CRM Cleanup Day: The Quarterly Database Detox

Your review-request text just went out to "Mike Johnson" twice — because Mike Johnson is in your system three times under three different phone numbers, one of which belongs to a guy who moved to Florida in 2022. Your missed-call text-back fired to a disconnected number. Your equipment-age campaign texted a customer about a furnace you replaced last year because the new install date never got recorded. None of your systems are broken. Your data is broken — and broken data makes every automation quietly dumber.

This post is part of the Hub: the 8-system framework — because CRM (customer relationship management — your database of customers and job history) hygiene isn't one system, it's the foundation all eight systems stand on. Clean data in, reliable automations out.

Why your database rots (and why quarterly is the cadence)

Database rot is not a character flaw — it's physics. People move, change numbers, get married and change names, sell houses to strangers who become your customers under a new name. Techs create duplicate contacts on the fly because they can't find the existing one. The office enters "Mike" on Monday and "Michael" on Thursday. Every month of normal business operations adds a few percent of noise, and nothing removes it — unless you schedule the removal.

Quarterly is the right cadence for a simple reason: monthly is too frequent to feel worth it (you'll skip it), and annually lets the rot compound until the cleanup is a multi-day nightmare nobody finishes. One day per quarter — put it on the calendar like a dentist appointment — keeps every cleanup session to a manageable few hours because you're never more than 90 days behind.

The five rot categories

Before the checklist, know what you're hunting. Nearly every dirty database contains the same five contaminants:

• Duplicates. The same person (or household) entered two, three, sometimes five times. Different spellings, different phone numbers, different addresses. Duplicates are the #1 automation killer: they cause double texts, split job history across records, and make your "customer count" a fiction.
• Dead numbers. Disconnected lines, wrong numbers, landlines that can receive texts (they can't). Every campaign you send to a dead number costs you money and skews your response metrics.
• Zombie contacts. Records for people who are no longer your customer in any meaningful sense: moved away, sold the house, switched contractors, or never actually became a customer (tire-kicker quotes from 2019). They bloat every list and every report.
• Fragmented households. "Mike Johnson" and "Lisa Johnson" at the same address, each with half the job history. Marketing goes to one, the service history lives on the other, and nobody has the full picture.
• Missing essentials. Contacts with no phone number, no address, no email, or no record of what you actually did for them. A contact you can't reach and can't reference is a row in a spreadsheet, not a customer relationship.

The quarterly cleanup checklist

Block the day. One person (usually the dispatcher or office manager — whoever touches the CRM most) works through this checklist top to bottom. Most shops finish in 3–5 hours for a database of a few thousand contacts.

1. Run the duplicate hunt (60–90 minutes)

Sort your contact list by phone number, then by last name. Duplicates cluster: same phone with two names, same name with two phones, same address with two households. Most CRMs have a built-in duplicate finder — run it first, then eyeball the near-misses it flags.

The merge rules — write these down and follow them every time, because inconsistent merging is how you lose data:

• Surviving record = the one with the most job history. Never merge into the empty record. The record with invoices, notes, and equipment history wins.
• Surviving phone = the mobile that last received a reply. If one number got a text reply in the last year and the other didn't, keep the live one. Mark landlines as landlines so they never enter a text campaign.
• Combine, don't delete, the notes. Before merging, copy any unique notes (gate codes, dog warnings, equipment details) into the surviving record. The merge tool may or may not carry notes over — assume it won't.
• Households merge to one primary contact. Pick the person who books the work as primary; add the spouse as a secondary contact on the same record, not a separate one. One household, one record, full history.

Example: you find "Mike Johnson — 555-0142 — 412 Oak St" with 6 jobs, and "Michael Johnson — 555-8871 — 412 Oak" with 1 job and a gate-code note. Per the rules: the 6-job record survives, the gate code gets copied into it, the 555-8871 number gets kept as a secondary number (it might be the spouse's or an old line), and "Michael" becomes a note, not a record. One contact, complete history, no data lost.

2. Scrub the dead numbers (30–45 minutes)

Pull your last 90 days of text-campaign reports. Every number that bounced, failed delivery, or replied "wrong number" goes on the scrub list. For each one: try to find the correct number (check the invoice, the work order, the tech's notes), update it if you can, and if you can't, mark the number invalid — don't just leave it sitting there to fail again next quarter.

Also run a landline check if your texting platform offers one. Landlines in a text list are pure waste: the message "delivers" and nobody ever sees it. Most A2P texting tools can flag landlines in bulk — one click, and those contacts get marked voice-only.

3. Retire the zombies (45–60 minutes)

Zombies are a judgment call, so use clear rules instead of gut feel. Archive (never delete — more on that below) any contact that meets all three: no job in the last 3 years, no reply to any text or email in the last 2 years, and no active equipment record or maintenance plan. If a contact fails any one of those tests, they stay.

Two exceptions to the archive rule: past install customers stay even if they're quiet — their equipment ages on your schedule, and the replacement campaign needs them. And anyone who ever left a review stays — they're social proof with a phone number, worth a win-back text someday.

4. Fill the gaps (30 minutes)

Filter for contacts missing a phone number or missing an address. For active customers, the office fills these from recent invoices — it's usually a 20-minute data-entry sprint. For the rest, mark them "incomplete" so they get skipped by campaigns automatically instead of failing silently. A campaign that skips 40 bad records is fine; a campaign that texts 40 bad numbers is a bounce-rate problem.

5. Audit the tags and segments (30 minutes)

This is the step everyone skips and the one that saves the most future pain. Open your tag/segment list and look for: tags nobody uses anymore ("2024-spring-promo"), tags with overlapping meanings ("hot-lead" vs "hot lead" vs "HOT"), and contacts sitting in campaigns they finished months ago. Consolidate the tag sprawl down to the tags your automations actually reference, and move finished contacts out of active sequences. Tag sprawl is how a customer ends up getting the "welcome" text twice.

Never delete — archive

The one hard rule of cleanup day: archive, don't delete. Deleted contacts are gone forever, including their job history, their invoices, and the paper trail you might need for a warranty dispute or a tax question. Archived contacts disappear from campaigns, lists, and reports — which is everything you want — while their history stays intact. Every major CRM and FSM (field service management software — your dispatch and scheduling system) supports archiving or an equivalent "inactive" status. If yours doesn't, use a tag like "archived-2026-Q3" and exclude it from every segment. Deletion is irreversible; archiving is a filter.

The three hygiene habits that keep it clean

Cleanup day fixes the past. These three habits protect the future — and they're what make each quarterly session shorter than the last:

• The search-before-create rule. Nobody creates a new contact without searching for the existing one first — by phone number, not by name, because phone numbers don't have spelling variants. This single habit eliminates the majority of new duplicates. Make it an SOP (standard operating procedure — a written step-by-step for how a task gets done): search phone, then name, then address, then create.
• The end-of-job note. When a tech finishes a job, they report back three things: anything that changed (new number, new gate code, moved), the equipment serviced, and anything weird. The dispatcher updates the record that day. Five seconds of capture beats an hour of archaeology next quarter.
• The monthly 15-minute spot check. Once a month, the office manager pulls 20 random contacts and checks them: correct number? correct address? job history intact? If more than 2 of 20 are wrong, something in the entry process is broken — fix the process, not just the records. This is your early-warning system between quarterly cleanups.

What clean data unlocks

Here's the payoff framing, because cleanup day is the easiest system to skip: every automation you've built — the missed-call text-back, the review requests, the equipment-age alerts, the reactivation campaigns — runs on this database. A 10% duplicate rate means 10% of your texts are doubles or dead ends. A list full of zombies means your "2,000 contacts" is really 1,200 reachable humans, and every metric you read is lying to you. Clean data doesn't just prevent embarrassment (the double text to Mike Johnson); it makes every other system measurably more effective, because now you're measuring reality.

And the compliance angle matters here too: texting dead or wrong numbers is how you text strangers, and texting strangers without consent is how TCPA trouble starts. A clean list is a compliant list. Keep your A2P 10DLC registration (the carrier approval for business texting — typically a 1–7 day approval and about $15–$20 one-time) current, include opt-out language in first texts, honor STOP requests immediately, and respect the quiet hours (no marketing texts before 8 AM or after 9 PM in the recipient's local time). This is general information, not legal advice.

Related guides in this series

This post is part of the Hub: the 8-system framework — the full picture of how the systems fit together lives there. For the automations this cleanup protects, read the missed-call text-back system and the equipment-age replacement pipeline next. And if you haven't documented your entry process yet, the SOP-library guide shows you how to write the search-before-create rule so it actually gets followed.

The done-for-you version

This post is part of the Leveraged Owner 8-system framework — the done-for-you Starter Pack with the full step-by-step SOPs, every script, the worksheets, and screen-by-screen setup instructions for all 8 systems.

Stats sourced as labeled: no third-party statistics used in this post.

Keep building your systems

Related guides from the Leveraged Owner blog: