Payback: how to judge the numbers

Every solar and battery proposal leads with a payback figure, and the figure is doing sales work. Payback is not a fact about the hardware; it is the output of a model, and the model’s assumptions are chosen by whoever wrote it. The same site can honestly support quite different payback figures depending on those choices, which means the number itself tells you little. What tells you something is the quality of the model behind it, and whether the person who built it profits from you believing it.

Published payback ranges for commercial solar vary widely between sources, and the variation is the message: these figures are functions of site, profile and assumption, not properties of the technology. Judging a proposal by comparing its payback to a number from the internet replaces one unverified figure with another.

The assumptions that decide everything

Five inputs dominate any payback model. The self-consumption rate: how much generated power is used on site rather than exported. The tariff assumption: what import price the generation displaces, which depends on your contract structure. The energy price trajectory: how fast prices are assumed to rise over the asset’s life. The export value: what surplus power earns and when. The cost tail: maintenance, monitoring, insurance and inverter replacement. Small optimism in each compounds into a headline that bears no relation to the site.

When we audit an installer’s model, the checklist is short and it maps to the same five assumptions. Is the self-consumption rate calculated from the site’s own half-hourly data, or assumed from a generic profile for the building type? Does the tariff assumption match the contract actually in place, including its structure, not just a headline unit rate? Is the energy price trajectory stated explicitly, and does it hold up against a flat or conservative case, not only the trajectory that makes the payback shortest? Is the export value modelled against the times power is actually likely to leave the site, rather than a single annual average rate? And does the cost tail include inverter replacement, monitoring, insurance and maintenance across the full asset life, not just the first few years when nothing has needed replacing yet? A model that survives all five questions is worth trusting. One that goes quiet on any of them was built to sell, not to inform.

The incentive problem

The party writing most payback models is the party selling the system. That does not make installers dishonest; it makes their models structurally untestable by the buyer, because every disputable assumption resolves in the sale’s favour and the buyer holds no reference model to argue from. The organisation’s finance function senses this, which is why self-interested models stall at sign-off even when the underlying project is sound.

Independent verification means rebuilding the model with the same rigour but a different incentive: the person doing the work is paid the same whether the answer is favourable or not. That single change alters which way every disputable assumption gets resolved. A self-consumption rate gets set from the metered data instead of rounded up. An energy price trajectory gets stated as a range instead of a single optimistic line. Export value gets modelled against the actual export profile instead of a flat annual rate that happens to flatter the numbers. None of this requires the independent model to be more sophisticated than the original. It usually needs to be less generous, because generosity was the point of the first one. What changes is not the arithmetic. It is who benefits from where the arithmetic lands.

Payback is also the wrong question asked alone

A payback period compresses an asset’s whole life into a single early milestone. Two projects with identical payback can have very different lifetime value, depending on what happens in the years after breakeven and how the asset interacts with tariff structure over time. A proper appraisal looks at lifetime cashflows under your real tariff, tested against pessimistic as well as central assumptions. If the case only works in the optimistic run, there is no case.

Before you sign off a proposal on the strength of its payback figure, have the model independently rebuilt against your own data. Book a review at /book or request a benchmark at /benchmark.