Can you trust app-based research in Ghana? How we verify players and weight results

Can you trust app-based research in Ghana? How we verify players and weight results

If you've ever bought research from a mobile app or online panel, you've probably asked two questions:

1. Are these real people, answering honestly?
2. Do they look like Ghana, or just the kind of people who download apps?

Both are fair. This post explains what An Ask A Day (AAAD) does about each, and where the limits are.

The short version

• Every player has a verified phone number, and one phone can hold only one verified account.
• Every answer goes through quality checks, and each player carries a trust score that research results can filter on.
• Results can be weighted to Ghana's 2021 census by region, sex and age band.
• Brands and researchers only ever receive combined results, never names, phone numbers or emails.
• We're open about what app-based research can't do. That's the last section of this post.

Part 1: Are these real people?

Phone verification

Players verify their phone with a 6-digit SMS code. One verified account per phone number. If a player changes their number, verification is cleared and they verify again.

This matters because most low-quality research comes from people who create many accounts. A verified phone makes that much harder.

Profiles that can't be quietly rewritten

Players give us their age band, sex, region, town, education, employment and relationship status. Two details are locked once set: date of birth and sex. Players can't change them, because they drive age checks and the research data itself.

Changes to a player's name, phone number and username are logged, so a hijacked or suspicious account can be spotted.

A trust score for every player

Each player has a trust score from 0 to 100 and a tier: trusted, standard, watch or low. It's recalculated every night. New players start at 70.

The score draws on signals such as:

• Check questions. Some side questions quietly compare an answer with the player's profile (for example region or age) or with what they told us earlier about stable facts.
• Attention checks in surveys, such as an instruction to choose a specific answer.
• Rushed answers, meaning faster than anyone could read the question.
• Straight-lining in grids, and low-effort or copied written answers.
• Two accounts on one phone.

Three things are worth knowing:

• Nothing happens automatically. The score is a tool for our team. It is used in reviews, in report filters and in who can take certain surveys.
• Honest life changes aren't punished. If something about a player has changed, such as moving region, they can tell us.
• Players never see their score or which answer is "expected", so there's nothing to aim at.

When you buy research, survey results and trackers can leave out low-trust answers, and a survey can be limited to trusted players only.

Brand studies have extra protection

Our brand-lift studies include one made-up brand to catch people who tick everything, and response times are recorded at every step so rushed answers can be excluded. You can read how a brand-lift report works in our guide to reading a brand lift report.

Part 2: Do they look like Ghana?

Honestly, not on their own. Like most apps, our raw player base skews younger and towards Greater Accra. If we reported raw percentages and called them "what Ghana thinks", we'd be misleading you.

So we weight.

How census weighting works

We weight results to Ghana's 2021 Population and Housing Census (Ghana Statistical Service), by region, sex and age band, for everyone aged 13 and over. We use raking, the standard method pollsters use to make a sample match known population figures.

In plain terms: if Greater Accra players are over-represented in the answers, each of them counts for a little less. If a region is under-represented, its players count for a little more. Weights are capped, so no small group can dominate a result.

Why we show the "effective sample"

Weighting has a cost. When some voices count for more, the result is worth fewer respondents than the raw number suggests. So reports show the effective sample: how many respondents the weighted result is actually worth.

Reports also tell you which groups had nobody in the sample. That lets you see where a weighted result is on firm ground and where it's stretched.

This is what allows a client to say "Ghana thinks X" instead of "app users think X", as long as the effective sample supports it.

A note on brand-lift studies: their comparison groups are matched by region, sex and age so that "saw" and "didn't see" players can be fairly compared. That is a different job from weighting a survey to the census.

Part 3: What brands and researchers can see

Partners and clients receive combined results only: counts, shares and breakdowns. They never receive names, phone numbers, emails or individual answers. Click tracking in offers uses a random code, not the player's identity.

Players under 18 are protected from adult-category advertising such as betting and alcohol, and the minimum age to play is 13. Players can hide their username, edit their details, or delete their account and data themselves.

Part 4: What app-based research can't do

No method is perfect, and a research partner should tell you where the edges are.

• Players are phone owners with app access. People without a smartphone or reliable data are harder to reach, even with weighting.
• Weighting fixes the mix, not everything. It corrects for region, sex and age. It can't correct for other ways app users might differ from the wider population.
• Small groups give small answers. A weighted result from a tiny sample can look precise and still be fragile. That's why we show effective samples and flag groups that are too small.
• It isn't a replacement for every method. For a nationally representative household survey, face-to-face fieldwork is still the benchmark. App-based research is strongest for fast, affordable, quality-checked answers, trends over time, and measuring advertising.

What this means when you buy research

Before you commission any research in Ghana, ask your provider:

1. How do you know each respondent is a real, unique person?
2. What do you do with rushed or inattentive answers?
3. Is the result weighted, to what, and how many respondents is it really worth?
4. What do I get to see, and what stays private?
5. What are the limits of this method?

You can read our answers on the methodology page.

Ready to ask Ghana a question?

We run sponsored questions, targeted surveys, monthly trackers and custom reports for brands, agencies, NGOs and public-sector teams. Tell us what you want to learn and we'll suggest the simplest way to find out.

• Research enquiries: research@anaskaday.com
• Partnerships and advertising: support@anaskaday.com
• Request a custom report: anaskaday.com/custom-reports

An Ask A Day. Your opinion, worth something. Made in Ghana.