Get the right data, get the data right
Five rules we hold research data collection to, and why validation at the point of capture beats any amount of cleaning afterwards.
AppNet Solutions · 27 January 2026 · 4 min read
Research budgets tend to allocate generously for analysis and thinly for collection. The ratio should usually be the other way round: no analysis recovers information that was never captured correctly.
We hold field data collection to five rules.
Get the right data
Instrument for the question the study is actually trying to answer, not for everything that might one day be interesting. Every additional field costs enumerator time, raises refusal rates, and adds a column somebody will later have to explain.
Get the data right
Validate at the point of capture. Ranges, skip logic, internal consistency, GPS plausibility — all of it enforced on the device while the respondent is still present. Cleaning afterwards is guessing with extra steps.
Get the data right away
The gap between collection and review should be measured in hours. A systematic enumerator error found on day two costs a retraining session. The same error found at analysis costs the study.
Get the data the right way
Consent, provenance and handling documented as carefully as the responses themselves. If you cannot say who collected a record, when, under what consent, and what has happened to it since, it will not survive review — and it should not.
Get the right data management
Storage, versioning and access control for the life of the study and beyond it. Most data loss we see is not dramatic. It is a laptop, a departure, and no second copy.
None of this is exotic. It is the difference between a dataset that supports a finding and one that merely accompanies it.
