Why Your Survey Design Is Killing Your Data Quality

1779337954085

Picture this.

It’s a Tuesday morning in Lagos. Amara, a field agent, walks into Oshodi with a tablet in hand. She has a target: 30 completed surveys before noon.

The survey? 90 questions. By question 12, the trader she’s interviewing is glancing at his goods. By question 20, he’s not answering with full attention. By question 35, he is saying anything and everything just to get back to selling.

Amara submits the response and moves to the next person.

The data looks complete. But what just got recorded? Not the truth.

Bad data is not a small problem. It is an expensive, invisible one.

Research from Qualtrics found that surveys longer than 9 minutes on a mobile device see substantial respondent drop-off. A separate study analyzing over 267,000 survey responses found that questionnaires with 15 or more questions have a completion rate of just 41.94% meaning more than half of respondents either abandon the survey or stop engaging meaningfully before the end.

That is not a field operations problem. That is a design problem.

And in Africa, where most field research is conducted on mobile devices, in fast-moving environments, with respondents who have limited time and a hundred other things competing for their attention the stakes are even higher.

According to ESOMAR, Africa and the Middle East combined account for just 3% of the global market research industry. That means every study we run, every dataset collected, carries extra weight. We cannot afford to waste it on poorly designed instruments.

So What Exactly Goes Wrong?

Let’s walk through the four most common ways survey design quietly destroys data quality and what to do about each one.

1. The Survey Question That Never Ends

Nobody wakes up excited to answer 47 questions. Yet somehow, research briefs keep producing surveys that read like tax returns.

The reality is simple: after about 10 minutes, people stop thinking and start clicking. They pick the middle option. They type “okay” in the text box. They select “agree” on everything just to move forward. Your data starts reflecting exhaustion, not opinion.

If you are really interested in accurate data then keep your surveys under 10 to 12 minutes. If a question does not directly answer your core research objective, it does not belong in the survey. Trim without mercy.

2. You Are Putting Words in People’s Mouths

Read these two questions carefully:

“How satisfied are you with our fast and reliable delivery service?”

“How would you describe your experience with our delivery time?”

Same topic. Completely different answers. The first question tells the respondent what to think before they have thought anything. It makes disagreement feel socially awkward. The result is inflated satisfaction scores that make your product team happy and your strategy team misled.

Keep your assumptions out of the wording. If the delivery is slow, let the respondent be the one to say so. That is the entire point of asking.

3. Designed for an Office, Deployed in the Field

There is a version of survey design that works perfectly on a laptop in a quiet office. That same survey falls apart the moment it reaches a market trader in Onitsha, a smallholder farmer in Plateau State, or a logistics worker at a depot in Apapa.

Asking someone who is giving out three-paragraph responses on an agent is going to type on a mobile screen is not research. It is wishful thinking. They will type “good” or “fine” or “no comment” and that will mean: please let me get back to my day.

Match the question format to the field context. Use simple single-choice or multiple-choice options in high-distraction environments. Save the open-ended, exploratory questions for structured one-on-one interviews where the setting is built for depth.

4. The Absence of Guardrails

If your data collection tool allows a respondent to say they are under 18 and also that they have three children and a car loan your design has failed technically, not just conceptually.

If someone indicates they do not own a smartphone in question 3, there is no reason they should be answering questions 4 through 10 about smartphone app preferences. Without skip logic, they will answer those questions anyway and your dataset will contain contradictions that no amount of cleaning can fully fix.

Build the rules into the technology. Validated input formats. Automatic skip routing. Logical consistency checks. These are not nice-to-haves. They are what separates clean, usable data from an expensive spreadsheet of noise.

Here’s What Better Looks Like

High-quality data does not begin in the field. It begins the moment you draft your first question.

The good news is that most of these problems are preventable not through more budget or bigger field teams, but through better design decisions made earlier in the process.

Start by setting a strict time limit for your survey and working backwards from it. Pressure-test every question by asking whether it directly answers your core objective. Have someone outside your team read the questions and flag anything that feels leading or confusing. And choose a data collection platform that has the technical guardrails built in  one that catches the errors your spreadsheet will not.

At WeCollect, this is exactly what we have built. Our platform includes dynamic skip logic, geofencing, automated constraint checks, and field-context-aware question formats so the instrument you design translates faithfully into the data you actually need, regardless of where in Nigeria or Ghana your field team is working.

Your research is only as strong as its foundation. Make sure the foundation holds.

What is the worst data quality issue you have had to deal with after a field exercise? Drop it in the comments. I would genuinely like to know what caused it.

Leave a Comment

Your email address will not be published. Required fields are marked *

Save to list

Scroll to Top