Placeholder Values Are the Data Quality Problem Nobody Measures
Your fields say 100% complete. The data says nothing.
Part 3 of the Clean Data for AI series | By Jeremy Carmona
A required field does its job. It blocks the save until something is entered. The problem is what gets entered.
One organization I assessed reported on industry segmentation for two years using their Industry picklist. The field was required. Completion rate: 100%. When we profiled the actual values, 34% said "Other." Leadership had been using those segments for strategic planning. A third of the data behind their decisions carried zero information.
That is the placeholder problem. Required fields create a false sense of completeness. Users enter "Other," "TBD," "N/A," a period, or a single space to get past validation. The field registers as populated. Reports count it. AI uses it. Nobody measures what the value actually contains.
My AI Readiness Checklist sets the red flag at 30% critical field blanks. Placeholder values create the same blind spot while passing every validation rule and every completeness report you run. The field is technically full. It is functionally empty.
Why AI Makes This Problem Expensive
Before AI, a placeholder in the Industry field was an inconvenience. A marketer running a segmented campaign might notice the "Other" bucket was suspiciously large and filter it out manually. The cost was one person's time and one imperfect campaign.
With Agentforce, Einstein, and Data Cloud, the AI does not filter. It does not suspect. It treats "Other" as a legitimate industry segment. Einstein recommendations for "Other" contacts blend patterns from every industry into a meaningless average. Data Cloud segments built on "Other" dilute every audience they join. An Agentforce agent personalizing outreach to an "Other" contact generates generic output at best, irrelevant output at worst.
The volume compounds it. A human might process 50 records in an afternoon and notice the "Other" pattern. An AI agent processes thousands in minutes and never looks back. Every placeholder value it encounters is a decision made on nothing, at scale.
The Excel Diagnostic
You can measure your placeholder exposure in 30 minutes. Export the object where placeholders are most likely (Contact, Lead, or Account), and open it in Excel.
Step 1: Count known placeholder strings.
For any picklist or text field, use COUNTIF to count specific placeholder values:
=COUNTIF(C2:C5000,"Other")+COUNTIF(C2:C5000,"N/A")+COUNTIF(C2:C5000,"TBD")+COUNTIF(C2:C5000,".")
This formula counts four common placeholders in the Industry column. Adjust the range and add strings for your org's specific patterns. Common additions: "None," "Unknown," "Test," "NA," "temp," single spaces, and single characters.
Divide by total records to get your placeholder rate:
=([placeholder count]/COUNTA(A2:A5000))*100
If the result is above 10%, you have a problem worth fixing. Above 30% is a red flag.
Step 2: Visual scan with Conditional Formatting.
Select the column. Apply Conditional Formatting > Highlight Cells Rules > Text that Contains. Enter each placeholder string one at a time: "Other," "N/A," "TBD," ".". This highlights every placeholder visually, making patterns obvious. You will often find clusters: records imported from a specific source, records created during a specific period, or records owned by a specific user who favored "TBD" over leaving fields blank.
Step 3: Profile by field.
Repeat Steps 1 and 2 for every field you consider decision-critical. The fields that matter most are the ones your organization uses for segmentation, routing, reporting, or AI personalization. Common high-impact fields: Industry, Lead Source, Rating, Account Type, and any custom picklist used in automation or assignment rules.
Build a simple table: Field Name, Total Records, Placeholder Count, Placeholder Rate. Sort by rate, descending. The fields at the top of that table are the ones degrading your AI outputs right now.
Common mistake: Checking only picklist fields. Free-text fields like Description, Notes, and custom text areas also accumulate placeholders, and those are harder to catch because the variations are unlimited. A Description field that reads "." or "test" or "asdf" is a placeholder with no standard spelling.
What Salesforce Should Prevent Going Forward
The audit finds the current state. Prevention handles the future.
Validation rules are the first line of defense. A validation rule that rejects known placeholder strings stops the problem at the point of entry:
CONTAINS("Other|N/A|TBD|.|None|Unknown", TEXT(Industry))
This formula fires when the value matches any string in the pipe-delimited list. In Salesforce validation rules, the formula evaluates to TRUE when the error should fire. When the rule fires, the save is blocked and the user sees the error message you configure: "Please select a specific industry rather than a placeholder value."Two implementation notes. First, this rule blocks "Other" from being saved, so you need to provide the user with a real alternative. If your picklist does not have enough specific options, add them before activating the rule. Second, test in a sandbox. A validation rule that blocks legitimate values creates more friction than the placeholders it prevents.
Picklist governance is the second control. Review your picklist values quarterly. Remove or rename generic options. If "Other" is a catch-all for values that should have their own entry, replace "Other" with those entries. If "Other" is genuinely needed (some fields have long-tail values that do not justify individual options), add a dependent text field that captures what "Other" means. The combination of a picklist value and a description field gives AI something to work with.
Quarterly placeholder audit report. Create a Salesforce report that counts placeholder values in your critical fields. Schedule it monthly or quarterly. Assign a named owner who reviews the counts and escalates when rates exceed your threshold. Without this cadence, the validation rules handle new records but the existing placeholder inventory sits unchanged.
The Governance Decision Nobody Makes
Who decides what counts as a placeholder? Who reviews the picklist values and determines whether "Other" should stay or be replaced with specific options? Who approves new picklist additions?
These are not technical decisions. They are governance decisions. In most organizations, nobody owns them. Picklist values accumulate over years. "Other" stays because removing it might break reports. New values get added without review. The field grows until it carries 30 options that could be 12, and 4 of those 12 are variations of the same thing.
The fix is a named owner for each critical picklist. That person reviews the values quarterly, approves additions, retires redundant options, and monitors the placeholder rate. This is a 30-minute task per field, four times a year. The alternative is AI making decisions on "Other" indefinitely.
What to Do This Week
Export your Contact records with all picklist fields
Run COUNTIF on each field for "Other," "N/A," "TBD," and "."
Calculate the placeholder rate per field
Apply Conditional Formatting to see the pattern visually
Identify which fields are used in segmentation, routing, or AI features
Prioritize the fields with the highest placeholder rates and the highest business impact
Draft a validation rule for one field and test it in sandbox
A field that says "Other" tells AI exactly as much as a blank field. The difference is that "Other" passes every completeness check you run, so nobody knows.
Related CCC resources:
Salesforce Duplicate Records: Find and Fix Before AI (Part 2)
AI Readiness Scorecard (free, 2-minute self-assessment)
Sources:
Salesforce Help: Validation Rules documentation (help.salesforce.com)
CCC assessment methodology (30+ field assessments, 2024-2026)
This article is Part 3 of the Clean Data for AI series. Part 2 covered duplicate records and identity resolution. Part 4 covers data standardization as a system, not a cleanup project.
Last updated: September 23, 2026. If you find an error, contact contact@clearconciseconsulting.com.
