What does cross-tool identity resolution mean in an ordinary business workflow?
It means someone, or some rule, decided that four records are one customer, and that decision is stored with its basis. Without it, a customer's order history, invoices and complaints sit in separate piles that cannot be read as one story.
AI agents that work across a company's tools run into this constantly: an agent asked to summarize a customer's open issues has to know which help desk organization, which CRM account and which ERP customer belong together. Companies that license workflow data to AI developers, a market Mercor also describes on its data page (checked October 18, 2026), are offering records where those joins either exist or do not.
Illustrative: a 310-person commercial HVAC distributor has served a regional hospital group for twelve years. Over that time the customer has been recorded four ways. Every identifier below is fictional.
| System | Identifier | Created | Strongest field that could link it |
|---|---|---|---|
| ERP (orders, invoices) | Customer 40417 | 2014 | Tax ID on the account |
| CRM (contracts, renewals) | Account record with "ERP no. 40417" field | 2019 | Tax ID, plus the ERP number typed in by sales ops |
| Legacy billing (retired 2021) | Account BL-9921 | 2016 | Billing address and main phone |
| Help desk (service requests) | Organization keyed on the email domain | 2020 | Email domain only |
The ERP and CRM records share a tax ID and an explicit cross-reference, so the link is close to certain. The legacy billing account agrees on address and phone, which is strong but not conclusive: the hospital group moved its accounts payable office in 2018, so older invoices carry a different address. The help desk record links only by email domain, and the hospital group's domain is shared by three affiliated clinics that buy separately.
That last row is where identity resolution earns its name. A reviewer who merges it blindly attributes the clinics' service complaints to the hospital group's contract.
Exact keys, scored evidence and a held link
Two methods appear in most companies' records, often side by side. Exact matching joins records only when a key is identical. Salesforce's documentation, for example, describes its exact method as case-insensitive and treats non-identical strings as non-matches. Scored matching weighs several partial agreements. The Fellegi-Sunter model, explained in the documentation for the open-source Splink library maintained by the UK Ministry of Justice's analytical services, scores a pair of records by how likely each observed agreement is among true matches compared with non-matches, and adds those per-field weights into one score.
Applied to the illustrative customer:
| Link | Evidence | Method | Decision |
|---|---|---|---|
| ERP to CRM | Same tax ID; CRM holds the ERP number | Exact key | Linked |
| ERP to legacy billing | Address and phone agree; name differs by abbreviation | Scored | Linked, with the 2018 address change noted |
| ERP to help desk | Email domain only | Scored, weak | Held for review |
| Help desk to clinics | Same domain, different purchase orders | Scored, weak | Held for review |
The decision column matters as much as the links. A company that recorded "held for review" and later resolved it has better records than one that silently merged everything. Salesforce also notes that its fuzzy methods support only Latin characters and recommends exact matching for international data, a reminder that the method a company used shapes which links exist at all.
The fields that record a link and its basis
A link is only reusable if the reason for it travels with it. In the example, these fields carry that reason.
| Field | Where it sits in the example | Why the example depends on it |
|---|---|---|
| Native ID in each system | ERP 40417, BL-9921, CRM account, help desk organization | The four things being joined |
| Cross-reference field | "ERP no. 40417" on the CRM account | An explicit, human-entered link |
| Shared strong key | Tax ID on ERP and CRM | Makes the exact match possible |
| Matching fields and their values | Address, phone and name on billing and ERP | The evidence behind the scored link |
| Address history with dates | 2018 accounts payable move | Explains why older records disagree |
| Link decision and status | Linked or held for review | Separates proven links from candidates |
| Who decided, and when | Merge log or crosswalk change history | Lets a reviewer trust or revisit each link |
| Retirement record for the old system | Legacy billing closed in 2021 | Shows BL-9921 will never gain new activity |
A spreadsheet crosswalk with only two columns, old ID and new ID, carries the first field and loses the rest. It records that a link was made but not why, which leaves no way to tell the tax ID match from the email domain guess.
What the example shows, and what it cannot
- Four systems is a small case. Companies that have been through acquisitions can hold the same customer under many more identifiers, and our guide to acquisition identity mapping covers that wider problem.
- A held link is not a wrong link. The help desk organization may well belong to the hospital group. The record shows only that nobody confirmed it.
- Merges can be undone, or not. Some tools overwrite the losing record when two accounts merge, and the evidence for the merge goes with it. Whether a given company's history survived is checked in review.
- The messy cases are the valuable ones. Clean exact matches are easy to invent; the address change and the shared clinic domain are the parts that are hard to fake, one reason synthetic data rarely replaces real data.
Who owns these records, and what stays unknown
Revenue operations or sales operations usually owns the CRM and any crosswalk, finance owns the ERP and billing records, and support owns the help desk. A data or IT team may maintain merge logs. Authority to discuss licensing usually rests with the CEO, CFO or owner.
The introduced company needs 50+ full-time employees at peak (contractors excluded), a documented operating history, rights to license the records, and an authorized contact. Companies in the U.S., Canada and the EU qualify, as do companies in the UK, Australia, New Zealand, India, Singapore, South Africa, Latin America, the UAE and Saudi Arabia when at least 70% of the data is in English. The company qualification checklist covers the basics.
What stays unknown until SourceX reviews it: SourceX's current data needs for these records, how many links hold up under review, the outcome of the rights review, and the offer terms and any reward. Industry interest in identity resolution does not mean every customer database qualifies. Discovery stays at metadata level; never ask for or handle raw customer records. For related record families, see CRM consolidation after an acquisition, sales and CRM workflows, tool-use data and the full list of data types.
Go deeper on cross-tool identity resolution
- Questions to ask a client about cross-tool identity resolution: a short discovery script for checking whether a company's linked customer records could qualify.
- Record gaps that make cross-tool identity resolution hard to assess: a guide to the bare crosswalks, overwritten merges and lost address history that weaken these records.
- Cross-tool identity resolution review packet: a list of the metadata a company prepares before SourceX reviews its crosswalks and merge logs.
Next step
Refer a company. No cap on rewards. Default reward on a company's first qualifying transaction: the greater of $50,000 plus 5% and 15% of Actual Supplier Payout, with no cap; further qualifying transactions first agreed within 12 months of the first Closing pay 15%. Eligibility and payment conditions apply, including the USD 100,000 Minimum Transaction, and a signed Schedule A or addendum can differ. A lead, meeting or signature alone does not trigger payment.
If a client keeps cross-references or merge logs that tie its customer records together across systems, register as a partner, confirm who qualifies, and read the program terms before introducing the decision-maker.