Section 4 · Nutshell — Turn 30,000 Records Into an Asset

4.4Clean duplicate and messy records

Built in the POC

What DMA asked for · Decision Book p.14

Find duplicate people and companies, and incomplete records.

Why it matters to DMA: Bad data undermines every score and outreach decision built on top of it.

In our words, from DMA's Product Decision Book v3, page 14.

Every review run compares accounts by name, domain and shared contacts and flags likely duplicates with a match score. It also flags messy records: deals stuck with no activity, deals with no next step, records owned by someone who has left, and bad email addresses.
It recommends; it doesn't merge (L1, as DMA's book sets it). A likely duplicate is capped at tier C so the same company is never contacted twice, and "duplicate of another account" appears in its reasons.
The flags are kept with the rule that raised them, so the team can clean up in Nutshell; a merge is always a person's decision.

It plays on its own while it's on screen; hover or use the controls to pause or step through.

Step 1 of 4

Compared

"Lakefield Roofing Co (old)" and "Lakefield Roofing Co" share a domain and a contact.

The same six stages on every page. Nothing reaches Nutshell except through the write gate.

1Nutshell records
Source
Accounts, contacts and deals.
2Sync
Copies Nutshell changes into the bridge database
Brings the records into the mirror.
3Bridge database
A copy of the CRM data, plus what the AI works out
Duplicate candidates and hygiene flags, with the rule that raised them.
4AI
Claude models, only through DMA's own gateway
Not needed: matching and checks are rules.
5Write gate
The only way back into Nutshell: checked, approved when needed, sent once
Nothing is merged or changed in Nutshell automatically.
6Nutshell
The system of record
The team cleans up where it works.

Drawn from the POC's database catalogue: structure only, no data.

PK primary keyFK reference the database enforces (solid line)ID reference kept by id (dashed line)

Table	What it holds	Columns
ai.duplicate_candidates	Pairs of records that look like the same company, with the match score and what matched.	10
ai.hygiene_flags	Messy-record flags, each with its rule and severity.	10

Synthetic demo data: every name, business and number is made up.

1 / 2

What you see

1."Duplicate of another account"

Why? on Lakefield Roofing Co (old): a decent score, capped at tier C because it duplicates another account.

localhost:13301/accounts

What happens in the system

Bridge databaseai.duplicate_candidates
Matched on name and domain with the active Lakefield Roofing Co record.
Rule engineHOLDduplicate gate
Score 60 would be tier B; the gate caps it at C.

The prototype's latest review over its synthetic accounts.

Record	Looks like	Matched on	Match
Wilowhill Pediatric Dentistry	Willowhill Pediatric Dentistry	Name, domain, contact	100%
Southside Home Goods	Southside Home Goods Inc	Name, domain, contact	100%
southside home goods	Southside Home Goods Inc	Name, domain	85%

localhost:13301/accounts

Why? on a likely duplicate, capped at tier C.

Dashboard · Accounts, a likely duplicate, why the score is capped. Captured from the running POC on September 25, 2026; synthetic data.

Opens the exact screen in the running POC, in a new tab.

Open	Where	What to try
Accounts	DMA AI dashboard	Duplicates show tier C with the reason.

Built in the POC

Built in the POC. This capability runs in the POC today, on synthetic data, end to end.