Case Study salmonrun.ai

Half a million accounts. No one trusted the CRM.

How Salmon resolved entities, mapped parent-child hierarchies, and deduplicated a 500K-record CRM for a leading real estate technology platform.

Years of acquisitions and CRM migrations left a major real estate technology company with over half a million account records — and no one trusted any of them. The same client appeared as six different accounts: "Cushman & Wakefield (HQ)," "C&W - Brisbane," "Cushman Wakefield Commercial Ireland Ltd." No parent-child linkage. Stale firmographics. Reps Googled everything anyway.

514K CRM records
38% duplicate or fragmented entities
40+ fields enriched per account
3 wks data handoff to delivery

Same company. Six different accounts. No connection between them.

Pipeline reviews turned into arguments about which account was the "real" one.

Entity Fragmentation

Global clients appeared as dozens of disconnected records — one per office, one per legacy system, one per acquisition. No parent-child mapping. No way to see total relationship value across the same organization.

Stale Firmographics

Revenue from 2019. Pre-pandemic headcounts. Inconsistent industry classifications. Prior enrichment vendors matched around 40% of accounts — and got abandoned.

Duplicate Records

Same company, different name variations, different addresses — sometimes different industries. Marketing hit the same org three times. CRM dedup rules couldn't keep up with the variation.

They'd tried this before. D&B's hierarchy product needed clean DUNS numbers they didn't have. An internal team spent three months manually deduping 20K records before the project stalled. Native CRM dedup caught exact matches and missed everything else.


Entity resolution is a research problem, not a matching problem.

Traditional dedup tools pattern-match on the fields you already have. When those fields are wrong, matching fails. Salmon researches each account the way an analyst would — then uses that research to resolve entities, build hierarchies, and flag duplicates.

01
Profile the data and map the mess. Ingest all 514K records. Identify naming patterns, variation clusters, and field quality. Map the problem before solving it.
02
Research and resolve each entity. AI agents cross-reference company registries, SEC filings, LinkedIn, and news to determine what the entity actually is. "C&W - Brisbane" and "Cushman & Wakefield (HQ)" resolve to the same organization — not through string matching, but through research.
03
Build parent-child hierarchies. Salmon maps corporate structure — HQ, subsidiaries, regional offices, divisions. A single account tree replaces scattered records. Total relationship value becomes visible for the first time.
04
Enrich, score, and deduplicate. Fresh firmographics on every resolved entity. Duplicates flagged with merge recommendations and confidence scores. The output is a clean, hierarchical, CRM-ready dataset.

Three weeks. From chaotic CRM to connected accounts.

W1
Onboarding & data profiling. Ingest the full export. Profile field quality, map naming patterns, identify entity clusters.
W2
Entity resolution & enrichment. AI agents resolve entities at scale. Hierarchies built, firmographics refreshed, duplicates grouped with merge recommendations.
W3
QA & delivery. Automated QA flags low-confidence resolutions. Final dataset delivered with confidence scores, source attribution, and hierarchy mappings.
Quarterly refresh. Firmographics decay. Companies restructure. Salmon re-runs resolution and enrichment quarterly — catching new duplicates, updating hierarchies, refreshing stale fields before they compound.

01

Raw CRM record arrives

A typical account from the CRM: a company name that's an abbreviation of a well-known enterprise, a stale address, outdated revenue, and no connection to the five other records in the system that represent the same organization.

02

Entity researched

Salmon's agents search company registries, the corporate website, SEC filings, and LinkedIn to identify this as a subsidiary of a global real estate services firm. The abbreviated name resolves to a known legal entity.

03

Hierarchy mapped

This record is linked to its parent organization and five other accounts in the CRM that represent offices, divisions, or legacy name variations. A single account tree emerges from what looked like unrelated records.

04

Firmographics refreshed

Revenue, headcount, industry classification, website, HQ address, and key contacts are all updated to current values. Stale data from 2019 is replaced with verified 2026 figures. Confidence scores attached to every field.

05

Salmon output

A resolved, enriched, hierarchically-linked account record. Duplicates flagged for merge. Parent-child relationships mapped. Every field sourced and scored. The ops team can finally trust what's in the CRM.

From fragmented records to a connected account tree.

Raw CRM record
Account C&W - Brisbane
Industry Real Estate
Revenue $4.1B (2019)
Employees 53,000
Website
Parent
Hierarchy
Duplicates
Address 123 Eagle St, Brisbane
Status Active

From "which account is the real one?" to a single source of truth.

Three weeks. Pipeline reviews stopped being arguments about data and started being conversations about strategy.

Entity Resolution at Scale

514K accounts resolved into distinct, verified entities. Abbreviations, regional variations, and legacy names mapped to canonical organizations. What took three months manually for 20K records — done across the full CRM in weeks.

Parent-Child Hierarchies

Corporate structures mapped — HQ, subsidiaries, regional offices, divisions. For the first time, the team could see total relationship value across an entire organization instead of fragmented deals across disconnected accounts.

Deduplication & Fresh Data

Duplicate clusters flagged with merge recommendations. Every surviving account enriched with current firmographics. 40+ fields where most had fewer than 10 before.

The alternative: years of manual cleanup. Their internal team stalled after three months on 20K records. At 514K, you're looking at a project that never finishes — and by the time you're halfway through, the first records are already stale again.


CRM cleanup at scale.

If your CRM has grown through acquisitions, migrations, or years of inconsistent data entry — different industry, same data debt.

What Salmon delivers for account-level CRM data

  • Entity resolution — name variations, abbreviations, and legacy records mapped to canonical organizations
  • Parent-child hierarchy mapping — HQ, subsidiaries, divisions, regional offices
  • Deduplication with merge recommendations and confidence scores
  • Full firmographic enrichment — revenue, headcount, industry, website, contacts
  • Quarterly refresh to prevent data decay from compounding

Send us a sample. We'll show you what's hiding in your CRM.

Export a few thousand accounts and see what Salmon finds — duplicates, broken hierarchies, stale data. No commitment, no contract — just proof.

kevin@salmonrun.ai

salmonrun.ai