Designing a KoboToolbox form seems simple until you receive the first field data. Here are five mistakes I still encounter regularly, and how to avoid them.

1. Forgetting validation on numeric fields

This is the most common error. A surveyor enters 250 instead of 25.0 for a child's weight, and no one notices until analysis — three weeks later.

The solution is simple: systematically add a range constraint to each numeric field. In XLSForm: 

type        | name   | label   | constraint           | constraint_message
integer     | weight | Weight  | . >= 2 and . <= 80   | Unusual weight

The error message appears immediately on the surveyor's device, who can correct on the spot.

2. Hard-coding choices instead of using an external file

When you have a list of 200 villages, don't enter them one by one in the choices tab. Use an external CSV file (itemsets.csv) that you can update without touching the form.

Why it matters

  • You avoid typos in locality names
  • You can add or remove choices without redeploying the form
  • The form stays fast to load on older phones
A well-designed form saves more time than ten dashboards could ever recover.

3. Not testing offline

Most rural field sites don't have stable connectivity. Always test your form in airplane mode before deployment. In particular, check:

  1. Media (audio, video) referenced in the form
  2. Cascading lists depending on external files
  3. Local saving of partial submissions

4. Neglecting labels in local languages

If your surveyors work in Hausa, Zarma or Fulani, translate the key questions. It's not just a comfort issue — it's a data quality issue. A misunderstood question produces a random answer.

KoboToolbox natively supports multiple languages via the label::fr, label::ha columns, etc. Official documentation here.

5. Designing the form alone, without a field pilot

Even with fifteen years of experience, I always run at least one pilot with two or three surveyors before deployment. You'll uncover issues you'd never have imagined: an ambiguous word, a question that takes five minutes instead of one, a skip logic that doesn't cover a real-world case.

The right time ratio

For a 30-day deployment, plan 3 to 5 pilot days. It's the best investment you can make.

These five pitfalls aren't exhaustive, but they alone account for 80% of the problems I encounter when asked to fix an ongoing collection. Anticipate them, and you'll save weeks.

For more, I've published a series of video tutorials on these topics on my Data Solution channel.