How Wyza works โ€” the honest details.

Research is only as good as the people, process, and platform behind it. Here's exactly how Wyza recruits, runs, and reports. No hand-waving.

Our panel

Wyza's African consumer panel is our core asset. We recruit through mobile-first channels across every country we operate in, retain members through fair compensation and relevant participation, and continuously validate them for authenticity and quality.

Nigeria landmark3 major languages5k+ users36 states

Nigeria

[N] active panel members

Kenya landmark3 major languages5k+ users36 states

Kenya

[N] active panel members

Ghana landmark3 major languages5k+ users36 states

Ghana

[N] active panel members

*Expand as markets grow

We publish active panel numbers quarterly. If a number isn't on this page, it isn't claimed.

What we don't do:

Rent panels from third-party aggregators without vetting.

Claim coverage in markets where we haven't activated.

Inflate respondent counts by including inactive accounts.

Quality controls

Every study runs through the same checks โ€” verified identities, response-quality scoring, and human review โ€” so the data you act on is data you can trust.

Identity verification

Every panel member is verified through phone number, device signals, and periodic re-validation.

Response quality scoring

Every response is evaluated on engagement depth, answer coherence, and known fraud signals. Low-quality responses are flagged and not paid out.

Attention checks

Studies include embedded attention checks. Respondents who fail are excluded from analysis

Panelist history

Every panelist has a score based on historical response quality. Studies can be gated to high-score panelists only.

AI moderation

Our AI conducts interviews in voice, video, and text โ€” in 40+ languages including Pidgin, Swahili, Yoruba, Hausa, Igbo, Amharic, Kinyarwanda, Zulu, and Afrikaans.

Unlike translation-layered tools, Wyza's AI is trained to moderate natively in each supported language. It asks follow-up questions that a human moderator would ask โ€” probing for depth, clarifying ambiguity, and adapting to the respondent's actual answers.

All AI-moderated interviews are reviewed by human quality analysts before delivery. If the AI missed something or went off course, we catch it before you see the report.

Wyza AI-moderated interview preview screen

Field agents

For offline communities, rural populations, and contexts where digital recruitment falls short, Wyza deploys a network of trained field agents. Agents conduct interviews in person using our mobile app โ€” uploading recordings, transcriptions, and metadata directly into the same analysis pipeline as our digital studies.

Agents are trained on ethical recruitment, informed consent, and quality standards. All agent-collected data is subject to the same quality scoring as digital data.

A field agent conducting an in-person interview

Analysis and delivery

Once fieldwork completes, Wyza's analysis layer processes responses across text, audio, and video. Themes are surfaced with statistical weight. Quotes are flagged by salience. Sentiment is scored at response and topic level.

Every study is reviewed by a human analyst before delivery. What you receive is not raw AI output โ€” it's an analyst-reviewed deliverable with the AI layer doing the heavy lifting underneath.

Standard deliverables: executive summary, full report, data exports (CSV, XLSX, SPSS), slide deck (PPTX), and highlight reel (video).

An analyst reviewing data visualisations

Data and privacy

Respondent data is collected with explicit informed consent. We comply with GDPR, Nigeria's NDPR, and equivalent regulations in every market we operate. Respondent PII is never sold, shared with third parties, or exposed to clients beyond anonymized research outputs.

For details, see Security & Trust
Two people reviewing documents beside a set of scales

Download full methodology

For enterprise buyers, RFPs, and research-minded partners: our full methodology document goes deeper than this page.

Download full methodology (PDF)