Data Quality Is Easier to Protect Than to Recover

Working with inherited data can sometimes be a challenge for evaluators. Someone else designed the tool, trained the enumerators, collected the data. And now it’s your job to make sense of it.

I’ve been doing a lot of this recently across different projects. And sometimes, the most interesting part of the analysis isn’t in the findings. It’s what you notice along the way.

On one project, a cross-check between two related datasets revealed hundreds of missing records. On another, completion records for a medical process that couldn’t have been completed in the reported timeframe. Neither issue was caught at the time of data collection. Both were spotted by slowing down and asking whether the numbers actually made sense. After years of working across multi-country data collection, I’ve learned to treat that instinct seriously.

When something like this happens, the temptation is to panic. After all, evaluations often have tight deadlines and clients expecting results. But in my experience, the right response is pretty straightforward. Stay calm, lay out the options clearly, and bring the team and client into the conversation early.

In both cases above, we were able to identify what had gone wrong, agree on a solution, and communicate it transparently. People generally respond well to “we found an issue and here’s how we’re handling it” far better than discovering a problem themselves later.

Two lessons from this kind of work. First: data quality is much easier to protect at the point of collection than to recover at the point of analysis. Second: when errors do surface, how you respond matters as much as how you caught them. Stay calm, present solutions, and treat it as a learning opportunity rather than a crisis.

In large, complex, multi-country evaluations, something will almost always go wrong. The question is whether you’ve built the systems to catch it early, and the relationships to handle it well when you haven’t.