The High Cost of Misreading Cocoa’s Numbers
Cocoa does not lack data. It lacks agreement on what much of that data means.
At a preview of the 2025 European Cocoa Association Forum, Christian Vollers, managing director of Vollers Group, described cocoa data as either ‘gold or dust’. The phrase captures a market highly capable of collecting information, but far less adept at connecting it.
Futures, crop forecasts, certified stocks, grindings and producer prices generate a constant flow of figures. Companies also assemble farm registries, supply-chain maps and data on deforestation, livelihoods, child labour and carbon.
More information ought to reduce uncertainty. Poorly connected information can instead encourage confident decisions based on measures that describe only part of the market, affecting margins, working capital, investment and supply.
One Market, Many Prices
No figure illustrates the problem better than ‘the cocoa price’. Futures are visible and influential, but they are neither the physical price of beans nor an automatic measure of what farmers receive. Origin differentials, quality, freight, finance, taxes, premiums, contract timing and national marketing systems all intervene.
A rise in futures may coexist with a smaller or delayed increase in farmgate income. Buyers can experience the same movement differently depending on hedging, inventories and contract dates. The useful question is not whether cocoa prices rose, but which price rose, for whom and over what period.
Numbers Need Context
Exchange-certified stocks measure cocoa available within a particular delivery system, not every tonne held in commercial inventories. Falling stocks may signal tight deliverable supply, but may also reflect financing incentives, quality constraints or the use of non-certified storage. The figure can be accurate while the story attached to it is wrong.
Quarterly grindings show beans being processed into cocoa liquor, butter and powder. They indicate industrial activity, but are routinely treated as a synonym for consumer demand. Processors can grind more while building product stocks; grindings can fall while retail demand remains relatively resilient. Facing high bean prices, manufacturers may change recipes, reduce pack sizes or draw down inventories before consumers visibly buy less chocolate.
Grindings must therefore be read alongside trade flows, product stocks, processing margins and retail evidence. Distinguishing weaker processing economics from weaker consumption affects decisions about capacity, procurement and pricing.
Precision Without Certainty
Supply estimates contain a different kind of uncertainty. The equation – production plus beginning stocks, less consumption, equals ending stocks – looks exact. Its inputs are not.
Crop forecasts depend on weather, disease, yields, quality, harvest timing and incomplete reporting. Cross-border trade can blur origin data. Demand is inferred from processing and downstream markets, while stock estimates may include modelled components.
Revisions show that new information has changed the market’s understanding; they are not necessarily admissions of failure. An intelligence system should preserve them rather than overwrite them, because beliefs held at the time shaped prices, purchasing and investment.
Disagreement is valuable, too. Different data or assumptions about disease, informal trade and the mid-crop can explain divergent crop estimates. The gap between credible estimates is itself a measure of uncertainty.
When Definitions Fail
Sustainability data poses an even harder problem: common words with uncommon meanings.
‘Mapped’, ‘traceable’ and ‘verified’ can describe different levels of evidence. A sustainability percentage may cover total purchases, direct suppliers or only cocoa enrolled in a programme. “No deforestation detected” is bounded by a dataset, methodology, geography and cut-off date; it is not necessarily proof that no deforestation occurred.
A farmer ‘reached’ might have attended training, received an input, completed a survey or entered a monitoring programme. None automatically demonstrates higher income. Mapping establishes geographic information, but does not prove that beans remained physically traceable through every transaction.
This is how good data becomes dust. Futures become ‘the cocoa price’; certified inventories become ‘global stocks’; grindings become ‘demand’; mapping becomes ‘traceability’; and training becomes ‘impact’.. The figures become misleading when their boundaries disappear.
Misreading supply can distort hedging and working-capital decisions. Confusing activity with outcomes can favour reportable coverage over resilience. Weak traceability can threaten market access or raise compliance costs. Overestimating farm economics can lead to underinvestment in future supply.
From Data Feed to Intelligence
This is the premise behind the CocoaRadar Intelligence Engine, due to launch in September. Its purpose is to record provenance, preserve revisions, distinguish observations from estimates and stakeholder claims, compare definitions, and connect evidence across market, regulatory and sustainability data.
A data feed reports that a number changed. Intelligence asks why, how much confidence to place in it, what evidence supports it and who bears the consequences.
Data can indeed be gold. But volume alone does not create value. Definitions must survive comparison, uncertainty must remain visible and separate signals must be connected. In cocoa, the most expensive mistake may be trusting a precise number before asking what, exactly, it measures.
A fuller essay examines CocoaRadar’s approach to data, evidence and uncertainty in more detail and is available to download here.
When Cocoa Data Is Gold - and When It Becomes Dust
Further reading:
Anthony Myers