Most retrofit plans don't fail in the boiler room. They fail at the capital allocation table. Here is what our own portfolio data says about why, and what changes when the plan is written in the language the money uses.
An asset manager walks into the investment committee with a technically sound retrofit plan. The measures are sensible, the savings are real, the consultant did good work. It still doesn't get funded.
The committee is not rejecting the engineering. It is rejecting a document that answers questions nobody in the room asked. Kilowatt-hours and CO2 savings where the committee needs euros. "Recommended measures" where it needs a phased CapEx line with a funding requirement attached. A decarbonization pathway where the credit desk needs to know what happens at refinancing.
The market is not short of retrofit ideas. It is short of investment-grade translation.
That distinction is expensive, because the deadlines are fixed and the clock is already inside the portfolio.
The recast Energy Performance of Buildings Directive (EU) 2024/1275 sets two different obligations, and they are routinely conflated.
Non-residential buildings face an asset-level standard. Member States must set minimum energy performance standards ensuring that the worst-performing 16% of the national stock is improved by 2030, extending to the worst-performing 26% by 2033. If an asset sits in that band, it has a renovation deadline.
Residential buildings face a stock-level trajectory instead. Average primary energy use must fall by at least 16% by 2030 and 20 to 22% by 2035 against a 2020 baseline. At least 55% of that reduction must be delivered by renovating the worst-performing 43% of the residential stock. No single building carries an individual deadline, but the policy pressure is pointed squarely at the same tail of the distribution.
Carbon pricing lands on top. ETS2 covers fuel combustion in buildings from 2027 auctioning, fully operational in 2028. Whatever the clearing price turns out to be, gas-heated assets have been handed a new operating cost with a known start date.
We analyzed a portfolio of more than 10,000 rated assets in blue auditor, and roughly ONE in FOUR (25%) non-residential assets sits in EPC class F or G. That is the band MEPS renovates first.
On the residential side the distribution is close to identical at the bottom: about 38% of assets sit in E, F or G. Against a worst-performing 43% cut, that is not a marginal exposure. It is most of the tail.
These are not modelled national averages. They are rated assets in institutional portfolios, which is why the number is uncomfortable. The renovation requirement is already in the stock people own today.
Here is the finding that changes how these plans should be built.
For the same assets, costed twice, once from the energy performance certificate and once from operational metered data, the two retrofit cost estimates diverge by a median of 68%. Not 68% high. 68% apart, and it runs in both directions. Sometimes the certificate-based number is far too low. Sometimes it is far too high.
The point is not which estimate is bigger. The point is that a number moving that much cannot be underwritten.
Put yourself on the other side of the table. A funder sees a capital requirement that could be wrong by more than half. The rational response is to add a conservative buffer, or to decline. Weak baseline data produces weak assumptions, and weak assumptions get priced.
This is why "get better data first" is not a technical footnote. It is the difference between a roadmap and a fundable plan.
Take one asset from the book. Retail, approximately 8,000 m², EPC class F. The deep retrofit costs €4.39M.
Left unfunded and undefined, that figure sits in the underwriting as a €4.39M hole. Pure downside. Nobody wants to touch it.
Funded, the same €4.39M answers three different people in the room at once.
Cash return: €3.1M. Around €258,000 a year in energy savings, a 5.9% direct return. Real cash that helps service the debt. This is the answer to the credit desk.
Value return: €350,000 to €530,000. The priority package of identified measures, at strong efficiency per euro spent. This is the asset manager's IRR story.
Risk return: €1.6M. A credible, costed, funded plan supports a cap rate roughly 100 basis points tighter, 8.5% to 7.5%. On this asset that is €1.6M of value, and none of it comes from income growth. Nothing about the rent changed. What changed is that the uncertainty is gone.
The biggest lever here is not the energy saving. It is removing the uncertainty that was forcing the discount.
Now take the same €4.39M of works on the same asset and change nothing except the timing. Four plans, modelled over a ten-year hold.
Everything at once returns 3.4% IRR and a €13.97M exit. Sequenced to lifecycle, priority package now and the heavy envelope work at its natural replacement window, returns 3.6% and the same €13.97M exit.
Identical destination. The difference is the funding requirement to get there.
All at once drives the owner's cash position to minus €3.9M in year one and holds a multi-million hole for years. Sequenced never goes deeper than minus €1.4M, because the priority package pays for itself early and the heavy spend waits for the moment it was needed anyway. Nearly three times less peak capital at risk, for the same value.
Deferring to years seven and eight looks comfortable early and gives up both return and exit value: 0.7% IRR, €13.10M exit. Doing nothing is cash positive the whole way until the exit collapses from €13.97M to €7.55M, an IRR of minus 7.1%.
Sequencing does not change whether the retrofit pays. It changes whether you can fund it.
This is the part most retrofit business cases get backwards, and it matters most to whoever signs off on the debt.
Same asset. Debt of €7.54M at 55% LTV, coverage stress-tested at plus 75 basis points, the way a credit desk works in 2026.
Stressed DSCR is 1.93x unfunded and 2.78x funded, against a covenant floor of 1.25x. Debt yield is 9.6% and 13.9%, against a floor of 8 to 9%. Both clear the covenants in both cases. On the day-one screen, the loan looks safe even with no retrofit at all.
Coverage is not where the risk lives.
It lives at maturity. Unfunded and stranded, the asset is worth €8.8M at a 9.0% exit cap. Against €7.54M of debt, that is 86% refi LTV, and it supports only €4.8M of new debt. A €2.7M shortfall: an equity call, or a forced sale into a market that does not want the asset. Funded and sequenced, the exit value holds at €14.0M, refi LTV drops to 54%, and the loan refinances with headroom.
The unfunded asset services its debt right up until the day it has to refinance against a stranded valuation, and then it cannot.
One note on method, because it always comes up: the exit differences here come from the cap rate, not from invented income growth. The 7.5% delivered, 8.0% deferred and 9.0% stranded steps are scenario assumptions, grounded in observed brown discount evidence. A University of Cambridge study drawing on more than 100,000 UK lease comparables found EPC F and G rated offices experiencing rental declines of 6 to 8%, with the discount appearing before the compliance date rather than after it. Capital does not wait for the deadline. It prices the deadline.
Six things, and a plan missing any of them tends to stall:
Engineering logic is necessary. Capital logic gets the check signed.
The full analysis behind these numbers, including the sequencing curves and the credit screen, is in our webinar recording: Retrofit Roadmaps That Actually Get Funded.
If you would rather test it directly: send us one asset. We return a costed, phased, scenario-tested roadmap written in the language your investment committee and your lender already use. If it changes the conversation on one asset, we scale it across the portfolio.
Send us one asset