When BESS suppliers submit proposals, one number immediately attracts attention: $/kWh. It is simple, familiar, and easy to compare. It is also, on its own, an incomplete basis for comparison.
The difficulty is not that suppliers are being evasive. Most are not. The difficulty is that two technically compliant proposals for the same project can be priced against materially different obligations — different definitions of the capacity being sold, different assumptions about who pays to maintain it over time, and different amounts of long-term service wrapped into the number. Rank those by price per kilowatt-hour, and you will not be choosing the cheapest system. You are choosing the supplier whose commercial structure happens to put the least cost inside the line item you looked at.
To make this concrete, take a typical utility-scale requirement: 100 MW / 400 MWh, a four-hour system, cycled about once a day, with a contractual obligation to deliver that 400 MWh of usable energy at the Point of Connection for the full term. Every supplier bidding it will quote an installed capacity, a price per kilowatt-hour, and a total. Those totals are the least comparable numbers in the package.
What follows is where the differences hide.
Start With What "Usable" Means
A proposal will quote installed capacity and, usually, usable capacity. The second number is the one that matters, and it is the one least likely to be defined the same way twice.
Between battery nameplate and energy actually delivered at the POC sit several distinct losses:
the state-of-charge window — the portion of nameplate the BMS will actually cycle, commonly around 97% for LFP, though it is a design and warranty decision rather than a physical constant
PCS conversion losses on discharge
transformer losses through to the POC
auxiliary consumption — thermal management, controls, fire systems — a real and continuous draw
Apply only the first and you are at roughly 97% of nameplate. Apply all four and you are typically somewhere in the low 90s — often around 92%, and lower still in hot climates or air-cooled designs where auxiliary draw is heavier. Both figures are referred to as "usable energy" in the supplier documentation. They describe different things, measured at different places, and the gap between them — four points or more — is wide enough to reorder a bid list on its own.
There is a further ambiguity worth catching early. Some suppliers quote container capacity as raw rated DC energy; others quote it with the SOC window already applied. Take the second at face value and apply the window again, and you have double-counted it. Confirm which convention each bidder is using before doing anything else with the number.
The effect is larger than the percentages suggest, because what this really changes is margin over the requirement. A proposal that appears to have comfortable headroom above 400 MWh may, once measured at the delivery point, have very little — and headroom determines when the system first needs capacity added, which in turn drives a large part of its lifetime cost.
Establish in writing, for every bidder: the measurement point, the ambient temperature basis, the C-rate, the SOC window, and whether auxiliary load sits inside or outside the quoted figure. Until those five match across all proposals, the prices are not comparable, and no amount of spreadsheet work will make them so.
Headroom Is a Commercial Decision, Not a Technical One
Every supplier oversizes. Cells fade, the contract requires a floor, and the gap has to be covered — either by installing more capacity at the start or adding it later. On a four-hour system on daily cycling, initial oversizing commonly falls between 8% and 16% above the contractual requirement, and that spread is not primarily a technical difference. It is a pricing strategy.
A supplier installing at the low end quotes a lower price and carries a heavier future augmentation program. One installing at the high end quotes more and augments later and less often. Both can be entirely compliant. Only one shows its full cost within the number being compared.
The trap is that oversizing percentages are quoted against installed nameplate, so they look directly comparable and are not. Two bidders quoting the same oversizing figure can have quite different actual margins over the requirement, purely because of the definition question above. A nominal 10% that is measured at DC terminals is a materially thinner cushion than a nominal 10% measured at the POC — and in a four-hour system on daily duty, the difference can be the gap between a first augmentation in year two and one in year five.
The more useful question is therefore not "how much oversizing did you include" but "in which year does this system first require augmentation, and what is the guaranteed usable-energy profile, year by year, for the full term?" That answer cannot be fudged by a measurement convention, and it exposes the capacity strategy.
Augmentation: Establish Who Owns the Obligation
Suppliers approach the augmentation commitment in fundamentally different ways, and the distinction is frequently lost because all three arrive in the bid as a line labeled "augmentation."
Structure
What the supplier is actually promising
Performance-guaranteed including augmentation
Supplier guarantees usable energy and carries the augmentation obligation at its own cost and risk
Firm-priced augmentation
Supplier guarantees a degradation curve; owner pays for augmentation at a contractually fixed price, often indexed
Indicative augmentation price
Supplier provides an estimated augmentation price, explicitly non-binding and re-priced at the time of the event
Different products, different risk profiles. Placing a firm-priced schedule next to a non-binding estimate in the same comparison row is the most common way a lifecycle analysis quietly becomes meaningless — and it favors the least committed bidder, because a non-binding number can be optimistic at no cost to the party quoting it.
There is a useful sanity check here. Augmentation often reuses the PCS, transformers, controls, land, and civil works. It is cells, racking, integration, and labor — not a new system. It should therefore price well below new-build, typically in the region of 60–75% of that supplier's own cost per kilowatt-hour for the original plant. A quoted augmentation that approaches or exceeds the new-build price is not really an augmentation price. It is an option premium on a scope the owner will have no practical choice but to buy, and it should be challenged before it enters any model.
Augmentation also costs more than modules. Container space, land, and foundations must be reserved, and often built, for blocks that arrive years later. Cabling and protection have to be sized for the final configuration rather than the initial. A cooling plant sized for the original heat load may not cover the augmented one, since aged blocks run hotter at the same current as their internal resistance increases. A future cell generation has to remain compatible with the installed DC architecture, BMS and firmware. And mixing states of health across vintages has consequences for the system-level guarantee that belong in the contract, not in a commissioning-stage discovery.
A proposal that defers capacity into the future also defers these costs. Several of them land on the owner regardless of what the supply agreement says.
Long-Term Service Is Routinely Underweighted
Maintenance and warranty are typically treated as secondary criteria — a single figure in a commercial annex, scored qualitatively, if at all. That is difficult to justify, because long-term service for a utility-scale BESS is commonly quoted at roughly $1 to $4 per kWh-year, depending on scope. Across a hundred-megawatt asset and a ten-year term, the spread between two bidders at opposite ends of that band is large enough to outweigh a meaningful equipment price difference.
The reason is structural rather than accidental. Cheap equipment recovered through an expensive, mandatory long-term service agreement is a well-established commercial model. A supplier using it is not mispricing anything. It is choosing where in the contract to make its margin — and choosing the part that receives the least scrutiny.
A widespread difference between bidders on service almost always signals a scope difference rather than a price difference. Before allowing that number to influence a decision, establish what each one actually buys:
preventive maintenance scope and frequency
remote monitoring, and the data rights attached to it
capacity testing — who performs it, how often, to what standard
warranty extension terms, and what voids them
software and licensing fees across the full term
spare parts holding, and who owns the inventory
whatever else is required to keep the performance guarantees enforceable
That last point deserves emphasis. Many performance guarantees are conditional on the owner maintaining a specific service agreement. Where that is the case, the service cost is not an operating choice. It is a committed cost and belongs in the comparison as a single item.
Read the Guarantees
Degradation. Guaranteed retention curves between suppliers running comparable LFP cells on the same duty cycle are usually close — commonly landing somewhere near 80% at ten years on daily cycling, with competitive bids often separated by only two or three percentage points at end of term. That spread is ordinary variation, not a differentiator, and it rarely deserves the review time it consumes.
What deserves the time is the structure around the curve. Is it a guarantee or an expectation? Verified by what capacity test, at what interval, under whose supervision? And critically — what is the remedy if the system misses it? A guaranteed curve with no meaningful remedy is a forecast with better formatting.
Availability. Utility-scale guarantees typically fall between about 97% and 98.5%, and the gap between those two figures looks trivial until it is priced. On a hundred-megawatt merchant asset, a percentage point or so of availability is not a rounding error in annual revenue terms.
But the figures are not comparable until the definitions are. Time-based and energy-based availability are different metrics and are not interchangeable. A guarantee measured monthly, with generous carve-outs for grid outages, curtailment, force majeure and scheduled maintenance, is a weaker instrument than a rolling annual energy-availability guarantee with narrow exclusions. Establish the definition, the measurement window, the exclusions and the liquidated damages formula before comparing two availability numbers at all. Note also that lost availability is only proportional to lost revenue if outages are uncorrelated with high-price hours, which in practice they often are not.
Round-trip efficiency. RTE must be normalized — DC-to-DC, AC-to-AC or POC-to-POC, at a stated C-rate and SOC window, with auxiliaries and thermal load either in or out. That is not optional. Whether it should then be monetized is a separate question, and the answer is usually that it belongs in the project's dispatch model rather than in a generic adjustment to procurement price. Efficiency value depends on charging cost, discharge price, cycle count, and market structure, none of which a supplier comparison needs to resolve.
The general principle: normalize everything, monetize only what can be monetized honestly, and keep the rest as separate technical criteria. Forcing an unquantifiable difference into a single financial figure produces false precision, which is worse than an acknowledged qualitative judgment.
A Guarantee Is Worth What It Can Be Enforced For
Everything above assumes the promise behind the number is collectible. Often the decisive question is not which supplier promises more, but which promise survives contact with a dispute.
Worth establishing before any award:
exclusions and carve-outs in each guarantee, read in full
the liquidated damages structure, and the caps on both LDs and total liability
the remedy mechanism — repair, replace, augment, pay — and who decides
supplier financial strength, and the entity actually signing
technology maturity and reference projects at comparable scale and duty
delivery capability and local service presence in the project's market
spare-parts strategy and lead times across the full term
A strong-looking availability guarantee with extensive exclusions and a low liability cap can easily be worth less than a more modest guarantee with clear remedies from a financially solid counterparty.
Payment terms belong here too and are routinely omitted entirely from lifecycle comparisons. On a contract of this size, the milestone schedule, advance payment percentage, retention, escalation provisions and currency exposure move the present value of the deal — sometimes by more than the equipment price difference under discussion. Discount the payment schedules, not just the headline prices.
One scope note. Tax credits, domestic content requirements, and foreign-entity restrictions sit deliberately outside this framework. In the US market, they can outweigh every consideration discussed here, and they turn on supplier-specific cell sourcing and manufacturing footprint. They require their own analysis.
On the Evaluation Period
There is no universal answer, but the choice matters and should be deliberate.
A twenty-year model is appropriate for evaluating the economics of the complete asset. For supplier comparison, that horizon introduces substantial uncertainty about future cell prices, market conditions, and operating strategy — uncertainty largely common to all bidders, which adds noise without adding discrimination. A ten-year window is usually long enough for differences in headroom, degradation, augmentation and service structure to become visible, while limiting exposure to assumptions nobody can defend.
Whatever period is chosen, the non-negotiable requirement is consistency: every supplier evaluated over the same term, on the same assumptions, at the same discount rate. And if the analysis is truncated before end of life, take an explicit position on what the asset is worth at the boundary rather than leaving it silently at zero.
How to Structure the Evaluation
First, normalize the cost. Put every proposal on a common scope boundary and a common measurement point. Build the full cost picture — initial capital, augmentation over the evaluation period, long-term service and warranty, and any owner-side costs implied by the bid structure. Discount the payment schedules. Normalize against the project's actual requirements rather than each supplier's nameplate, and disclose every assumption used so the analysis can be interrogated rather than believed.
Second, normalize the performance. Fix the definitions before comparing the figures. Ask for the year-by-year guaranteed usable-energy profile rather than a single number at COD. Monetize the differences that can be monetized, applying the same method symmetrically to every bidder, and keep the remainder as explicit technical criteria.
Third, overlay contractual risk — then decompose your own answer. Test enforceability and remedies. Then identify which single input is carrying your conclusion, and how far it would have to move to reverse it. This step is skipped almost universally, and it is the most valuable one, because lifecycle comparisons frequently resolve to a margin thinner than the uncertainty in the assumptions that produced them.
When the Result Is Too Close to Call
Suppose the analysis comes out with one bidder a couple of percent cheaper over the evaluation period. Before treating that as a result, ask what the answer would look like if one of your inputs were wrong.
If a plausible change to a single assumption — the service scope, the augmentation price, the discount rate — moves the outcome by more than the gap between the bidders, then you have not established that one supplier is cheaper. You have established two different things: that the two are close, and that one assumption is deciding the tender is decided.
That is a useful finding rather than a failed analysis. But it calls for something other than a recommendation.
Go and get the missing information. You now know exactly which input the answer depends on, which is more than most evaluations ever establish. Go back to the supplier and have it specified — the service scope written out line by line, the augmentation price confirmed as firm or indicative. A week spent closing that gap is worth more than any further refinement of a model built around a number nobody has pinned down.
Use the closeness as leverage. Two bids within a couple of percent is a negotiating position, not only an analytical problem. Tell the preferred bidder which part of their offer is carrying the gap and ask them to close it. A supplier that wants the job will often move on a line item you can name precisely.
If it still does not resolve, decide on the qualitative factors — enforceability of the guarantees, financial strength of the counterparty, delivery record, service capability in the project's market — and state clearly in the recommendation that these were the deciding factors, rather than the cost model.
The failure to avoid is presenting a two-percent difference as though it were a finding. It lends the award an appearance of rigor it has not earned, and it rests the decision on the least reliable number in the analysis.
Conclusion
A good BESS procurement process does not ask "which supplier has the lowest $/kWh?" It asks "which proposal delivers the required performance at the best risk-adjusted economic value over the period relevant to this project?"
Answering that means establishing what each supplier is actually selling before comparing what each one charges. Most of that work is definitional, and most of it happens before any model is built. The suppliers are not hiding these things. They are answering the questions they were asked, in the terms their own commercial structure favors — and a tender that asks only for price per kilowatt-hour will get exactly that, correctly, from everyone.
The cheapest BESS bid may well be the best one. But first, you have to prove it.
Daniel Schwartzberg is an electrical power engineer and independent BESS advisor. He supports developers, IPPs, and investors with BESS procurement, owner's engineering, and technical due diligence.