Source.ag: greenhouse forecasts are supply signals, not binding offers

External-source research synthesis, not an OfferSignal client result. Based on the retained October 2026 research and source-specific dates.

A greenhouse grower can have extensive plant, climate and cultivation data and still struggle to answer a commercial question: how much product will be ready, when, and with what confidence? A wrong harvest expectation affects labor, packaging, logistics and commitments to retailers. Source.ag’s case is useful because it connects operational data to that planning problem, while showing why a prediction is not automatically a quote or a purchase decision.

From plant observations to supply planning

Source.ag, founded in Amsterdam in 2020, describes a platform for commercial greenhouse operators, cooperatives and connected supply-chain partners. Plant App digitizes observations; Workspace centralizes operational data; Cloud connects cultivation information to systems such as Power BI. Inputs can include sensors, cameras, climate-control equipment and resource-price records. Harvest Forecast produces daily yield predictions up to eight weeks ahead, while Cultivation Management lets growers simulate variables such as stem density, pruning, temperature and lighting before changing a strategy. 1 3 4

That creates a practical B2B/B2B2C workflow: a grower can use an expected weekly volume to plan labor, packaging and transport, and to inform a sales or retail discussion. The forecast remains evidence for an indicative availability signal. The retained record does not show autonomous purchasing, quoting, contracting or binding customer-service authority. Grade, quantity, delivery window, price, quality and capacity still need the responsible commercial and operations owners.

The control layer is more consequential. Source Irrigation Control combines historical data, predictive analytics and real-time monitoring to generate climate-computer settings against grower-defined irrigation targets. In a provider-reported pilot, dry-down targets were reached on 87% of days with the system versus 37% without it. The greenhouse count, crop, comparison period and causal design were not disclosed, so this is a pilot signal, not an ROI estimate. 5

In October 2026, Source.ag announced the acquisition of Optimal, extending the stated roadmap to heating, ventilation, screens, lighting, carbon dioxide and irrigation through existing hardware. The announcement said first autonomous deployments at Source customers in Europe and North America would go live in winter 2026–27. That was announced future availability at the 27 October research cutoff, not evidence that those deployments were already live. 8

What the reported forecast results establish

Source.ag reported that a next-generation tomato model, field-tested with a small group before phased rollout in March 2026, improved mean accuracy at the three-week horizon by 33% versus the previous model. Forecasts more than 20% off target fell 25%, and severe outliers fell 50%. Absolute error, sample size, outlier definitions and independent validation were not published. 6

A commercial pepper forecast launched in April 2026 for six varieties. Source.ag reported that pilot growers Harvest House and Jansen Paprika saw average three-week forecast error fall by more than 40%, with wide-miss weeks 28% lower. The workflow required weekly entry of rounds and percentage-coloured data alongside plant, climate and cultivation-strategy information. The grower statements connect the signal to retail alignment and replacing manual Excel counting, but do not establish realized profit, waste reduction or price improvement. 7

Optimal’s acquisition announcement separately claimed 13% more yield and 27% less energy than a benchmark Dutch grower in a commercial greenhouse. The customer, crop, period, baseline method and independent reporter were not named. Treat this as a company-reported acquisition claim, not a general performance guarantee. 8

An illustrative food-supply pilot

For an illustrative greenhouse cooperative, start with one crop and a matched block or phased rollout. Record greenhouse, block, variety, units, timestamps, cultivation-plan revisions, sensor provenance and user permissions. Validate freshness and duplicates before evaluating the model. Reconcile each forecast with actual harvest, preserve the model and input versions, and report confidence bands rather than one unqualified number.

Let the grower or agronomist own cultivation targets and high-impact control changes. Let sales and operations approve any retailer commitment. Use manual override, rate limits, allowlisted hardware actions, dual approval for consequential changes and an audit trail of recommendations and overrides. Measure absolute forecast error, wide misses, target attainment, labor hours, energy and water, waste, contract fulfillment and margin. Public pricing, integration terms and ROI/payback were not disclosed; Plant App is described as no-extra-cost within the suite, but commercial access and tier prices require vendor contact. 4 11

Sources and next reading

• Source.ag main site / demo: public website; product access and pricing require vendor contact/demo.
• Source.ag product tiers: public product overview; no public prices.
• Source Irrigation Control: public product and pilot description; operational access is commercial.
• Source Plant App: public product page; deployment terms are not fully public.
• Source.ag FAQ: public FAQ; detailed API and integration answers require confirmation.
• Optimal: public acquired-company site; post-acquisition availability is not publicly priced.

Explore OfferSignal · Focused Business & AI Workshop · All research

References