Not Every Data Problem Deserves Fixing: A Leader’s Guide to Data Priorities
Some data problems are expensive. Some are annoying. Some are just digital lint.
The trouble starts when organizations treat all three the same. A duplicate customer record that affects quarterly revenue reporting is not the same as a misspelled warehouse code from 2009 that nobody uses. One deserves money. The other deserves a polite nod and maybe a sandwich.
Every enterprise has messy data. Legacy systems, acquisitions, manual workarounds, old product hierarchies, half-retired spreadsheets, and fields named `FINAL_FINAL_USE_THIS_ONE` all leave marks. The question is not whether the data is imperfect. It is whether the imperfection changes a decision that matters.
That distinction should sit at the center of modern data strategy.
Imperfect data is not the same as harmful data
Bad data has a reputation problem. The phrase makes every row sound dangerous. In reality, data quality exists on a spectrum.
A customer table may contain stale phone numbers. That matters if sales teams call those numbers daily. It matters less if the table only supports an annual historical analysis. A product catalog may have inconsistent abbreviations. That matters if pricing rules depend on those abbreviations. It matters less if the field appears only in an old lookup report that nobody has opened since the last season of a canceled TV show.
The economic question is simple:
Does this data issue change a decision, delay a decision, or create material risk?
If the answer is no, fixing it may feel responsible but still be a poor use of capital.
That is hard for data teams to say out loud. Many data professionals were trained to pursue clean, complete, controlled data. That instinct is useful. It prevents chaos. But when taken too far, it creates a different problem: expensive perfection.
Enterprise-wide cleanup programs can take months or years. They pull business experts away from daily work. They require system changes, mapping rules, governance meetings, testing, training, and maintenance. None of that is free. The invoice may come from a vendor, but the real cost often sits inside the business.
A better target is decision-grade data. That means data good enough for the decision it supports, given the financial exposure, frequency, and risk involved.
That bar changes by use case.
Monthly executive revenue reporting needs high trust. So does regulatory reporting. So does pricing. By contrast, a one-time exploratory analysis for a low-value internal question may not need gold-plated source data. It needs honest caveats and a short shelf life.
Perfect data is a lovely idea. So is a garage with no random cables in it. Good luck with both.
The five tests that separate noise from value
Data problems deserve investment when they threaten business outcomes. The trick is to judge them using business criteria, not only technical criteria.
Here are five tests that work well in executive prioritization.
Business criticality
Ask what process, decision, or commitment depends on the data.
If the data drives billing, revenue recognition, pricing, supply planning, financial close, customer commitments, or board reporting, treat it seriously. If it supports a low-use reference report, the fix can wait.
Critical data often hides in plain sight. A small mapping table may look harmless until it feeds margin reporting across every region. Think of it like a tiny screw in an airplane door. Size is not the point.
Decision frequency
Frequent decisions multiply the impact of bad data.
A weekly demand forecast with flawed product substitution logic can create repeated purchasing errors. A daily pricing feed with inconsistent customer segments can leak margin every day. The same error in a once-a-year archive extract may not justify urgent work.
Frequency turns small errors into recurring business costs. It also makes remediation more valuable because the benefit repeats.
Financial exposure
Estimate the dollars at risk.
This does not require false precision. A reasonable range works. Could the issue affect $50,000, $500,000, or $5 million in margin, working capital, penalties, or missed revenue? If nobody can connect the issue to a financial range, pause before funding a fix.
CFOs do not need a poetry reading about “trusted data.” They need to know whether the leak is a teaspoon or a fire hose.
Downstream dependencies
Some data issues spread.
A broken customer hierarchy may feed sales compensation, account profitability, credit limits, service levels, and executive dashboards. Fixing it once can improve many outputs. Ignoring it can create a game of spreadsheet whack-a-mole.
Map where the data travels. The more decisions it touches, the more serious the defect becomes.
Regulatory and audit needs
Some data has a legal or compliance standard attached to it.
Examples include SOX-related financial controls, HIPAA-protected health information, tax reporting, industry-specific retention rules, and privacy obligations. If a data issue creates compliance exposure, the debate changes. The question becomes how fast to reduce risk, not whether perfection has a positive ROI.
Smart data governance keeps these obligations visible without turning every field into a constitutional amendment.
A hypothetical company with 47 problems and five that matter
Picture a national manufacturer called Northstar Components. It has grown through three acquisitions. It runs two ERP systems, one CRM, a planning tool, a data warehouse, and enough spreadsheets to qualify as a protected species.
An internal review finds 47 legacy data issues. The list includes:
Duplicate customer accounts
Inconsistent product names
Missing supplier lead times
Region codes that changed after an acquisition
Old cost centers still appearing in reports
Different definitions of “active customer”
Incomplete product lifecycle dates
Manual overrides in pricing files
Forecast categories that do not align across systems
The first reaction is predictable: “We need an enterprise cleanup program.”
Maybe. Maybe not.
Northstar’s leadership team asks a sharper question: Which issues materially affect pricing, forecasting, or executive reporting?
After two weeks of interviews, report tracing, and financial sizing, the team finds that only five issues drive the bulk of risk.
Customer segment mismatches affect pricing discounts
Some large customers have inconsistent segment assignments across CRM and ERP. That causes the pricing engine to apply the wrong discount tiers. The issue affects current orders, margin, and customer negotiations.
Product substitution rules distort demand forecasts
Several legacy products have newer replacements, but the forecasting tool treats them as unrelated items. Demand looks split across old and new SKUs. Inventory teams buy too much of one item and too little of another.
Manual pricing overrides lack expiration dates
Temporary deal pricing sometimes remains active after the promotion ends. The revenue team notices only when margin reports look odd. That is not a control. That is a smoke alarm with commitment issues.
Executive revenue reports use conflicting customer hierarchies
Sales reporting groups accounts differently from finance reporting. The board deck and operating dashboard can show different revenue by strategic account. Nobody enjoys explaining that on slide 6.
Supplier lead time fields are stale for critical components
A handful of critical parts still use old lead time assumptions. Forecasts look fine, but production misses occur when suppliers cannot meet those dates.
The other 42 issues are real. They are irritating. Some should be fixed later. But they do not materially change high-value decisions right now.
Northstar funds the five high-impact fixes first. It assigns business owners. It measures before and after results. It updates governance rules only where the defects can recur. It leaves the rest in a managed backlog.
That choice creates more value than a broad cleanup program because it targets the parts of the data estate tied to margin, forecast accuracy, leadership trust, and operational risk. It also delivers faster proof. Executives can see what changed. Teams stop arguing about whose spreadsheet is “more right,” which is excellent because nobody should have to win that trophy.
How to decide what to fix now, later, or never
A practical data prioritization framework should fit on one page. If it needs a 90-minute explanation and a commemorative binder, it will not survive budget season.
Use this scoring model to sort data issues into three buckets.
Decision factor | What to ask | High-priority signal |
Business use | What decision depends on this data? | Pricing, forecasting, financial close, customer commitments, board reporting |
Frequency | How often is the decision made? | Daily, weekly, or monthly decisions with repeated impact |
Financial exposure | What dollars are at risk? | Margin, revenue, working capital, penalties, or material labor cost |
Dependencies | What systems or reports consume it? | Multiple downstream tools or executive dashboards |
Risk | Are there audit, privacy, or regulatory requirements? | Legal, compliance, financial control, or contractual exposure |
Remediation cost | What will fixing it require? | Low effort compared with business value, or high effort with clear risk reduction |
Recurrence | Will the issue come back? | Weak controls, unclear ownership, or manual workarounds |
Then assign each issue to one of three categories.
Fix now
These issues affect important decisions, carry meaningful financial or risk exposure, and occur often enough to justify action.
Typical examples include pricing errors, failed financial reconciliations, unreliable executive metrics, regulatory reporting gaps, and master data defects that feed many systems.
“Fix now” does not mean boil the ocean. It means define the specific defect, its business effect, the owner, the target state, and the control that prevents recurrence.
Fix later
These issues matter, but not enough to interrupt higher-value work.
Maybe the process runs quarterly. Maybe the affected business line is small. Maybe a system replacement will solve the issue in six months. Maybe the data supports analysis but not execution.
Place these items in a visible backlog. Review them during planning cycles. Do not let them become data purgatory, where issues go to wear tiny sweaters and haunt dashboards forever.
Do not fix unless conditions change
Some issues should stay unfixed.
That sounds reckless. It is often disciplined.
If the data is rarely used, has low business exposure, has no regulatory need, and would cost more to fix than the decision is worth, document the risk and move on. Archive it, label it, restrict its use, or add a warning to the report.
A known flaw with low impact is better than an expensive fix with no business return.
The cost of remediation can be bigger than the defect
Every data fix has a total cost. The visible cost is technology work. The hidden cost is everything else.
Common remediation costs include:
Business time to define correct values and rules
Data profiling and root-cause analysis
System changes or integration changes
Manual cleanup
Testing and reconciliation
User training
Ongoing stewardship
Governance meetings and exception handling
If an issue costs $25,000 a year in noise and $400,000 to fix, the answer may be “not now.” If an issue costs $2 million a year in pricing leakage and $300,000 to fix, order the pizza and get moving.
This is where many programs go wrong. They rank defects by technical ugliness rather than economic harm. A field can be horrifying and still harmless. Another field can look clean but contain the wrong business logic.
Pretty data can still lie. It just lies in a nicer font.
The best remediation plans start with the decision, then trace backward to the minimum data improvement needed. That may mean fixing one key field, one mapping rule, one integration point, or one ownership gap. It rarely means cleaning every table in sight.
A focused plan also protects credibility. When leaders see data teams solving painful business problems, support grows. When they see endless cleanup with fuzzy benefits, patience burns fast.
What leaders should ask before funding a data cleanup effort
Before approving a broad data quality project, ask direct questions.
What decision or process fails because of this issue?
How often does that decision happen?
What is the estimated financial or risk exposure?
Which reports, systems, and teams depend on the data?
What is the root cause?
What would remediation cost, including business time?
What control prevents the issue from returning?
What happens if we do nothing for 12 months?
The final question is the best one. It clears the fog.
If doing nothing creates material risk, fund the work. If doing nothing creates mild discomfort, schedule it later. If doing nothing creates no meaningful consequence, stop feeding the project monster.
Useful analytics strategy is not a crusade for spotless data. It is a set of choices about where better data will improve decisions.
That is the work that pays.
FAQ
Does this mean data quality does not matter?
No. It means data quality should be tied to business use. High-impact data needs strong standards. Low-impact data may only need labels, caveats, or retirement.
How often should data priorities be reviewed?
Review top data issues at least during quarterly planning. Review them sooner when the business changes, such as after an acquisition, system migration, product launch, or regulatory change.
Who should own data issue prioritization?
Business and technology should share it. Business leaders define impact and urgency. Data and IT teams assess root cause, cost, dependencies, and feasibility.
Should companies still invest in enterprise data platforms?
Yes, when the investment supports clear business use cases. Platforms help when they reduce repeated integration work, improve trusted reporting, or support high-value analytics. They disappoint when used as a substitute for priority setting.
What is the fastest way to start?
Pick one high-value decision, such as pricing, forecasting, or executive reporting. Trace the data behind it. Identify the defects that change the outcome. Fix those first.
A pragmatic way forward
The strongest data programs do not try to fix everything. They focus on the data problems that change decisions, increase risk, or drain money.
That takes discipline. It also takes a willingness to leave some mess alone. Every company has an attic. The goal is not to alphabetize every box. The goal is to know which box contains the tax records, the family photos, and the raccoon. Deal with those first.
Data Consultants Inc. helps organizations connect data work to business outcomes, especially in data strategy and business intelligence. That means finding the issues that matter, sizing their impact, and building a practical roadmap for what to fix now, later, or not at all.
For a closer look at how DCI approaches this work, explore its data strategy and business intelligence services.
Fix the data that changes the business. Manage the rest with clear eyes. That is not lowering the bar. It is finally putting the bar in the right place.
Comments