# Data qualityData qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → metrics that predict bad decisions
A partner at a mid-size consulting firm signs off on a client's Q3 margin report showing 32% utilization. Three weeks later, finance discovers 15% of consultant timesheets were logged 10+ days late, some backfilled after invoices had already gone out. The real utilization was 27%. Every staffing decision made off that report, from hiring to bench management, was built on a number that was wrong before anyone opened the spreadsheet.
This is not a rare glitch. In professional services (consulting, law, accounting, agencies), the core operational data (time, staffing, billing) is manually entered by people who have no incentive to do it promptly. Data qualityData qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI. problems here don't crash systems. They quietly bend every downstream report.
Professional services firms run on three interlocking datasets. Get familiar with all three before trusting any metric built on top of them.
1. Timesheets. The atomic unit of the business. Captures who worked on what, for how long, billable or not. Usually lives in systems like SAP Fieldglass, Replicon, or practice-specific tools (Aderant and Elite for law firms, Deltek for consulting and government contractors).
2. Staffing/resourcing data. Who is allocated to which engagement, at what rate, for what duration. This feeds utilization (percentage of available hours billed to clients) and bench (unstaffed, non-billable capacity).
3. Billing and WIP (work in progress). WIP is unbilled work already delivered, sitting as a kind of inventory. Billing data converts timesheets into invoices; WIP is the gap between work done and cash requested.
These three datasets should reconcile. In practice, they rarely do cleanly, and the gaps are where bad decisions get seeded.
Definition: the share of expected records that actually exist and are populated with required fields.
Worked example: a 200-person consulting practice expects 200 weekly timesheets. If 170 are submitted with all required fields (project code, hours, task category), completeness = 170/200 = 85%.
Anything below roughly 90 to 95% completeness (a common internal benchmark cited by professional services operations teams, treat as an estimate rather than a regulatory standard) should trigger review before reports go out. Low completeness doesn't just mean missing data, it usually means the missing records are systematically different (people avoid logging unbillable or awkward hours), which biases the average.
Definition: the share of records submitted after the expected reporting cutoff.
This is the metric in the hook. If 15% of timesheets are entered more than a few days after the work occurred, any margin report generated at the standard cutoff is missing that slice of hours entirely, or worse, includes them retroactively and creates silent restatements.
Simple lag rate formula:
Timeliness lag rate = (records submitted after cutoff) / (total expected records)
Example:
Expected timesheets this week: 200
Submitted after the Friday 5pm cutoff: 30
Lag rate = 30 / 200 = 15%Why this matters mechanically: utilization and realization (billed revenue as a percentage of standard rate value) are both calculated as ratios with hours in the denominator or numerator. If 15% of hours are missing at report time, utilization looks artificially low that week and artificially inflated the following week when the late entries land. A manager reading week-over-week trend lines sees noise that looks like a real signal, like a project ramping up or down, when it's purely a reporting artifact.
Definition: the share of records that represent the same underlying event logged more than once.
Common in professional services when consultants log hours in both a client-specific tracking tool and the firm's internal system, or when an engagement code changes mid-project and old and new codes both accumulate entries. Duplication inflates apparent hours worked, which can overstate cost-to-serve on an engagement even while understating margin, a confusing combination that often triggers a wrong root-cause investigation (people blame pricing, when it's data hygiene).
Rough calculation: if a reconciliation check finds 12 duplicate entries out of 1,000 total logged entries, duplication rate = 1.2%. Firms with integrated single-entry timesheet systems typically report sub-1% duplication (industry estimate); firms running parallel systems across a merger or multi-tool stack can see this climb into the mid-single digits.
Here's the mechanism worth internalizing. Margin reporting in professional services typically flows:
Timesheets → Utilization → Realized revenue → Cost allocation → Reported margin
A quality defect at step one doesn't stay at step one. A 15% timesheet lag rate means step two (utilization) is wrong, which means step three (realized revenue, since unbilled hours can't yet be invoiced) is wrong, which distorts step four and five. By the time a partner or engagement manager sees "margin," it has passed through four compounding rounds of a garbage-in problem.
This is the professional services version of a broader data qualitydata qualityThe degree to which data is fit for purpose: accurate, complete, consistent, timely, valid and unique. Poor quality data undermines analytics, reporting and AI.Voir la définition complète → principle: errors propagate downstream and get harder to trace the further they travel from their source (general public-sector framing, but the propagation logic applies directly here).
Mature firms treat these three metrics as a standing dashboard, not a one-off audit, with named owners:
European firms operating under GDPR (General Data Protection Regulation, the EU's data protection law) face an added layer: timesheet and billing data often contain personal data about employees' working patterns, so retention periods and access controls need documenting alongside the quality metrics themselves. This is a governance overlay, not a reason to skip the quality checks.
Vérification des acquis
1. Why do data quality problems in professional services firms tend to go unnoticed rather than trigger obvious alarms?
2. In the Q3 margin report example, what was the fundamental problem with the decision-making process?
3. What does WIP (work in progress) represent in the context of billing data?
4. Select ALL correct answers about the three core datasets in professional services firms.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers about why late or backfilled timesheet entries are dangerous for decision-making.
Sélectionnez toutes les réponses correctes.
As-of-2025/2026 estimates, treat as directional, not audited industry standards:
None of these are regulatory requirements. They're operational hygiene benchmarks that firms like Deloitte, Accenture, and mid-market consultancies build into internal operations reviews; treat the specific percentages as industry-observed estimates rather than published standards.