Farmer interviews by WhatsApp, in English and Swahili

You send the questions as a spreadsheet and the list of farmers. We put the questions to them and hand back a table, not a folder of recordings. Every answer carries a label that says how it was captured.

Priceon request
Unitper completed interview
WhereKenya, Tanzania, Uganda, Rwanda

The deliverable

3 labels instead of a number

Every answer comes back with one of 3 labels. Tapped from options means the farmer chose from a list: nothing to interpret. Read from a typed reply means a rule matched a number, a date or a yes. Interpreted and reviewed means free text or a voice note was read by software and then checked by a person before it reached your table.

There is no confidence score anywhere in that. A decimal implies calibration: 0.85 ought to mean that 85 of every 100 such answers are right. Nothing in this product has been measured against a hand-labelled set, so the decimal would be invented. An auditor would be right to say so. A certifier can act on the 3 labels; nobody can act on a fabricated probability.

2 channels, and the stronger one

The WhatsApp conversation is the one people ask for. It works best on a warm list: the exporter’s own growers, told in advance by the exporter that we will be in touch. Sent cold, it performs the way any cold contact performs.

The stronger channel is quieter. Your field officer opens a link while sitting with the farmer and works through the questions together. That is an interview with a person present and a device that enforces the structure. Among the farmers visited it reaches nearly all of them. If you already run field staff, plan around that channel. The data arrives in the same table as the WhatsApp answers, with the capture label saying which route it took.

The record as the product

The satellite half of a validation study checks land. This half is a record. A structured, timestamped, per-farmer answer traceable to one conversation is a different object from a clipboard and a field officer’s memory of last Tuesday. It is the half that survives being audited. Where a study in Kenya pays the people who answer, the respondent incentives run through us as well, with a signed statement for your finance team.

Your list stays yours

You remain the controller of the farmer list you share. We process it under a written agreement signed with you. That agreement is separate from our own registration with the Office of the Data Protection Commissioner in Kenya: the registration covers us as a controller of our own data, the agreement covers what we may do with yours. The list is never re-used for anything else and never sold.

Limits of the evidence

  • No confidence score on any answer. Nothing here has been calibrated against a labelled set, so a decimal would be invented; you get the capture label instead, which is true.
  • No completion rate quoted as fact before your study has run. Published figures for other countries and other modes are not a promise about your list.

Work still in development

None of this is available yet, and none of it is quoted until it has run on a real study.

Voice interviews
The same questionnaire conducted as a spoken conversation, for farmers who do not use WhatsApp. The interview engine already treats the channel as swappable, so a call will arrive on the same path as everything else. It is not available, and we do not quote for it until it has run.

Scoping and quote

To quote an interview study we need the number of farmers, the questions and what kind of answer each expects, the channel, and the languages. A free-text question costs more than a tapped one, because a person reviews every answer that had to be interpreted.

Email info@nuasense.com WhatsApp +254 707 523 321