How early-stage energy analysis
goes wrong — and why
everyone lets it.
This is not a product page. It is an attempt to describe, as honestly as possible, a problem that has cost the renewable energy industry billions in misallocated development time — and that almost nobody talks about directly, because almost everybody is implicated.
Picture a project review meeting. A developer presents a 120 MW wind site in a market they've been watching for two years. The LCOE is $47/MWh. The available PPA is $52/MWh. There is a $5 margin. The recommendation is to proceed to detailed development.
Across the table, a financial analyst pulls up their own model — built independently, on the same project, with inputs sourced from the same benchmark reports. Their LCOE is $54/MWh. The project is underwater by $2/MWh. Their recommendation is to pass.
Both models are technically correct. Neither analyst made an arithmetic error. The $7 difference — the entire decision — sits in three inputs nobody agreed on beforehand: the WACC, the capacity factor, and what counts as "all-in" CAPEX.
The meeting does not resolve the disagreement on its merits. What happens instead is one of three things: the more senior person's number wins; the decision is deferred pending "further analysis" that never quite arrives; or — most commonly — a compromise number is agreed that neither analyst would defend independently.
The project's actual viability has not been determined. The meeting has produced a decision, but not an answer. Those are different things, and the industry treats them as the same.
This is not an edge case. This is the standard operating condition of pre-feasibility analysis in renewable energy development. It happens in small independent developers and in the project development arms of major utilities. It happens in markets with mature financing ecosystems and in frontier markets. It happens to experienced analysts who know exactly what they're doing.
It happens because the industry has never agreed on what a pre-feasibility analysis actually is — what inputs it uses, what assumptions it holds fixed, what uncertainty it acknowledges — and so every team builds their own version of the same thing, inconsistently, repeatedly, from scratch.
"The renewable energy industry has spent twenty years getting very good at building projects. It has spent almost no time getting good at screening them."
LCOE is, at its core, a discounted cash flow calculation. Which means the discount rate — the WACC — is not one input among many. It is the most sensitive single number in the entire model. A 1% change in WACC moves LCOE by 3 to 5 $/MWh on a typical utility-scale wind project. A 3% difference — entirely plausible between two analysts choosing slightly different assumptions — can swing the result by $10–15/MWh. That is the difference between a project that looks viable and one that looks marginal.
Published benchmarks make this worse, not better. IRENA publishes weighted average cost of capital figures by region. They are useful for understanding the broad landscape. They are not useful as inputs for a specific project in a specific market in a specific year — because WACC is not a regional average. It is a function of the specific capital structure being contemplated, the perceived country and offtake risk at the moment the analysis is being done, and the cost of debt available to this developer, in this market, today.
The third number in that row is the one to sit with. 8.1% is the published benchmark. A large Indian conglomerate with strong rupee-denominated bank relationships might genuinely achieve a blended WACC near that figure. A foreign-currency-funded independent developer entering India for the first time, in a year where RBI policy is tightening, might be looking at 12–13% on a realistic all-in basis.
Those are not the same project. But if both analysts reach for the IRENA benchmark — because it's published, because it's credible, because it's easier than defending a project-specific assumption — they will produce the same LCOE, which will be wrong for at least one of them, and possibly both.
The correct approach at pre-feasibility stage is not to use the benchmark as the input. It is to use the benchmark as a sanity check on your own estimate — and to be explicit about where your estimate sits in the distribution and why. That distinction sounds minor. In practice it is the difference between an analysis that survives a challenge and one that collapses the moment someone asks "where did that WACC come from?"
The reason analysts use published benchmark WACCs as inputs rather than as references is not laziness. It is deniability. If your WACC comes from IRENA, nobody can blame you for it. If your WACC comes from your own judgement about this project in this market at this moment — and you're wrong — that is on you. The incentive structure actively rewards anchoring to published numbers, regardless of whether they're appropriate. The result is a generation of pre-feasibility analyses that are defensible in process and unreliable in substance.
The LCOE goes into a deck. The deck goes to a committee. The committee approves or declines the project. Nobody in the committee meeting asks where the WACC came from. The analyst who built the model is not in the room. The number is on slide 7, next to a green checkmark indicating it clears the hurdle rate.
This is not a failure of the committee. They are doing exactly what committees do — reviewing a summary prepared by people they trust, making a decision within the time available. The failure happened two steps earlier, when the model was built to be presentable rather than to be challenged.
A number that has never been seriously questioned is not a robust number. It is a number that has been lucky so far. The projects that expose this are the ones that reach detailed development, commission a proper resource assessment and a full financial model, and find that the LCOE is $12/MWh higher than the pre-feasibility said. By then, development costs have been spent. Sometimes a lot of them.
That sentence is going to bother some people. It should. It is also, in the majority of cases, accurate — and the people it bothers most are the ones who have written one, or commissioned one, and know exactly what went into it.
The pre-feasibility report as a genre has a recognisable structure. Executive summary. Methodology section — which references IRENA, Lazard, and GlobalWindAtlas in roughly that order. Assumption table. LCOE output. Sensitivity chart. Conclusion. Appendix with the model. The model is an Excel file. The Excel file has the same structure as the last one the consultant produced, for a different project, in a different market, with the inputs changed.
There is nothing wrong with Excel. There is nothing wrong with that structure. The problem is what the report costs, how long it takes, and — most importantly — what the cover page implies about the rigour of what's inside. A pre-feasibility study that takes six weeks and costs $40,000 carries an implicit promise of depth that the analysis, at this stage of development, cannot and should not fulfil. The cover page says "feasibility study." The contents say "we ran the benchmarks."
A developer has three sites under preliminary consideration. They need to decide which one, if any, to advance to detailed development. The internal team does not have capacity to model all three properly. They commission a pre-feasibility study.
The consultant delivers three reports. Each one is formatted identically. Each one uses regional benchmark CAPEX and OPEX from the same published source. Each one uses the same WACC. The capacity factors come from GlobalWindAtlas, unadjusted. The loss stack is the consultant's standard assumption, not site-specific.
The developer has spent $45,000 and six weeks to produce three analyses that any competent analyst could have produced in a day, using publicly available data, with no loss of rigour at this stage of development. What the developer paid for was not analytical quality. It was the comfort of having a named firm's logo on the cover page.
This is not a criticism of the consultant. They delivered exactly what was asked for, at the price the market bears. It is a criticism of the market — which has confused the credibility of the deliverable with the rigour of the analysis.
There is a version of this story that is more uncomfortable still. Pre-feasibility consultants have, in aggregate, a financial incentive to keep pre-feasibility analysis difficult. Not through any active conspiracy — it is simpler than that. If a developer believes that credible pre-feasibility requires a specialist, they will keep hiring specialists. The moment a developer realises they can produce a defensible Class 3–4 result themselves, in an afternoon, that revenue stream narrows.
The consultant ecosystem is not going to tell you this. So we are.
Published benchmarks are averages. Averages hide the most useful information — which is where a specific market diverges from the average, and why. But before getting to the market-specific detail, there is one observation that applies equally across India, SE Asia, and MENA, and it is the one nobody in the consultant ecosystem is going to volunteer:
In all three markets, pre-feasibility is routinely priced and scoped as if it were Class 1 work. The reports are long. The timelines are measured in weeks. The fees reflect the brand on the cover page rather than the complexity of the analysis inside. And because developers in these markets are often navigating unfamiliar regulatory environments, they accept it — because the alternative feels riskier than the cost.
It is not. Here is what actually varies by market at pre-feasibility stage, and what doesn't.
India's published LCOE benchmarks are among the most competitive in the world. They are also among the most misleading for anyone who isn't one of the five largest developers in the country. The benchmark reflects a market dominated by entities with rupee-denominated debt access, government relationships, and balance sheets that let them absorb currency and offtake risk that a smaller or foreign developer cannot.
The WACC range in practice — across the full population of developers actually active in Indian renewables — is wide. Very wide. And yet almost every pre-feasibility analysis conducted for an Indian wind or solar project reaches for the IRENA benchmark WACC, uses it as the input rather than the reference, and produces an LCOE that is competitive on paper and potentially fictitious in practice for that specific developer.
The consultant does not flag this. The report moves forward. The committee approves the screen. The project reaches financial close discussions and the cost of capital reality arrives — sometimes two years and significant development spend later.
Southeast Asia has some of the most attractive renewable resource conditions in the world and some of the most constrained grid infrastructure. Pre-feasibility analyses in the region consistently model the resource well and model the grid not at all. The capacity factor comes from an atlas. The curtailment assumption is either zero or a standard loss stack percentage that does not reflect any knowledge of the specific grid zone.
This is not a criticism of the analyst. Grid constraint data in SE Asian markets is difficult to obtain, inconsistently published, and changes with investment cycles. The honest response to that uncertainty is to model it explicitly — to run a base case and a curtailment-adjusted case and to be clear about which one you are making decisions on. The standard response is to use the atlas CF and say nothing about curtailment.
The difference between those two approaches does not show up in the pre-feasibility. It shows up in year three of operations when actual generation is tracking 15% below the investment case and nobody can explain why the model didn't see it coming.
MENA wind resources, as shown in GlobalWindAtlas and similar tools, are among the highest capacity factors in the world. Some of those numbers are real — Saudi Arabia, Oman, and parts of Morocco have genuinely world-class wind resources that have been confirmed by measured data campaigns. Others are atlas artefacts: the 250-metre resolution of publicly available tools cannot capture the orographic effects, marine layer dynamics, and diurnal variability that characterise coastal MENA sites.
The pattern that repeats: a developer screens a MENA site at 40–44% capacity factor, the economics look compelling, development spend is committed, a resource assessment is commissioned, and the bankable P50 comes in at 35–37%. The project still works, but the margin is different. In some cases the project doesn't work at all at the realistic CF, and would not have passed the screen if the screen had used a more conservative assumption.
A pre-feasibility consultant working in MENA knows this pattern. The report does not always reflect it — because flagging the uncertainty means the client might not proceed, which means the engagement ends at pre-feasibility rather than continuing to detailed study. That is not a conspiracy. It is an incentive.
Every tool embeds its builder's opinions. The honest thing to do is name them. Here are the decisions in preFeasibility's methodology that were not obvious, with the reasoning behind each one and the trade-off we accepted.
Good pre-feasibility is not more detailed than what we have described. It is more honest. It names its assumptions explicitly. It acknowledges the inputs it is sensitive to and shows you what happens when they move. It does not dress up a screening result in the language of a bankable study, because that conflation is where most of the damage happens.
A good pre-feasibility result tells you three things clearly: what LCOE you get at the central case, how much it moves if the two or three most sensitive inputs are wrong, and where that LCOE sits relative to the market context. If those three things are visible, traceable, and honest about their uncertainty — you have what you need to make a good screening decision.
Everything beyond that is either detailed feasibility work — which belongs at a later stage — or decoration.
"The goal is not a number you can't argue with. It is a number you can argue with intelligently — because you can see exactly where it came from."
That is what we have tried to build. Not a tool that produces an answer, but a tool that produces an auditable answer — one that shows its work, acknowledges its limits, and invites the challenge rather than pre-empting it with a consultant's letterhead.
If you have read this far and found yourself agreeing, disagreeing, or — best of all — recognising something from your own experience, we would genuinely like to hear from you. The methodology page is versioned. The thinking is not finished. And the people who know where this analysis breaks down are the ones who have watched it break.
See the methodology behind the platform
Every formula, every fixed assumption, every limitation — documented in full.