Who we are, how we work, and what we do when we are wrong
Harmattan Market Monitor is a cocoa market-intelligence service built around a single proposition: the numbers that move this market are made at origin, and almost none of them are published. We measure them.
Ownership
Harmattan is owned and operated by Melody Research, a research company owned by Melody Global, an investment firm based in Lehi, Utah. Melody Research designs and runs the field panel, holds the respondent data, and is responsible for everything published here. It already works with 225 field staff across 19 countries speaking 43 languages, which makes it an unusual instrument for the developing-world markets where ground truth is hardest to establish and where Harmattan operates. No data source, exchange, trading house or producing-country institution has any ownership interest in Harmattan or any say over what it publishes.
The field panel
Our core asset is a panel of roughly 400 respondents — farmers, purchasing clerks, depot staff, input dealers and hauliers — interviewed repeatedly across the cocoa districts of Ghana, then Côte d'Ivoire and Nigeria. Because the same people are asked the same questions every wave, the series measures change rather than sampling noise.
Interviews are conducted in person by enumerators drawn from BYU–Pathway Worldwide, the online degree programme whose students in Ghana, Côte d'Ivoire and Nigeria live in the communities the panel covers. That is the point: a farmer or purchasing clerk is far more likely to answer a question about late payment or beans held back honestly when it is asked by someone from the same district, in the same language, who will be back next wave. The programme also runs the other way: for the students it is an early-career job in data collection and data science, with training in survey method, geolocation and quality review that is portable to any employer, in a region where such work is rarely available to people at the start of their careers. Each enumerator works under an individual credential, submits from a configured handset, and is assigned respondents by our research staff rather than choosing their own. Every interview carries the enumerator's identity, a timestamp and a location, and our quality engine compares each enumerator's answer distributions with their peers' and each respondent's answers with their own previous waves. Interviews that fail those checks are excluded before anything is computed.
Respondents consent separately to research use and to any sharing of their farm's location, and may withdraw either at any time. A refusal is recorded as a finding, not treated as a gap.
Editorial standards
- Every figure carries its source and date. Where the sourced record is silent, the cell reads "—". We do not model, allocate or interpolate to fill it.
- Estimates and projections are labelled as such, and our own assessments are dated so a reader can see how old the judgement is.
- Licences are enforced, not noted. The build refuses to publish anything a source's terms do not permit. The source ledger shows the status of every input.
- Methodology is pre-registered. Weights, caps and thresholds for the field indices are fixed before data exists and never refit to the data afterwards. See how the indices are built.
- Nothing here is investment advice.
Corrections
When we get something wrong we say so on the page where it appeared, with the date and what changed. Hand-entered figures are the most likely place for an error and the easiest to fix; tell us at hello@harmattanmarketmonitor.com and we will respond within two working days.
Contact
Questions about method, data access or the design-partner programme: hello@harmattanmarketmonitor.com, or request field-data access.