Data quality projects stall because they are framed as one heroic cleanup. Someone spends a month deduping accounts, declares victory, and six months later the org is just as messy. Quality is not a project; it is a habit with a feedback loop. Here is a version small teams can actually sustain.
Measure before you clean
Start with a handful of quality metrics that map to business pain: duplicate account rate, percentage of open opportunities missing a close date, contacts without a valid email. Build a simple dashboard for these and note the baseline. Without a number, "the data is bad" is an opinion; with one, you can show progress and defend the time spent.
Fix the intake, then the backlog
Cleaning history while new bad data pours in is bailing a leaky boat. Address the source first: validation rules on required fields, restricted picklists instead of free text, and duplicate rules that block or warn at the point of entry. Only once the intake is controlled does it make sense to work through the historical backlog, because your cleanup will finally stay clean.
Prioritize by what the business uses
Not all fields deserve equal effort. A field that drives routing, scoring, or a report leadership reads every week is worth enforcing; a field no process touches may not be worth cleaning at all. Rank remediation by downstream impact, and be willing to retire fields that exist only to collect noise.
Make ownership explicit
Data decays when no one owns it. Assign each critical object or field area an owner responsible for its definition and health, and give them the dashboard to watch. Quality that depends on everyone caring equally tends to decay to no one caring; a named owner with a visible metric does not.
Close the loop with a recurring review
A one-time cleanup fades; a standing checkpoint holds. Put a short data-quality review on a regular cadence, monthly for most teams, where the owner walks the dashboard, notes what moved, and picks the next fix. Keep it lightweight enough that it actually happens. The point is not a perfect org but a trend line that stays flat or improves instead of quietly sliding backward between projects.
The takeaway
Treat data quality as an operating rhythm: measure, control the intake, work the backlog by impact, assign ownership, and review on a cadence. It is less dramatic than a big cleanup and far more effective, because the improvements hold instead of quietly eroding until the next fire drill.