A controlled three-site pilot offers complete visibility, while fleet growth introduces operational uncertainty. The operator knows each site's equipment configuration, export limit, inverter behavior, and battery chemistry. Dispatch decisions use direct knowledge of every site's state, and all three sites can be checked before the morning dispatch call.
Moving from three sites to fifteen changes the operation qualitatively, not merely quantitatively. Coordination requirements, information management, and failure modes all change. Successful operators identify where pilot-era processes become insufficient and redesign them before those weaknesses create problems.
The configuration diversity problem
A three-site pilot is often conveniently uniform for testing: the same inverter brand, battery model, similar array orientations, and similar grid connection terms. That consistency is partly deliberate and partly a result of early-stage operators buying from a small supplier group.
A fifteen-site fleet rarely remains uniform. Sites arrive at different times, from different equipment suppliers, and with different inverter interfaces. Export limits vary. Battery chemistries vary too: some sites use lithium iron phosphate, others nickel manganese cobalt, with different round-trip efficiency curves and degradation behavior. Some have operated for eighteen months; others are newly commissioned.
Applying one dispatch rule across the fleet therefore creates suboptimal results at a meaningful share of sites. An overnight charge target of 70% may suit a calibrated new site, but be too low for one with battery capacity degradation and too high for one with a restricted morning solar window. Uniform settings simplify operations, yet the resulting revenue gap grows with portfolio size.
Where dispatch coordination fails as the fleet grows
SoC visibility is usually the first coordination failure beyond the pilot stage. At three sites, an operator can open three browser tabs and check SoC within seconds. At fifteen, opening fifteen tabs, reading current SoC, and retaining a coherent mental model is not a realistic daily workflow. Decisions then rely on the last SoC viewed rather than the current value.
The second failure is intraday response latency. When MIBEL session 3 clears at 14:00 and shows higher-than-expected prices for 17:00 to 19:00, each site's response depends on current SoC and the remaining time to charge before the target window. Assessing fifteen sites and sending revised charge instructions while the batteries can still reach target SoC requires automation or an operator faster than practice allows.
Exception handling is the third failure. In a small fleet, an inverter going offline or behaving abnormally is easy to spot. In a larger fleet, a site may quietly follow stale settings from three days ago after a brief connection interruption, while the dispatch system provides no clear error state. It can go unnoticed for days. One site using stale settings for a week loses modest absolute revenue, but the effect accumulates across a full fleet over annual timescales.
Sequencing infrastructure alongside physical expansion
Operators that scale effectively treat dispatch infrastructure as a parallel investment to site acquisition, not something added after the fleet becomes large. Per-site data aggregation and a centralized SoC view should be ready before manual tracking exceeds the operator's cognitive capacity. Automated intraday response should also precede the point where missed sessions become a material revenue issue.
The error is to defer dispatch precision to the pilot and dispatch infrastructure to the scale phase. The order should be reversed: infrastructure must precede scale because installing it after a large, active fleet is disruptive. The operator is changing dispatch rules for active sites, not configuring new ones.
A practical sequence is to add automated overnight target-setting and per-site SoC monitoring at five to eight sites. Manual effort stops scaling even modestly for these functions. Add automated intraday session response at ten to twelve sites, when session-by-session manual work becomes the daily bottleneck. Above fifteen sites with material configuration diversity, add per-site configuration management, including battery degradation tracking and tariff structure per site.
What does not automatically improve with better infrastructure
Dispatch infrastructure cannot correct basic site configuration errors. A battery with the wrong export limit settings will behave incorrectly under manual or automated dispatch. If an inverter's API returns stale data, automation receives the same stale data that the manual operator saw. Infrastructure is the scaffold; accurate site-level configuration is the substance.
Automated dispatch also does not make every site equally valuable. Sites with small battery capacity, constrained grid connections, or limited solar irradiance contribute less fleet revenue per site than larger, better-situated installations. Automation does not change that relationship. During fleet expansion, operators should assess each acquisition's marginal revenue contribution with dispatch infrastructure applied, not only its headline capacity.
A realistic transition from pilot to fleet operations
Moving from pilot-scale to fleet-scale operations is gradual, not a step change. Complexity accumulates until manual processes can no longer handle the fleet, while automated infrastructure is still absent. Planning can avoid that gap by investing in infrastructure slightly before it becomes necessary.
Operators we work with who managed this transition successfully report the same pattern: they chose dispatch automation at eight to ten sites, while manual work was still manageable. They acted on the trajectory, not because they were overwhelmed. The change was smoother because they could still configure each site carefully and compare automated behavior with manual expectations.
