead.
- 3
Get Claude Code to write scripts for the mechanical fixes: split phone and mobile into separate fields, standardise casing, normalise company names.
- 4
Decide what counts as a duplicate before you dedupe. Same email is easy; the same company under three spellings needs fuzzy matching, and that is the fiddly bit. Matching on website domain gets you most of the way.
- 5
Enrich the survivors against live sources: scrape each company’s own website for current details rather than buying a stale data list.
- 6
Score every record (reachable, verified, classified) before loading it back, so you know what the cleaned database is actually worth.
I have a messy customer database export at [file.csv]: duplicate companies, phone and mobile mixed in one field, dead emails, inconsistent names. Write scripts to: 1) profile the data and show me how bad it is, 2) normalise names and split the phone fields, 3) fuzzy-match duplicates on company name and website domain, showing me candidate merges before merging anything, 4) flag rows that need email verification. Never delete: mark and merge, keeping the originals.
Copy it into Claude Code, fill the brackets, and it will plan the build with you before writing a line of code.
We would rather show you how than bill you. The whole ladder of free help, answers, guides and the weekly build-along, is on the do-it-yourself page.
Go deeper