Data Quality
At one stationery retailer, cleaning up the customer data resulted in 7.5 million corrections in just a few months. No new model, no new campaign—just getting the basics right.
Data quality is the least visible part of your data strategy, and yet it’s the part that determines whether everything else works. You rarely notice corrupted data until you filter for it. And by then, it’s too late.
Data quality is about the usability of your customer data. In practice, it comes down to four areas: address, email, name, and phone number. Each area can contain errors that seem minor—an address without a ZIP code, an email with a typo in the domain, a name spelled differently in two places, or a phone number in the wrong format.
On their own, these details don't amount to much. Together, they determine two things: whether you reach your customer, and whether your system recognizes them as a single customer.
Poor-quality data makes customers untraceable and hides duplicates. The latter is the most insidious. If the same customer is listed in the system with a slightly different address, a typo in their email, or an incorrect year of birth, your platform will see two or three customers where there is actually only one. Your segmentation then doesn’t categorize your customers—it categorizes your errors. And a predictive model built on top of that data will mainly predict noise.
At the stationery retailer, this became apparent the moment the data was accurate. The number of correctly merged duplicate profiles jumped from an average of 1,500 to 2,000 per month to eight to ten times that level. Each merge represents a customer who was previously fragmented in the system and is now a single, accurate, usable profile. That is the foundation upon which a 360° customer view truly takes on meaning.
The full case study
3.7 million records, 7.5 million corrections, and two databases that were finally merged into a single customer view. Read how Data Quality Services got things back on track for a stationery retailer.
Data quality isn't a one-time project you can just check off your list; it's an ongoing process. If you have a solid foundation, you can build every selection, every segment, and every model on solid ground rather than quicksand. Data quality comes first, and only then the model.
