Can we just use the phone number as the unique ID for a customer?
It is the best single identifier available in home services, and still not enough on its own. Numbers are shared inside households and small businesses, get reassigned by carriers, and one person may hand you three different ones over five years. Use a normalized phone as the primary match key, require a second signal before merging, and never let a phone match alone silently overwrite an existing record.
Why phone wins anyway
Email is optional and frequently mistyped. Names are ambiguous and change. Addresses are recorded inconsistently. The phone number is the one field that is almost always present, almost always correct, and shared across your website, your phone system, your CRM and your field service platform.
It is also the field that ties marketing data to operational data, because call tracking and booking records both carry it. Most practical identity matching in this industry leans on phone first and uses everything else to confirm or deny.
Four ways it fails
- Sharing. A landline belongs to a household. A main line belongs to a business. Two genuinely different customers, one number.
- Reassignment. Carriers recycle disconnected mobile numbers. A number in your database from years ago may now belong to a stranger — which is a marketing accuracy problem and, for outbound texting and calling, a compliance one.
- Multiplicity. The same person gives you a cell on the first job, a work number on the second, and a spouse’s number on the third.
- Formatting. Extensions, country codes, punctuation and leading ones. Two records with the same number do not match if you compare raw strings.
Normalize, then require corroboration
Store phone numbers in one canonical digit-only or E.164 form, and keep the raw entry alongside it for reference. Match on the canonical form.
Then set the merge rule higher than the match rule. A phone match plus a matching service address, or a phone match plus a matching last name, is strong. A phone match alone should surface a suggestion for a human, not trigger an automatic merge. The cost asymmetry justifies the caution: a missed merge leaves you with a duplicate you can fix later, while a wrong merge fuses two families’ service histories into one record you may never untangle.
Design for many numbers per customer
The cleanest structure is not one phone field on the customer. It is a list of phone numbers attached to the customer, each with a type, a source, a first-seen date and an active flag.
That structure makes several everyday problems tractable at once: caller ID lookups match on any known number, outbound campaigns can prefer the most recently confirmed one, and disconnected numbers can be retired without deleting history. It also lets a tracked marketing number be recognized as a marketing number rather than mistaken for the customer’s own — a small detail that otherwise corrupts source reporting inside a marketing intelligence rollup.
Topics: identity · phone matching · customer key · data quality
Have a version of this question about your own business?
The useful answer usually depends on which systems you run and how they're connected. That's a conversation, not a blog post.