7 measured answers about Kenyan crop markets

These 7 read the crop record, where the other groups read fertilizer, and every answer is an observation or arithmetic on observations. None of them predicts a price, because the version that tried lost to the rule of carrying the current figure forward and was never deployed.

7 questions under /api/partner/v1/market. Every one is priced per answer, at 4¢ a call.

The 7 questions

QuestionWhat comes backCallFiltersPrice
Has the season started here — early or late, and is the rain that starts one coming? Eight years of satellite greenness against this unit's own normal. Also covers TZ, UG and RW. Kenya answers add rain_outlook: expected rain for the next 7 and 15 days with a wet/dry band, whether planting-window rain is unlikely/possible/likely, the expected dry spell and heat days — derived from a 64-member ensemble forecast (Google DeepMind WeatherNext 3, experimental; the response carries the required attribution). /analytics/season county, country, index 4¢
How reliable is this county's harvest, and which years failed? Mean, variability and the actual failure years since 1997. The insurance and credit question. /analytics/yield-risk county, product, season 4¢
What does coffee grade AA earn over the market? Premium over the whole-auction average, in percent, per sale. /analytics/coffee-grades sales 4¢
When Nairobi moves, how long until my market follows? Measured on price changes, not levels. Only lags with correlation ≥ 0.2 mean anything. /analytics/price-lag commodity, reference, window_days, max_lag_days 4¢
Where can I buy cheap and sell dear — and does it cover the lorry? Price gaps between markets, checked against your own transport cost. Pairs flagged likely_grade_mismatch are two different grades under one name, not a trade. /analytics/spread commodity, window_days, transport_cost_kes_per_kg, limit 4¢
Which counties are getting expensive fastest? Year-on-year per county against the official KNBS national rate where one exists. /analytics/inflation commodity, window_days 4¢
Which KTDA factory beats its neighbours? Ranked on factory_effect, not the raw premium — 86-89 % of the raw spread is which side of the Rift Valley the factory sits on. /analytics/tea-league grade, sales, min_sales 4¢

Buying cheap and selling dear, with the lorry included

The arbitrage spread call finds the gap between markets for the same product and then does the part most spread calculations skip: it checks whether the gap survives the cost of getting there. A wide gap between 2 markets at opposite ends of the country is not an opportunity, it is a freight bill. The answer reports the gap and the distance together, so the second half of the question is never left implied.

How a price travels

Price transmission is the question of what a move in the reference market means for a market 2 counties away, and when. The answer gives a lag in days and a strength, both measured against the historical record rather than assumed. Some markets follow closely within a day. Others take weeks, and a few barely follow at all, which usually means different ground supplies them.

Note what this is not. A lag that held for 3 years is a description of the past, and reading it as a prediction of next month is the mistake it most invites. The figures behind this call, including a worked example, sit on the price intelligence page.

County inflation

This ranks counties by how fast a basket of staples is getting more expensive. It is the call behind a question a programme officer asks in a different form: where is the cost of eating rising fastest right now.

Harvest reliability

This call predicts no yield. It describes how dependable a county’s harvest has been: the spread across recorded seasons and which specific years failed. A lender pricing risk over a district gets more from that than from a single-number projection, and it has the advantage of being checkable.

Season onset

Has the season started here, is it early or late against the usual, and is the rain that would start one on the way. The first 2 parts are measured from the rainfall record. The third is an outlook, it is labelled as one in the reply, and it carries its own attribution.

Coffee and tea

2 calls are specific to the auction floor. One reports what a given coffee grade earns over the market as a whole, which is the premium a washing station captures at auction. The other places a tea factory against its neighbours and says whether it is beating them. Both read auction records, and neither attributes the difference to a cause.

The data underneath

Every answer on this page is computed on the datasets below, and each of them can also be pulled row by row with the same key if you would rather do the arithmetic yourself.

Questions not on this list

Spread and transmission for the exact pair of markets you trade between, harvest reliability narrowed to the wards you lend into, or a factory comparison for a crop the auction record covers and the league does not. Work like that is scoped in a conversation and quoted; the per-answer rate does not apply to it.

Ask on request Create a free account