The price frame misses the actual decision structure
When fertilizer affordability comes up in Indonesian agricultural policy discussions, the conversation usually lands on price: the HET (Harga Eceran Tertinggi) for subsidized urea, the markup from kios resellers, the seasonal price spikes before planting windows. These are real issues. But they are not the whole problem, and in some cases they are not the primary problem for smallholder farmers who want to get a better return from their inputs.
The more common issue we observe is not that farmers cannot afford fertilizer. It is that they spend more on it than they need to, because the recommendation they follow is generic, and generic advice for rice in West Java covers the worst-case scenario, not the average case and certainly not the best case for a well-managed plot with good soil structure.
Fertilizer waste and fertilizer affordability are the same problem viewed from different angles. If you apply 230 kg/ha of urea because that is what the input dealer recommends, and your plot's nitrogen uptake capacity only supports meaningful response up to 180 kg/ha, the remaining 50 kg is cost with no yield return. Multiply that by IDR 2,400 per kilogram for non-subsidized urea and a 1 ha plot, and you are looking at roughly IDR 120,000 per season per hectare in direct waste before any secondary costs like labor for application.
Why generic advice over-estimates the dose
Agronomic recommendations for fertilizer dosing are typically built around yield potential under favorable conditions. The logic is: if a farmer follows the national recommendation and has reasonable soil, they should hit the target yield. This means the recommendation is calibrated for the average-to-good case and pads slightly for safety.
The problem is that smallholder plots in West Java are not a uniform population. They range from well-irrigated Andosol plots in Cianjur that respond strongly to additional nitrogen, to compacted clay-heavy Latosol plots in parts of Karawang that have poor water retention and hit nitrogen saturation at significantly lower doses. A single recommendation cannot serve both well. It either under-doses the Cianjur farmer or over-doses the Karawang farmer.
Extension programs know this. The standard answer is "get a soil test." But a soil test at a provincial laboratory costs between IDR 150,000 and IDR 300,000 per sample, takes one to three weeks, and requires the farmer to know which parameters to test for. For a farmer with a 0.8 ha plot earning IDR 3 to 5 million per season, this is not a practical step before every planting cycle.
The data route that does not require a lab
The alternative we are building does not replace the soil lab. It asks a different question: what do we already know about this plot and the plots around it that reduces the uncertainty enough to give a better recommendation than the generic national dose?
In practice, this means combining three things we can observe or estimate without a physical sample. First, the soil-class map for the plot's location, which gives us a baseline nitrogen-holding capacity range. Second, the farmer's prior-season yield record, which tells us whether the last applied dose produced a yield consistent with what we would expect, or whether there is evidence of over- or under-application. Third, the rainfall timing during the last growing period, which tells us how much nitrogen was likely lost to leaching between application and uptake.
None of these three signals is as precise as a lab result. Together, they are precise enough to shift the recommendation meaningfully for most plots. In our early-access work across Subang and Karawang, we found that a substantial share of farmers we worked with were applying urea at doses above what our model estimated their plots required. For those farmers, the data-informed recommendation was a cost reduction, not an agronomic compromise.
What this does not solve
We are not saying that price policy is irrelevant. For farmers near the income floor, even a correctly sized input purchase can be a cash-flow problem if the timing requires payment before the harvest sale. The credit access gap in Indonesian smallholder agriculture is real and will not be fixed by better dosing recommendations.
We are also not saying that our model eliminates the need for physical soil testing. It narrows the uncertainty enough to give a better default recommendation. For plots with unusual histories, unusual soil profiles, or sharp performance outliers, a lab test is still the more reliable route to a precise answer.
What we are saying is that for the median smallholder rice farmer in the regencies we cover, the fertilizer affordability problem is partly a data problem, and the data problem is solvable at a cost that is proportional to what the farmer spends on inputs in a season. That is the problem Elevarm is working on, and the dosing recommendation is the place where the work becomes concrete.
Where the Mitra dealer fits in
A useful side effect of plot-specific dosing recommendations is that they give Mitra dealers a better basis for the conversations they have with farmer customers. Instead of a dealer recommending "take 200 kg urea per hectare" based on general habit, the Mitra dashboard surfaces the Elevarm recommendation for that specific farmer's plot and variety. The dealer who follows through on that specificity builds a relationship that is harder for a competitor to displace on price alone.
This is not a small thing in the context of rural input distribution. The traveling rep model, where a dealer builds relationships by showing up regularly and offering credit, is under pressure as more farmers have phone access and can compare prices. The dealers who will stay relevant are the ones who can offer advice that is actually calibrated to the farmer's situation, not just to the truck's inventory.