A study can have a brilliant design and still fail in the field. The questionnaire reads well in an office but confuses respondents. A translation shifts the meaning of a key question. A tired enumerator skips a section. None of this shows up in the final dataset as an error. It shows up as data that looks fine and is quietly wrong.
We have managed field research in Uganda since 2016, for universities in the United States, evaluation firms, NGOs and local government partners. Over those years, a handful of habits have done more for data quality than any piece of software. Here are five.
1. Pilot until it is boring
Every instrument gets tested on real people before real fieldwork starts. Not once, but until the pilot stops teaching us anything new.
In our air pollution study with the University of Virginia and UC Santa Barbara in Nansana, we ran three separate pilot studies of the waste-burning assessment before baseline data collection began. Each round changed something: how a Cell was divided for counting, what counted as evidence of burning, how locations were recorded. By the time the baseline started, the method was dull, predictable and repeatable. That is exactly what you want.
The habit: budget time and money for more than one pilot, and treat every change from a pilot as a finding.
2. Translate, then translate back
Uganda is multilingual, and most surveys are carried out in several languages. A question that is clear in English can become leading, vague or simply different in Luganda, Runyankore or Acholi.
We translate instruments into the languages of the study area, then have a different person translate them back into English, and compare. Where the meaning has drifted, we fix it with the translators and the research team together.
The habit: never let a translation go to the field without a back-translation check.
3. Train for the hard cases, not the easy ones
Enumerator training often walks through the questionnaire from top to bottom. That covers the easy cases. The hard cases are what go wrong: a respondent who asks what the study is really about, a household where the selected person is absent, a question that upsets someone.
For a 2024 survey experiment with the University of Texas at Austin, the design only worked if respondents did not know the exact purpose of the study. So our training spent real time on how to introduce the study honestly without revealing the experiment, how to obtain informed consent properly, and how to follow a random-walk protocol for selecting respondents. We practised these moments until they were second nature.
The habit: build training around the situations most likely to bias your data, and practise them through role play.
4. Check the data every day
The cheapest time to fix a problem is the day after it happens. If an enumerator is misreading a question, you want to know after five interviews, not after five hundred.
We sync data daily and review it daily: completion rates, missing values, unusual patterns, interview durations, GPS locations. Our supervisors call enumerators the same evening if something looks wrong. We have since built this into an automated service, FieldPulse, which runs these checks overnight and sends a report before the next morning’s fieldwork.
The habit: decide your quality checks before fieldwork starts, and run them every day it lasts.
5. Write down every decision
Fieldwork never goes exactly to plan. In Nansana, a national 42-day lockdown stopped all activity in mid-2021. A survey of local leaders showed that only 45 of 130 Cells had formal waste collection, which changed which areas the study could ethically target. One Cell’s chairperson chose not to take part, and their wish was respected.
Each of these changes was documented when it happened, with the reason. That record is what lets an investigator defend the final analysis to reviewers, years later.
The habit: keep a field log of every deviation from the plan, why it happened and who agreed to it.
The common thread
None of these habits is new or expensive. What they share is discipline: doing the unglamorous checks every time, even when the timeline is tight. That discipline is the difference between data that looks complete and data you can trust.
If you are planning a study in Uganda and want a field partner who works this way, see our research services or get in touch.