Cost

Costing and Quoting a Master Data Cleanup or Governance Program

What actually drives cost per record in master data cleanup and governance, how to build a defensible quote, and where estimates blow up.

Master data cleanup is priced per record, and the dominant cost is fully loaded steward labor, not software. A data steward at $38 per hour base carries to roughly $57 fully loaded once benefits, supervision, and overhead are added. If a simple field fill takes 5 minutes, that record costs about $4.75. A research record requiring supplier confirmation and engineering sign-off can take 45 minutes, costing $42.75, a ninefold spread. Quote a blended rate only when you have profiled the mix; otherwise you either lose money on the hard records or price yourself out on the easy ones. Segment the backlog into simple, moderate, and research tiers before you attach a number.

Tooling is a fixed cost that must be amortized across the population, and getting the denominator wrong is where quotes go soft. A de-duplication or data quality platform at $6,000 to $40,000 setup is trivial per record across 200,000 items (3 to 20 cents each) but crushing across a one-time 800-record cleanup ($7.50 each). The Duplicate Item Cost and Work Center Data Completeness calculators split variable from fixed for exactly this reason. When you quote, state the tooling cost as a separate line, not buried in the per-record rate, so the buyer sees that scaling the population lowers unit cost and a tiny one-off job carries an unavoidable fixed penalty.

False positives are the silent margin killer. Fuzzy matching flags candidates, but only 30 to 60 percent are true duplicates before rule tuning. If you quote 800 flagged records as if all need merging at $22 each, you promise $17,600 of work; at a realistic 45 percent true rate the actual merge labor is $7,920, but the review time on the other 440 false positives is real and unbilled if you forgot it. Price review separately from remediation: assume 2 to 4 minutes to disposition each flag whether or not it is a true duplicate, then add merge cost only to the confirmed share.

Rework and approval routing inflate every quote by 15 to 40 percent, and leaving it out is the most common estimating error. Records bounce back for missing attributes, approvers sit on queues, and rejected changes get redone. A 10 percent allowance suits clean, well-defined field updates; mixed backlogs need 25 percent; research-heavy or cross-functional work needs 30 to 40 percent. On a 240-hour base cleanup, a 30 percent allowance adds 72 hours, or about $4,100 at loaded rates. Quote the allowance as an explicit contingency line tied to the backlog complexity, and defend it with your profiling data rather than a round guess.

Machine and system time is minor in this category but not zero. Profiling runs, match jobs, and mass-update batches consume ERP and data-quality license time plus an administrator to run them. Budget 8 to 20 hours of technical setup per cleanup wave for query building, rule configuration, and validation, at a $75 to $110 loaded engineer rate. That is often $1,000 to $2,000 that estimators forget because they focus on steward heads. It scales with the number of domains, not records: cleaning items, BOMs, routings, and suppliers each needs its own rules, so a four-domain program carries four setup blocks.

Build the quote bottom-up in four layers so it survives scrutiny. Layer one, variable remediation: true-defect records times loaded minutes per tier. Layer two, review labor: all flagged records times disposition time. Layer three, fixed tooling and technical setup as separate lines. Layer four, the complexity allowance as a stated percentage. For a 12,000-record item cleanup at 20 percent research records, expect roughly 400 to 600 steward hours, $6,000 to $12,000 tooling, and a 25 percent allowance, landing near $35,000 to $50,000. The Item Master Cleanup Effort and Master Data Governance Workload calculators feed layers one and two directly.

Estimates go wrong in five predictable places. Using an easy-record rate on a hard-record backlog understates hours by 3x to 10x. Pricing every flag as a true positive overstates merge labor by roughly double. Burying fixed tooling in the per-record rate hides that small jobs are inherently expensive. Ignoring the review cost of false positives eats the whole margin. And omitting demand-splitting or downstream savings understates the payback, which is a quoting sin in reverse because it kills projects that should be funded. Cross-check every quote against actuals from the first 200 records before committing to the full population.

Frame the quote against avoided cost, because governance work competes for budget it does not directly earn. A duplicate part splits demand so MRP under-orders and a planner carries phantom safety stock on two records; a wrong UoM factor can over-purchase by 10x. Those recurring losses often dwarf a $40,000 cleanup, and pairing your cost quote with Data Error Cost and Data Governance ROI figures turns a spend request into an investment case. Present the cleanup price, the tooling amortization, and the annual avoided loss side by side so finance sees payback in cycles, not just the invoice.

Published 2026-07-01.