Practical framework / 5 MIN READ
Keep missing values from masquerading as poor performance
Give absent data a meaning before choosing a replacement value.
AI-assisted research and analysis for the SNS Creative Fieldbook
Sources checked 2026-11-06
Zero can mean an observed absence of activity. A blank may mean unavailable, inapplicable, delayed or uncollected information. Collapsing those states changes calculations and can send attention to the wrong problem.
What you will leave with: Create a missing-value contract that makes aggregation, display and escalation rules explicit.
Use this when
- A dashboard turns blank cost or revenue fields into zeros.
- An average changes when previously empty fields are populated.
This lesson does not provide a universal statistical imputation method or justify guessing unavailable customer information.
Name the missing state
For each important field, ask why it can be absent. A campaign that delivered no impressions differs from a campaign whose connector failed. A customer with no purchase differs from a customer whose purchase status is not yet known. A measure can also be inapplicable to a particular campaign type.
Represent those meanings using documented status fields or display labels where needed. Avoid inventing a detailed reason from a blank alone. If the source cannot distinguish causes, preserve unknown and investigate the system rather than assigning a commercially convenient interpretation.
Inspect aggregation semantics
PostgreSQL distinguishes counting all rows from counting non-null values; its average uses non-null inputs, and a sum over no rows returns null. These technical rules explain why apparently similar calculations can yield different results. Check the actual behavior of the tool in use.
Write the intended business calculation independently. Should an average order amount include only known valid amounts? If so, report coverage. Should a count include records with unknown status? Answer that question explicitly. Replacing null with zero before averaging changes the population and should be a deliberate, justified decision.
Test empty and partial cases
Build examples with observed zero, missing value, no matching records and an inapplicable field. Confirm that both the computation and the chart preserve the distinction. A blank bar rendered as zero can mislead even when the warehouse stores the value correctly.
For incomplete reports, show the number or share of records with known values and explain whether coverage differs across groups. A ranking based on substantially different coverage may need to be withheld. Do not infer that a smaller reported total means worse performance when missingness could explain it.
Choose action rules by meaning
Observed zero may warrant performance diagnosis; unavailable data may warrant connector repair; inapplicable data may need no action. Assign an owner and response to each meaningful state. Keep a clear path from the displayed label to the underlying evidence.
If the business needs a planning estimate despite missing information, label it as a scenario with assumptions and keep the observed report separate. Never insert an estimated value into the source record without an authorized, traceable process. Revisit the contract when the provider changes how absent fields are encoded.
Make the reasoning concrete.
A hypothetical set of three campaign revenue fields contains $100, $0 and unknown. Averaging known values yields $50 across two campaigns. Replacing the unknown with zero yields $33.33 across three, while neither calculation reveals the true three-campaign average.
Fill every blank with zero
Opening: Show $33.33 as a complete average.
Hypothesis: The chart becomes easy to render.
Risk: Missing evidence is falsely treated as observed failure.
Show value and coverage
Opening: Show $50 among two known campaigns, with one unknown.
Hypothesis: The reader sees the calculation’s scope.
Risk: The known subset may not represent the missing campaign.
Use the known-value result only with coverage and its limitation. Investigate the missing record before comparing the complete portfolio with another period.
Take the questions into your next brief.
Use these prompts in your own notes. Nothing is submitted from this page.
Meaning
Which absent states can this field represent?
Calculation
How does the tool treat each state?
Coverage
How many eligible records have known values?
Display
Can the chart distinguish zero, unknown and inapplicable?
Action
Who resolves each type of absence?
Watch for the wrong conclusion.
- Replacing every blank with zero.
- Comparing groups with hidden coverage differences.
- Treating no rows as proven no activity.
- Writing an estimate back as an observed fact.
Your next action
Audit one dashboard’s blank-value behavior with four deliberate test cases and document the correct display for each.
Sources & scope
Frameworks, worksheets and hypothetical examples are original editorial synthesis. Linked primary sources support only the stated concepts. Source checking is not an independent audit of effectiveness. No named human author or reviewer is claimed.
- PostgreSQL: Aggregate functions →Primary technical documentation · Supports: Distinguishes row counts, non-null counts, averages of non-null values and null results for empty sums. These are PostgreSQL semantics, not universal business definitions.