We have had the same ROI conversation with eight utility operations teams now. It goes roughly like this: the team understands the technology, they have a feeder problem, and they are trying to figure out whether the business case supports procurement. The number that unlocks the conversation is almost always the same: how much are you spending on capacity reservation to cover forecast uncertainty?
The cost of forecast uncertainty
Distribution utilities carry reserve capacity margins that are sized around their worst-case forecast error. When your 4-hour ahead feeder load forecast is accurate to plus or minus 18 percent, you build reserve capacity to absorb that range. When the forecast is accurate to plus or minus 6 percent, the reserve requirement compresses. The delta between those two states is real dollars: capacity reservation contracts, demand response activation costs, and the capital cost of infrastructure that gets built to cover a risk that better forecasting would have reduced.
For a utility with 50 to 150 feeders, that margin compression is often worth $400,000 to $1.2 million annually, depending on feeder load profiles and regional capacity market rates. We have seen numbers on both ends of that range in our pilot data.
How the pilot phase actually works
Every Ampgrove engagement starts with a bounded pilot covering 5 to 15 feeders over 60 to 90 days. The pilot has two goals: establish a measured forecast accuracy baseline on your feeders, and quantify what that accuracy improvement is worth in your specific cost structure.
We do not ask you to change your operational procedures during the pilot. Operators keep doing what they do. We run shadow forecasts alongside your existing process and compare outputs. At the end of the pilot, you have a document that says: on feeder group X, your current forecast error is Y percent, our model achieved Z percent error, and the implied capacity reservation savings is N dollars annually. That document either supports a production deployment or it does not.
What moves the room
The capacity reservation number is almost always the number that moves the room. Reliability improvements and operator time savings are real, but they are harder to put in a procurement justification. A specific dollar savings on capacity reservation is a number the CFO can evaluate directly.
The second number that matters is integration cost and timeline. Every utility we talk to has had an experience with a software vendor that promised a 3-month integration and delivered something unfinished 18 months later. Our 2-week integration claim is testable during the pilot. If we get your SCADA or EMS data flowing and the model running in two weeks, that validates the timeline claim before you commit to production.
The honest version of the ROI case
Not every utility will see the same numbers. Feeders with relatively stable, predictable load profiles get less benefit from ML forecasting than feeders with high EV penetration or significant distributed generation. We tell prospects this up front. If your load profile is straightforward and your current forecasting tools are handling it adequately, Ampgrove may not be the right tool for you right now.
We have passed on a few opportunities where the pilot data showed modest accuracy improvement but insufficient volume to justify the cost. That is the right outcome for everyone. We would rather build a reference customer who gets real results than sign a contract where the economics barely pencil.
If you are evaluating the business case and want to walk through the capacity reservation calculation for your specific situation, contact us directly. It is a 30-minute conversation that usually gives you the number you need to decide whether a pilot is worth doing.
What a production contract actually looks like
After a successful pilot, the production engagement is a subscription to the Ampgrove platform sized by feeder count, with the first year's pricing fixed at the rate we quoted before the pilot. We do not re-price after a good pilot. The pricing you see on our pricing page is the pricing. Utilities that go to production typically start with the feeders covered in the pilot and add adjacent service territory in the first 12 months as operator confidence grows.
The contract term is 12 months, not multi-year. We think that is the right structure for a bootstrapped company that needs to earn continued business. It also matches how utility procurement tends to work at this budget level: annual software subscriptions go through a different approval channel than multi-year commitments, and getting through that channel faster is worth more to both parties than the slight certainty premium from a longer contract.
Implementation is included in the subscription cost. We do not have a separate professional services engagement for integration and onboarding. The integration work we did during the pilot continues directly into production. This matters for the total cost of ownership calculation.
Measuring ROI after production deployment
After 6 months of production operation, we review the financial performance against the pilot estimate with our utility partners. The review covers three numbers: actual capacity reservation cost reduction (comparing the 6-month period against the same period prior year, adjusted for weather), operator time spent on peak period load management, and the count of near-overload events that were addressed proactively versus reactively.
In our three current production deployments, the capacity reservation savings have tracked within 20 percent of the pilot estimate in each case. One deployment came in above the estimate because a feeder experienced higher-than-projected EV charging growth during the period, and the forecast-enabled switching operations that kept it in bounds avoided what would have been a costly demand charge event. Two came in within 10 percent of the estimate.
We share this review openly because we believe our long-term business depends on the ROI actually materializing for our utility partners. A utility that is not seeing real savings has no reason to renew, and no reason to refer us to the next utility. The numbers need to work in practice, not just in a pilot estimate.
The numbers we publish and why
We publish our pilot accuracy metrics and the capacity reservation savings range on our website. The 94 percent forecast accuracy figure and the $400,000 to $1.2 million annual savings range come from our internal pilot data across three early-access utility partners. We include those qualifications because the numbers are only useful to you if you know their source and limitations.
We do not publish the names of our pilot utilities. All three are in early-stage enterprise software relationships where they have not yet completed internal processes for public reference approval. We will begin publishing named case studies as those approvals come through. In the meantime, we are willing to arrange direct introductions to our pilot contacts for utilities that want to speak peer-to-peer about the experience. That is a more valuable signal than a published testimonial anyway.