TL;DR
- The CRM admin usually owns deduplication, while other teams decide the unclear cases.
- Duplicate searches work best set up per object, on two to four strong fields.
- Merge rules decide the surviving record and which value wins in each field.
- Automation suits searches that have been tested on hand merges and stay stable.
- Dedupely gives admins searches, merge rules, safety settings, automation, review queues and merge records.
In many companies, one person ends up running deduplication. It's the CRM admin, or someone in RevOps or marketing ops who knows the CRM best. Sales and marketing feel the duplicates, but the admin is the one who finds them, merges them and answers for every merge.
This guide is for that person. It assumes you run Dedupely day to day, or are about to, and want a setup that holds up without taking over your week.
You'll see how to divide the work, set up searches for each object, decide what a merged record keeps and keep risky matches out. You'll also see when to automate, how to hand cases to colleagues and which records to keep.
Working alone? Our post on deduplication when you're the only admin covers how to fit it into one person's time.
What the admin owns, and what others decide
Deduplication goes smoother with one owner. The admin builds the searches, sets the rules, runs or schedules the merges and keeps the record of what changed.
Other teams own the answers. Sales knows which two accounts are one customer, and finance knows which companies must stay separate. What counts as a duplicate is case by case, and they're the ones who know the case.
Many teams also keep reps out of merging altogether. Reps send questions to the admin instead of merging what they find on their own.
Set up duplicate searches for each object
Duplicate Searches are where you decide what counts as a duplicate. Every synced object comes with preset searches the first time you log in, and an account can hold up to 50.
- One search per pattern: keep "same email" and "same name and company" as separate searches, so each can have its own review and automation.
- Two to four strong fields: one field rarely proves two records are the same, and several that agree make a safer match.
- The right match option per field: Exact for email, Similar for typos and Fuzzy for names that sound alike. Domain Root suits company domains, and None leaves out fields you don't trust.
- Clones for testing: clone a search to try new criteria without touching the one you rely on.
Every search shows its number of matches and the records in them. A count far above what you expected usually means the criteria are too loose. For Contacts, Companies and Leads, Dedupely Intelligence also finds complex duplicate patterns with AI-powered matching, without match settings.
Decide what each merged record keeps
Merge Rules decide the outcome. Without them, Dedupely uses your CRM's own merge logic, which is fine for a first set of exact matches. Once you want a specific outcome, set it once and every merge follows it.
| Rule | Options | What to think about |
|---|---|---|
| Primary Record | Oldest, Newest, Last Updated, Most Field Data or Custom | Which record your integrations and reports point to |
| Field-level rules | Primary, Only Primary, Append (multi-value fields) or Custom | Owner, lifecycle stage, email and phone usually need their own rule |
| Win and Lose Rules | Custom conditions; Lose Rules run first, then Win Rules from top to bottom | For example, an inactive owner should never win |
Blank fields never overwrite existing data, and rules can't be set on fields that are read-only in your CRM. Preview a few merges after every rule change, since the in-line preview shows the winning values in bold.
Let safety settings hold back risky matches
Safety Settings are on by default for every object. They hold back matches with more than 20 records. They also hold back matches built on example data, like test@test.com, John Doe or placeholder phone numbers.
Held-back matches go to the Excluded tab, where you can merge them by hand, reject them or send them to Manual Review. Each object has its own thresholds, and a change takes effect on the next scan. In HubSpot, you also choose what happens when a merge would pass HubSpot's association or merge-count limits.
Check the Excluded tab now and then. A large excluded set is often a shared inbox or a placeholder value worth fixing at the source.
Automate in two steps
Dedupely has two kinds of automation, and admins tend to use them in order.
- Auto Match checks a search on a schedule and sends an email when its matches reach the number you set. It never merges, so it's a safe first step for any search.
- Auto Merge merges new duplicates from one search automatically on a schedule, using its fields, filters and Merge Rules. In the free trial, its merges count toward the 20 free merges.
A search is ready for Auto Merge after a few hand merges and a bulk page. False matches should be rejected, and results should stay stable.
Auto Merge never merges excluded, rejected or Manual Review matches. You can pause it, disable it or cancel a running merge from its dashboard.
Hand unclear sets to the right person
Manual Review is a queue you fill by hand. You send any duplicate set there, from any search, and it stays until someone merges or removes it. Matches can't be assigned to one person, so agree on who checks the queue and how often.
Colleagues join as Users. They can review, merge, change Merge Rules and turn on Auto Merge, while only Admins invite people, change roles and manage company settings. Every plan includes unlimited users, with no per-user fees, so the person who knows the accounts can decide the close calls without a separate seat.
Keep the Admin role for the one or two people who manage users, security and billing. Our guide on working as a team in Dedupely covers how other teams split the work.
Keep the records you'll be asked about
Sooner or later, someone asks why two records merged. These are the places to look.
| Record | What it holds |
|---|---|
| Merge History | Every merge, with the original records, the winning values and a link to the merged record in your CRM. Exportable as a CSV that keeps the original data and record IDs |
| Audit Log | Merges, scans, search changes, invites, role changes and logins, with the user and time. Admin-only, kept indefinitely |
| Duplicate Report | An email on your duplicates, weekly by default, monthly or off, set per person |
| New Duplicate Alerts | An email when an Auto Match search reaches its threshold, on by default for each user, at most one per day |
Most merges can be undone for 30 days from Merge History, but treat every merge as permanent. If a merge goes wrong, undo it there or restore values from the Merge History CSV. Our playbook on what to do if you merged the wrong records walks through it.
Running more than one CRM
If you connect several CRMs, or several accounts of the same CRM, each one is independent. Each has its own searches, Merge Rules, Safety Settings, Merge History, users, Security Policy and Audit Log. Dedupely never matches records across two CRMs.
The record limit and billing are shared across all of them. No email warns you before the limit, and if you go over it, you can keep working with up to 10% extra records for 10 days. After that, the account becomes read-only until you upgrade or get back under the limit. Searches, scans and Auto Match keep working, so keep an eye on usage before a large import.
How to get your Dedupely setup ready as an admin
Start with the preset searches on your busiest object, then merge a few sets by hand and set Merge Rules from what you see. Once a search stays stable, put it on Auto Match, then Auto Merge, and bring in colleagues for the sets in Manual Review. Our page for CRM admins shows the rest of what Dedupely does for your role.
A setup that runs without you watching
Running deduplication well comes down to a few decisions made once. Decide what counts as a duplicate per object, what a merged record keeps and which matches are too risky.
With those in place, the daily work gets small. Tested searches merge automatically on a schedule, colleagues decide the close calls, and every merge leaves a record you can point to.
Frequently asked questions
Who should be responsible for CRM deduplication?
Usually the CRM admin, or someone in RevOps or marketing ops who knows the CRM well. One owner builds the searches, sets merge rules and runs the merges. Other teams decide unclear cases, since they know which records are one customer and which must stay separate.
How many fields should a duplicate search match on?
Two to four strong fields is a good range. One field rarely proves two records are the same, while several fields that agree make a safer match. Start with the field that identifies each record, like email or domain, and add fields that confirm it, like name or company.
When is it safe to automate merging duplicates?
When the search has been tested. Merge a few sets by hand and a page or two in bulk, reject the false matches and check that results stay stable. Start with duplicates you'd merge without looking, like exact email matches, and keep less certain patterns in review.
How do I find out who merged two records?
Check your merge log first, which should show the original records, the values that won and the merged record. An audit log shows who ran the merge and when. Keep both, since merges are usually permanent in the CRM and questions often come months later.
Should sales reps merge duplicates themselves?
Many teams prefer they don't. Reps merging what they find leads to inconsistent outcomes and few records of what changed. In a common setup, reps send questions to the admin, who merges with agreed rules and asks the owner about unclear sets.
Or sign up with Google
HubSpot Marketplace 
