What one row of the agro-dealer table records
You won't find pesticide prices in Kenya as a public series like maize or fertilizer prices. The official market table behind the county commodity and fertilizer dataset has fertilizer at market level for 29 counties, but nothing for crop protection. That gap is why the platform's agro-input shelf price dataset pulls from another kind of source: Kenyan agro-dealer catalogues. Dealers publish these lists with a pack and a shilling figure for each product.
Four details matter for each product. Pack price is the shelf figure as listed, and pack size is the volume that figure buys. Price per litre is the arithmetic between the two, worked out only where the pack holds a liquid. That is why this slice is called liquid crop protection rather than pesticides in general. A listing count records how many catalogue entries and days stand behind the product's figure. One product can appear on several lists at several prices in the same week. When that happens, the figure served is the median over those listings and days, never the lowest and never the first seen.
Household pest products stay out of this slice. A can of fly spray has a litre price too, but it is not crop protection. Including it would drag the household class into every ranking. Everything here is Kenyan: shilling prices from Kenyan dealers. Nothing in this slice should be read as Tanzanian, Ugandan or Rwandan. Units follow the site convention, weight per kilogram and volume per litre whatever the pack quoted. That puts a small bottle and a large jerrycan on the same scale.
The latest window runs from 12 October 2026 to 25 October 2026 and holds 332 products with a per-litre price. This is a snapshot of the shelf, not a time series of a market. Hold on to that distinction. Most of what follows depends on it.
The latest fortnight, counted
The window breaks out three classes. Insecticides are the largest group at 61 products, with a median of 7,399 shillings per litre and a range from 2,000 to 33,399. Fungicides number 33, median 7,205 per litre, range 1,600 to 22,000. Herbicides number 30, median 3,415 per litre, range 620 to 8,419. Add the three and you get 61 plus 33 plus 30, which is 124 products. The window holds 332, so 332 minus 124 leaves 208 rows. Those rows carry a per-litre price but sit outside the three-class split. Anyone pulling the full table should expect that remainder and decide what to do with it before computing an overall median. The three class medians do not describe it.
The data names the cheapest and dearest products per litre. Among insecticides, Dizon lists at 2,000 shillings per litre, Profile at 2,400 and Mesurol SC at 2,414. At the top, Calypso SC reaches 33,399, Occasion star 31,596 and Panthenon 31,200. Among fungicides, Katerina is cheapest at 1,600, then Decis ec25 at 2,100 and Proplant at 2,650. Absolute star tops the class at 22,000, ahead of Miravis duo at 13,600 and Cabrio at 12,520. Among herbicides, Atlantis maxx OD sits at 620, Roundup 360k at 1,799 and Greenlife governor at 1,905. Integrity is dearest at 8,419, then Butisan gold at 6,454 and Sencor SC at 5,438.
A caution on names. Class labels follow the catalogue listing. If a product looks like it belongs in a different class, check the label before you trust the class median that includes it. A misfiled product moves a 30-product median further than it would move a larger one. A single mispriced entry is damped by the median. A consistent misfiling is not.
The spread is the figure you did not expect
The medians are not the surprise. Insecticides at 7,399 and fungicides at 7,205 per litre land close together. Herbicides sitting lower at 3,415 is what most readers would guess. The surprise is the width of each class. Calypso SC at 33,399 shillings per litre and Dizon at 2,000 are both insecticides in the same fortnight, from the same catalogues, on the same per-litre scale. Fungicides run from Katerina at 1,600 to Absolute star at 22,000. Herbicides, the narrowest class, still run from 620 to 8,419.
That much spread within one class tells you the median is a weak proxy for any single product. Take a lender's cost-of-production template with one insecticide line set at the class median. For most products on the shelf it will miss, and miss in both directions. A median only marks where the middle of the catalogue sits. It says almost nothing about what a particular tomato grower will pay in a particular week. That grower buys a product, not a class.
Put the two gaps next to each other. Between classes, the insecticide median of 7,399 sits against the herbicide median of 3,415. That is a real gap, and it is the one most summaries report. Inside insecticides, the top of the range is 33,399 and the bottom is 2,000. That gap is of a different order, and it is the one most summaries hide. The table supports this opinion: for liquid crop protection, the product column matters more than the class column. Any analysis that collapses to class medians has thrown away the information the dataset actually carries.
Why is the shelf spread so wide? Catalogue listings do not say. The platform does not guess either. Active ingredient, formulation, brand, pack size and dealer margin are all plausible reasons. But none of them is a field in the table, so none can be tested from it. One thing can be said: the per-litre column is honest about the pack and silent about everything else.
This is not a trend. The window is a fortnight in October 2026. A second window would let you say whether the top of the insecticide range moved. One window cannot do that. The platform publishes observations, not extrapolations.
Why a cheap litre is not a cheap treatment
The dataset's own note puts it plainly: a litre of one product is not a dose of another. Concentrations and application rates differ. A cheap litre is not a cheap treatment. Two bottles at the same shelf price can hold different concentrations of active ingredient and carry labels calling for different volumes per hectare and different numbers of sprays per season. Price per litre normalises the pack, not the treatment.
That consequence is sharpest where losses are high. A survey of 141 agro-dealers in Kirinyaga County, where tomato production is predominant, states that the tomato leaf miner Tuta absoluta can cause 80 to 100 percent losses if not properly managed. Against a loss that size, the difference between a cheap litre and a dear litre is small next to the difference between a product that controls the pest and one that does not. Calypso SC, at the top of the insecticide list, can be the cheapest treatment if its label rate is low and it works. Dizon, at the bottom, can be the dearest if it needs repeated sprays or fails. Nothing in the price table decides between those cases.
The same Kirinyaga study is useful for a second reason. Dealers were asked what they would pay for the fungal biopesticide ICIPE 20 before its introduction. Their benchmark was Coragen, described there as the most popular insecticide for managing Tuta absoluta. A higher proportion of dealers were willing to pay above that benchmark. They did not price the newcomer against a per-litre class median. They priced it against the incumbent product for the job. That is how the trade thinks, and it is a better guide to reading this dataset than any ranking by volume. It also means the useful comparison is usually between two or three named products that do the same job. The table supports that well. It supports comparisons between classes badly.
So use the per-litre column honestly: as a first filter. It shows which products are on the shelf, at what pack and price, and how they compare on the one axis the catalogue provides. Treatment cost is a derived figure. It needs a label rate the dataset does not hold and does not estimate.
Per season is the unit the trade compares in
Per litre is only a first filter. The number people actually weigh is spend per cropping season. One Kenyan figure speaks to that, from a CABI-led survey of smallholders in areas of Kenya's Central, Eastern and Rift Valley Provinces, published in Pest Management Science in May 2020. There, 87 percent of farmers said they used chemical pesticides to manage crop pests. Only 10 percent used biopesticides. And 98 percent of those who did not said they would be willing to try them.
The pricing finding is the one that matters here. Farmers surveyed were willing to pay, on average, 9.6 percent above what they currently spend on a chemical pesticide. They would do that for a biopesticide that is just as effective. The study puts that at US$5.7 above current chemical pesticide expenditure per cropping season. Read the unit carefully. This is a premium over season spend, in dollars, among smallholders in three former provinces. It comes several years before the October 2026 catalogue window. You cannot turn it into a per-litre price. It should not be attached to any product in the table. What it establishes is the scale on which buyers reason: a season's outlay, and a tolerable margin above it.
The same survey carries a cost the price column cannot show. Among farmers using chemical pesticides, 42 percent said someone in their household had negative health effects after applying them in the previous 12 months. Skin irritation was reported by 38 percent. Headaches by 28 percent, dizziness by 25 percent and stomach ache by 8 percent. The CABI piece also notes that biopesticides then covered approximately four percent of the global pesticide market. So the shelf the Kenyan catalogue describes is overwhelmingly conventional chemistry. If your programme team models input cost, carry season spend as the target variable. Treat the per-litre table as the price input to it.
Does geography move the price, and can this table tell you?
Readers used to the county commodity table will want a county breakdown. This slice does not carry one. Every figure above is a national Kenyan catalogue figure. Nothing in the window should be assigned to a county. Kirinyaga appears here only because a dealer survey was run there. It says nothing about Kirinyaga shelf prices.
There is reason for caution even where county columns exist. A 2025 study in Frontiers in Artificial Intelligence by researchers at Strathmore University, the African Population and Health Research Center and ICIPE, which modelled Kenyan food prices with machine learning, used an international humanitarian food price series covering January 2006 to September 2024 alongside central bank exchange and inflation rates. Its SHAP feature ranking put unit quantity, price type, commodity and currency ahead of geographical factors such as county. That is food prices, not agrochemicals. A feature ranking is a statement about explained variance rather than about markets. Still, the direction is instructive: check the unit and the price type first, and the county column after.
Where county comparison is possible on this site, it is on the commodity side. The commodity data piece worked through how the official market table reports wholesale and retail per market and day across all 47 counties since May 2021. It also showed how a move in a national figure can come from which counties reported that week, not from a price change. The agro-dealer table has no county mix to worry about. Its composition risk is which products were listed, not which markets reported.
Where the table is thin and what it cannot answer
Three limits deserve naming. Depth first: 30 herbicides and 33 fungicides are small groups. A median of 30 moves when two products change. Time second: one fortnight. Third, a catalogue price is an asking price. Nobody in this dataset has confirmed a sale at the listed figure. Dealer credit, discounting and stock-outs sit outside it.
Those limits belong to a wider measurement problem. FAO's chapter on the economic impacts of transboundary pests and diseases states that data on crop losses from pests are not very reliable in developing countries. That data has generally come from site-specific tests, not systematic government-sponsored research. Lack of record-keeping by farmers adds uncertainty about how much of a production change is due to pests, weather or management. The chapter also warns against multiplying estimated physical loss by the market price, because that exaggerates the likely economic impact. Put the two together, and the value side of any pesticide cost-benefit sum in Kenya is weaker than the cost side this dataset provides. Knowing the price of a litre precisely does not make the loss it averts any better measured.
FAO also notes that outbreaks tend to line up with climatic shifts, drought and early rains among them. After an outbreak, prices can rise where most of the output is eaten at home, or fall where most of it is exported and quarantine blocks the shipment. Anyone reading an agrochemical price alongside a commodity price should keep both mechanisms in mind.
On coverage: an FAO methodology note on pesticide use and trade statistics sets out how net trade data can stretch a use dataset further back in time and across more geography. Trade figures tell you what came into a country. Shelf figures tell you what it costs on arrival. Neither one tells you what was applied to a field.
Pulling the rows
This slice is a single table in a catalogue that covers Kenya, Tanzania, Uganda and Rwanda under one API key. A free account needs no payment details. Confirm the address and the market layer, the full catalogue and the documentation are open to you. Analyses, meaning the answers the platform computes rather than the rows it serves, cost USD 40 per 1,000, with the first 10 each month free. The Claude connector is USD 10 a month per key.
Say you are a developer building a cost-of-production tool. You would take product, pack size and per-litre price, and use the listing count as a confidence flag. A product with one listing on one day is a weaker figure than one with several across the fortnight, even though both are served as medians. Join the per-litre column to a label rate table of your own and you have a treatment cost. Field-level weather and soil rows come from NuaSense's own sensor network, the IoT hardware deployed in fields plus the intelligence layer above it, and they answer only for the devices on the account that owns them. Nothing on the platform predicts. A price forecast was built, tested against carrying today's figure forward, lost, and was not shipped.