Seven data quality numbers every Salesforce admin should know
The short answer: The seven Salesforce data quality metrics worth tracking are duplicate rate, key-field completeness, stale-record share, orphaned records, records owned by inactive users, completeness of newly created records, and picklist sprawl. Standard reports can produce all seven. Each one is a share of a total, so you can repeat it every quarter and see which way it is moving.
Somebody asks whether the data is good. You say “mostly”, because the only evidence you have is the complaints that reached your inbox. A complaint count measures how loud people are, not how clean the data is.
A number you can repeat does more. It doesn’t need to be precise. It needs to be the same measurement every time, so that “7%, down from 9%” means something. These seven are the ones I would set up first. Every reading below is a rule of thumb, which means a starting point for a conversation and not a standard. No industry body publishes the right figures, and I’m not going to invent some.
1. What is your duplicate rate?
Duplicate rate is the share of records that have a twin. It tells you how much history, pipeline and ownership is split across records that should be one.
To get it, build a summary report on Contacts grouped by Email, sorted by record count, and see how many records sit in groups of two or more. For accounts, a tabular report sorted by name and a hand count of twins in a few stretches of rows gives you an estimate. If you use duplicate rules with the Report action, Salesforce Help describes a custom report type with Duplicate Record Sets as the primary object, which lists the sets your rules found.
The caveat is that those sets only contain what your matching rules detect. Records the rules never flagged won’t be there, and sets can linger after records stop matching. Rule of thumb: under 2% of accounts with a twin is comfortable, and over 8% is a project.
2. How complete are your key fields?
Key-field completeness is the share of records where the fields you rely on actually have a value. It tells you whether routing, segmentation and reporting have anything to work with.
Choose five fields per object, say Title, Email, Phone, Mailing Country and Lead Source on Contacts. Build a report, note the total, add a filter on one field with the operator “equals” and the value left empty, and note that count. Repeat for each. The share filled, averaged, is your number.
The hard part is the choice of fields, not the report. Pick the ones a process reads, not the ones that look important. Rule of thumb: 90% or more filled is healthy, and under 70% means people are working around the gaps.
3. How many records are stale?
Stale-record share is the portion of records nobody has touched in a long while. It tells you how much of your database is memory and how much is working material.
Salesforce’s LastActivityDate is a system field that holds the date of the most recent logged task, event, call or email, and standard objects such as Leads, Contacts, Opportunities and Accounts include it in reports. Filter on it being older than your cutoff, say 90 days for leads, and compare with the total. For accounts with no logged activity at all, Salesforce Help describes an Account report with a cross filter that returns accounts without Activities.
Two limits. LastActivityDate only reflects logged activity, so teams that work records through other channels look quieter than they are. And for shared activities, the date appears only on the primary contact. Cross filters are available in Enterprise, Performance, Unlimited and Developer editions. Rule of thumb: for open leads, under 10% untouched for a quarter is good, and over 40% means the list has become a museum.
4. How many records are orphaned?
Orphaned records have lost the parent they should belong to. A contact with no account drops out of account views, roll-ups and most reports.
For contacts, report on Contacts and filter on Account Name equals blank. For the other direction, accounts with no contacts, use a cross filter with Accounts as the primary object, “without”, and Contacts as the secondary object. Salesforce Help documents WITH and WITHOUT cross filters and gives “contacts without activities” as its own example of the pattern. An account with no contacts isn’t always a mistake (a prospect may be exactly that), so read the list before you act.
Rule of thumb: under 2% of contacts with no account is fine, and over 10% means something upstream is creating them.
5. How many records belong to inactive users?
This number shows how much of your data is sitting with someone who can’t see it, answer it or be routed to. Salesforce Help describes a custom report type on Leads with the Lead Owner’s Active field added, then a filter of “Lead Owner: Active equals False”.
The same idea works for accounts, opportunities and cases. For an estimate without setup, list the users on the Inactive Users view in Setup, then group your main objects by owner and look for those names. There is a whole post on this one, including what to do about it: records owned by inactive users. Rule of thumb: zero is the target, because every record here is one someone could have reassigned in a minute at the time.
6. How complete are new records at the moment they are created?
New-record completeness is key-field completeness measured only on records created recently. It separates a cleanup problem from a process problem.
Run the report from number 2 again with a Created Date filter for the last 30 days. If the old backlog is patchy but the new records are full, you have cleanup to do and the door is fine. If new records are as patchy as the old ones, cleaning is pointless, because the form, the integration or the import is still letting them in incomplete. The fix there is a required field, a validation rule or a better file template, not a campaign. A load that skips these checks is the usual cause, which is why importing a CSV safely matters upstream of every number here.
Rule of thumb: new records should score at least as well as your overall figure. If they score worse, the problem is current.
7. How badly have your picklists sprawled?
Picklist sprawl is the number of distinct values in use, compared with the values you meant to allow. It tells you whether a grouped report can still be trusted.
Group a report by the picklist field and count the groups, using the Add group menu in the Lightning report builder. Compare them with the active values in Setup, and look for “Other”, blanks and near-twins like “Tech” and “Technology”. Salesforce’s Help explains why this happens: an unrestricted picklist stops users entering new values in the user interface, but the API, automation and other apps can still add them, while a restricted picklist keeps them out. Admins with permission to edit the field can still add values to either kind.
Mind that the Lightning builder lets you group by up to two row fields and two column fields, so check one field at a time. Rule of thumb: under 5% of records in odd buckets is fine, and over 20% means the field has stopped being a field.
What to do Monday:
1. Pick two objects, Accounts and Leads for most teams, and run numbers 2, 4 and 5 on them.
2. Write the three figures in a note with today’s date. That is your baseline.
3. Put a reminder in the calendar for the same day next quarter, and run the same reports.
Where we come in
Our 20-minute Salesforce data health scorecard walks through ten checks with standard reports and gives a score, so you can try this today. It is an estimate. If you want every record counted, with the records behind each figure, the Data Health Check is a two-week fixed-scope audit. It measures and recommends, and it doesn’t fix anything.
Questions people ask
How do you measure data quality in Salesforce?
Pick a handful of measurable checks, such as duplicates, completeness, staleness, orphans and ownership, express each as a share of a total, and run them on the same schedule. Standard reports do most of it. The trend matters more than the first reading.
What is a good data completeness rate in Salesforce?
There’s no published standard, so treat any figure as a rule of thumb. For fields your processes depend on, 90% or more filled is a reasonable target. Under 70% usually means people are compensating by hand. Measure only the fields something actually reads.
How often should you check Salesforce data quality?
Quarterly is enough for most teams, with new-record completeness checked more often if you’ve recently changed a form or import process. Keep the reports saved, so each run is the same measurement as the last.
Sources
- Filter across objects with cross filters: WITH and WITHOUT cross filters, editions.
- Accounts with no activities using cross filters: the stale-account report.
- Report to display records owned by inactive users: custom report type and Owner Active filter.
- Create a field or report that shows days since Last Activity: what LastActivityDate holds and its caveats.
- Create reports on duplicate records: duplicate record set reporting (summary read via search excerpt).
- Build a report in Lightning Experience: grouping rows, two row and two column groups.
- Guide to creating custom picklists: restricted and unrestricted picklist behavior.