Key takeaways
- Most duplicates are made at booking, when the person answering can't find the customer in a few seconds and creates a new record.
- Search by the caller's number, then house number and street, then last name, and ask one direct question before creating anyone.
- Records go stale even when nobody slips: renters move about four times as often as owners[1], and around 35 million phone numbers are disconnected and reassigned each year[2].
- Merge on two strong signals, never on a name alone, and do it weekly in small batches: pick the survivor, move every piece of history, archive the loser.
Where the second record comes from
The duplicate is almost never made by the software. It is made at 8:10 on a Monday, when a CSR has a caller talking, two more lines ringing, and a search box that returns nothing for “Bill Okafor”. The account is under “William Okafor”, or under his wife's name, or under the landlord's company with Bill as the tenant. Creating a new customer takes fifteen seconds. Finding the old one would have taken forty-five. So the shop gets a second Okafor, and the water heater installed in 2021 now sits on a record nobody will open.
Every duplicate splits history. The tech arrives without the note about the crawl space hatch. The warranty lookup misses the original install date. The membership renewal goes to one record while the work goes to the other, and the customer gets a reminder for a plan they already paid for. Reports drift too: software reports on customer count, repeat rate and revenue per customer all overstate new customers and understate loyalty when one household sits on three records.
Some duplicates have nothing to do with a rushed booking. People move, numbers change hands, households split phones. In the Census Bureau's 2023 survey, 16.2% of people living in rented homes had moved in the past year, against 4.1% of people in owned homes[1]. With the homeownership rate at 65.5%[3], roughly a third of the homes you serve are rented, and those addresses turn over their callers fast.
- Owned homes
- Rented homes
Source: U.S. Census Bureau, 2023[1]
Show the data
| Survey year | Owned homes | Rented homes |
|---|---|---|
| 2015 | 5.1% | 24% |
| 2017 | 5.5% | 21.7% |
| 2019 | 4.9% | 19.7% |
| 2021 | 4.9% | 15.9% |
| 2023 | 4.1% | 16.2% |
Phone numbers drift as well. The FCC estimated that about 35 million numbers are disconnected and made available for reassignment each year[2]. A number on file can belong to a stranger within months, which is how a new caller's number pulls up somebody else's record and the CSR, quite reasonably, makes a fresh one.
Six patterns behind most duplicates
| Pattern | What it looks like | Booking habit that prevents it |
|---|---|---|
| Name variant | “Bill” and “William”, “Tran, Kim” and “Kim Tran”, a typo in the surname | Search by phone and street before searching by name |
| Second phone | The customer calls from a mobile; the record only has the old home landline | Store several numbers per customer, each one labelled |
| Household member | A spouse or adult child books under their own name | Add them as a contact on the account, not as a new customer |
| Tenant and owner | The tenant calls, the landlord pays, and both become customers at one address | One billing customer, the property as a service location, the tenant as a site contact |
| Address format | “12 Oak Ave” and “12 Oak Avenue Apt 2” as two places | USPS abbreviations, and the unit number in its own field |
| New owner, old address | The house sold and the new owner calls | A new customer, linked to the existing service location history |
The second-phone pattern is fading, but it hasn't gone. In CDC's late-2023 survey, 75.2% of adults were wireless-only, with no landline at home, yet among adults 65 and over the share was 54.6%[4]. Older customers are the ones most likely to have two numbers on file, and the most likely to call from whichever phone is nearest.
Addresses cause more quiet duplicates than names do. The Postal Service's list of common street suffix forms gives seven ways people write Avenue, from AV to AVNUE, all standardized to AVE[5]. Depending on how your software searches, “Oak Avenue” may never find “Oak Ave”. Pick the USPS standard abbreviation and use it every time; a standardized address includes every required element and is built to match postal files exactly[6], which also helps the tech's map app find the right house.
Search before you create
Search the number the caller is using
If caller ID shows a number, search it before asking anything. A match is a strong lead, not proof: confirm the name and street before reading any details back.
No match? Ask for the service address
Search by house number and street name. People give their address the same way every time and their name several ways, so the street finds more returning customers.
Then try the last name alone
Drop the first name and scan the results for the same street or a nearby one.
Ask one direct question
“Have we been out to this address before, maybe under another name?” Tenants, spouses and new owners answer that one honestly.
Only then create the customer
Fill the standard fields the standard way: mobile in the mobile field, unit number in the unit field, USPS abbreviation for the street suffix.
This adds perhaps twenty seconds to a booking. That is cheap next to a tech who doesn't know about the dog, or a renewal letter sent to the household that moved out. For what a returning caller's record should show the moment the phone rings, see recognizing repeat customers.
Example script: “I'm not finding that number in our system. What's the address where you need us?” (search) “I can see we were at 412 Linden Avenue in 2023. Was that you, or the previous owner?” If it was the previous owner: “No problem, I'll set you up as a new customer and keep the house history attached, so the tech can see what was done.”
Match rules: which signals mean the same customer
Before merging anything, agree what counts as a match. Some signals are strong on their own; others only mean something in combination. Write the rule down so the office manager and the newest CSR make the same call.
| Signal | Strength | When it misleads |
|---|---|---|
| Same service address and unit | Strong | A new owner or new tenant at the same address |
| Same mobile number | Strong | A reassigned number, or one phone shared by a family |
| Same email address | Strong | A shared household or business inbox |
| Same last name and street | Medium | Relatives on one street; large apartment buildings |
| Same landline only | Medium | The number moved to a new household after a move |
| Same name only | Weak | Common names; a father and son with the same name |
The last column matters as much as the strength. Merging two different customers is worse than leaving a duplicate: one household's invoices, notes and access details end up in front of another. When in doubt, link the two records with a note instead of merging them, and confirm with the customer on the next call.
A weekly merge routine
- System
- Your team
01System
Pull the likely pairs
Run a duplicate report, or sort customers by phone and by street and scan.
02Your team
Check two signals
Same address and same mobile, or another strong pair. A name alone never counts.
03Your team
Pick the survivor
Keep the record with the most job history and any active membership.
04Your team
Move the history
Jobs, invoices, notes, photos, equipment, memberships and open estimates.
05Your team
Fix contact fields
Keep every live number, labelled; remove numbers the customer says are gone.
06Your team
Archive and log it
Archive the merged record and note who merged it, when and why.
Example routine of 20 to 30 minutes a week. If your software has a merge tool, try it on two dummy records first and check what it does with notes and balances.
Do it weekly, not yearly. A pile of four hundred duplicates is a project nobody finishes; fifteen a week fits into a quiet half hour. Give the routine to one person, usually the office manager or the CSR who books the most, so the judgment calls stay consistent.
- Open balances. The balance follows the survivor, and no payment is counted twice.
- Memberships and service agreements. One household, one plan, one renewal date.
- Equipment records. Water heater install dates and model numbers drive warranty and replacement conversations, so they must survive.
- Notes with a shelf life. An access note from three years ago shouldn't come across just because it sat on the losing record.
- Consent and opt-outs. If either record asked not to be texted or called, the merged record keeps that request.
Tenants, landlords and property managers
Rentals produce the messiest records because three parties touch one address: the tenant who calls, the owner who pays and sometimes a manager who approves. People in rented homes made up 63.8% of everyone who moved in 2023[1], so the person who called about the kitchen sink last year is often not the person calling now, while their mobile number stays on file.
Structure beats cleanup here. Make the owner or management company the billing customer, the property a service location under it, and each tenant a site contact with a start date. When a tenant moves out, end the contact; don't delete the location or its history. The commercial contractor example shows the same discipline at portfolio scale, starting from an export of every agreement customer with its sites before a single call is answered.
A house sale is the one case where a new record is right. The new owner is a new customer; the house is not a new house. Link the new customer to the existing service location so the pipe material, the water heater age and the sewer camera footage stay with the address.
Calls answered outside the office
Night and weekend calls are a steady source of duplicates, because whoever answers can rarely see your customer list, and the morning person keys every message in as a new customer. The fix is a rule, not a tool: the morning entry starts with a search, never a create.
Whether nights go to an answering service or an SI Employee, ask for the address and callback number to be read back to the caller, so the morning search runs on clean data. With Ava, the summary arrives by email with the recording and transcript in the account; creating jobs straight in field-service software is still planned, so matching is a person's job for now. Call data can also go to other tools by webhook (how integrations work).
If you are still moving off paper or spreadsheets, set the search-first rule during the paper-to-digital switch, before habits form. And because notes are what a careless merge loses first, keep them short, dated and categorized, as described in what to record in customer notes.
Monthly checks that keep it clean
- Duplicate count. Note how many pairs the weekly routine found. A rising count means a booking habit has slipped somewhere.
- Records created without an address or a phone number. Each one is a future duplicate.
- New-customer share. If “new” customers jump without a marketing push, look for returning customers entered again.
- Portal accounts. Two portal logins for one household usually mean two records behind them; see customer portals.
- Old data. The FTC's guidance for businesses is to keep only the personal information you need[7]. Archive customers with no contact for years under a written rule, rather than letting them sit as merge candidates forever.
Example: an office with 4,200 customer records runs its first duplicate report and finds 310 likely pairs, about 7% of the list. Clearing 15 a week, it works through the backlog in about five months. After that the weekly report settles at two or three new pairs, nearly all of them night calls that were entered as new customers, which tells the office manager exactly which habit to fix.
Frequently asked questions
Should we merge or delete duplicate customer records?
Merge. Deleting a duplicate throws away whatever history it held: jobs, invoices, photos, warranty dates. Move everything onto the surviving record, then archive the empty one so there is a trail of what was merged and when.
How often should a plumbing office clean up duplicates?
Weekly, in small batches. A short routine catches pairs while the people who booked them still remember the call. A yearly cleanup turns into a project that rarely gets finished.
Can two customer records share a phone number?
Sometimes, correctly: a landlord and a tenant's emergency contact, or a family phone. Label each number and confirm before merging, because numbers also change hands; the FCC estimated about 35 million are disconnected and made available for reassignment each year[2].
What if we merge two different customers by mistake?
Split them as soon as you notice: move each job, invoice and note back to the right customer and check that access details didn't cross over. Some software can't undo a merge, which is why the two-signal rule exists.
Does standardizing addresses really reduce duplicates?
Yes, because many duplicates are the same address typed two ways. USPS lists seven common forms of Avenue that all standardize to AVE[5]. One format for every booking makes the search find the existing customer.
References
U.S. Census Bureau (2023). Table A-4. Annual Geographic Mobility by Tenure: 1988-2023 (CPS ASEC) (opens in a new tab) ↑
Federal Communications Commission (2018). Advanced Methods to Target and Eliminate Unlawful Robocalls, Second Report and Order (FCC 18-177) (opens in a new tab) ↑
U.S. Census Bureau (2026). Quarterly Residential Vacancies and Homeownership, Third Quarter 2026 (opens in a new tab) ↑
National Center for Health Statistics (CDC) (2024). Wireless Substitution: Early Release of Estimates from the National Health Interview Survey, July-December 2023 (opens in a new tab) ↑
U.S. Postal Service (2024). Publication 28, Postal Addressing Standards: C1 Street Suffix Abbreviations (opens in a new tab) ↑
U.S. Postal Service (2024). Publication 28, Postal Addressing Standards: 211 Standardized Delivery Address Line and Last Line (opens in a new tab) ↑
Federal Trade Commission (2026). Protecting Personal Information: A Guide for Business (opens in a new tab) ↑