Will sellers actually provide raw telemetry?

Increasingly, yes — and refusal is itself diligence information. The practical friction is more often format and export mechanics than willingness; requesting the data early in exclusivity, with a specified schema, is the single best process improvement available.

Does this replace the independent engineer?

No — it arms them. IE scopes have historically been document-bound; telemetry analysis gives the IE measured evidence for the assumptions they were previously asked to bless on reputation. Several IEs now incorporate reconciliation outputs directly into their reports.

Is two years of data enough to measure degradation?

It bounds it, weakly — three or more years of weather-corrected trend is where the confidence interval becomes decision-grade. With less history, the honest move is to widen the assumption range and price the uncertainty explicitly.

Can this be done on wind and storage assets too?

Yes — the method is technology-agnostic: expected output from measured conditions, deviations attributed. Wind adds power-curve and yaw analysis; storage adds state-of-health measurement against the augmentation schedule, where OEM aggregate reporting hides at least as much as solar availability does.

What if the target has no string-level telemetry?

Inverter-level analysis still recovers most of the valuation signal — peer deviation, degradation slope, thermal trends, fault archaeology. String granularity sharpens the recoverable-loss inventory but is not a precondition for the analysis.

Technical Due Diligence for Renewable M&A: What Acquirers Miss in the Operating Data

Virat Damania — Co-Founder & Managing Director, Ellume Technology

The data room tells you what the seller measured. The telemetry tells you what the asset is. They are rarely the same document — and the difference has a clearing price. Conventional technical due diligence for operating renewable assets relies on seller-provided reports — availability, budget variance, O&M summaries — which systematically understate operational problems. Interval telemetry analysis reveals what those reports cannot: chronic string-level losses, real degradation rates, gamed availability, sensor decay, and deferred-maintenance signatures. For a typical acquisition, telemetry-based DD moves the energy assumption by 1–4%, which at portfolio scale is a valuation event.

Key Takeaways

→The data room is a record of what the seller's measurement regime could see; its blind spots transfer to the buyer at closing, priced as if they didn't exist.
→Five valuation-moving discoveries recur: real degradation slope, chronic sub-threshold losses, availability methodology, telemetry quality itself, and deferred-maintenance signatures.
→Telemetry-based DD reconstructs expected generation from measured weather across the trailing 2–3 years and attributes every deviation — turning document review into measurement.
→The output is an inventory, not a report card: recoverable MWh (a buyer's value-creation lever), a measured degradation slope (a model input), and remediation capex (a negotiation line).
→Quality-of-energy analysis is becoming for renewable transactions what quality-of-earnings became for corporate M&A: eventually standard, currently an edge.

A fund acquires an operating solar portfolio. The independent engineer reviewed the data room: availability above 99%, generation within two percent of budget, clean O&M summaries. Eighteen months later, the asset is underperforming its acquisition case and the new owner is discovering — one truck roll at a time — years of accumulated string faults, a degradation rate half a point above assumption, and an availability figure that was measured generously. None of this was fraud. It was the ordinary gap between reported performance and physical performance — the same gap this publication keeps returning to, now with a purchase price attached. In M&A, that gap has a clearing price, and the party who measures it first keeps it. Increasingly, sophisticated buyers are measuring it. This essay is about how.

What does the data room systematically miss?

Five things, in rough order of valuation impact.

•Real degradation. Sellers report nameplate-adjusted generation; only a weather-corrected performance trend over multiple years reveals the true slope, and the difference between 0.5% and 0.8% annual degradation compounds directly into terminal value. It is the single largest number in the model that the data room typically cannot evidence.
•Chronic sub-threshold losses. Dead strings, derating inverters, and soiling drift never appear in availability but permanently subtract two or three percent of energy. Unidentified, the buyer inherits them silently, priced into the historical generation baseline as 'plant characteristics.' Identified, they are the opposite: a value-creation lever, because the buyer acquires recoverable energy the seller's price never captured.
•Availability methodology. Exclusion rules, force-majeure definitions, grace periods, and the measurement point can each add a flattering point to the headline figure. An availability number is only meaningful next to its formula, and the formula rarely travels with the number.
•Data quality itself. Decayed irradiance sensors, telemetry gaps, and re-baselined meters mean parts of the operating history are simply not evidence. A DD process that doesn't audit the instruments is auditing their errors.
•Deferred-maintenance signatures. Rising thermal trends across the inverter fleet, fault-frequency creep, aging remaining-useful-life profiles, and repair backlogs predict capital expenditure the model has not budgeted — visible in the telemetry years before it lands in a budget.

Reported metric versus what telemetry-based DD examines
Data-room artifact	What it hides	Telemetry counterpart
Availability %	Partial faults, exclusion rules	Energy-based loss attribution
Generation vs budget	Resource luck offsetting asset decline	Weather-corrected performance trend
O&M reports	The detection stack's blind spots	Fault archaeology in interval data
Degradation assumption	Actual site-specific slope	Multi-year corrected PR regression
Warranty status	Claims never detected, now expiring	Component-level anomaly history

What does telemetry-based DD actually involve?

Request interval data — inverter-level at minimum, string-level where it exists — plus on-site meteorological measurements, for the trailing two to three years. Reconstruct expected generation from measured weather across the entire history, and attribute every deviation. The output is not a report card; it is an inventory. This many megawatt-hours per year of recoverable loss — a value-creation lever the buyer can execute in the first two quarters of ownership. This real degradation slope, with confidence intervals — a model input replacing an assumption. This availability figure, recomputed on a stated methodology — a negotiation input. And this list of latent issues with estimated remediation cost — a capex line, priced before it becomes a surprise. Two to four weeks of analysis, one asset-level truth, and DD converts from document review into measurement. The asymmetry favors whoever runs it: a buyer prices the flaws into the offer; a sophisticated seller runs it first and prices the flaws out of the discount.

How Ellume Vector runs acquisition analysis

Everything Vector does for an operating fleet — the physics model, the energy reconciliation, the fault attribution — runs equally on historical telemetry, which is what makes it a due-diligence instrument: the analysis works pre-close, from data-room exports, without site access.

•Historical reconciliation: the energy waterfall runs across the full trailing history — theoretical generation from measured weather, down through every loss bucket, to metered actuals — producing the recoverable/unavoidable/asset-fault partition for each year of the record. On a representative asset, that reconciliation reads 315 MWh theoretical to 278 MWh actual over a sample period, with 7,804 kWh explicitly recoverable: the buyer's day-one work queue, priced.
•Fault archaeology: the physics rules run retrospectively across the record, surfacing every event the seller's monitoring never flagged — including the chronic cases, like an inverter that ran 48.5% below its fleet peers for 121 days, accumulating 6,973 kWh of loss that appears nowhere in any O&M report.
•Degradation measurement: multi-year weather-corrected performance regression with the seasonal thermal wave removed — the measured slope that either confirms the model's assumption or reprices it, with the evidence attached for the IE and the lender.
•Instrument audit: telemetry-quality scoring flags sensor decay, data gaps, and implausible readings, so the buyer knows which parts of the history are evidence and which are noise — before relying on either.
•Fleet condition inventory: per-inverter health scores and remaining-useful-life estimates aggregate into a replacement-timing profile — the capex schedule the model should have contained, derived from thermal history rather than calendar assumptions.

The negotiation math: suppose telemetry DD on a 100 MW target finds 1.5% of annual energy in recoverable faults, a degradation slope 0.2 points above the model, and a two-year inverter replacement pull-forward. The first is upside the buyer captures post-close; the second and third are price adjustments with evidence attached. Any one of the three typically exceeds the entire cost of the analysis by two orders of magnitude. The question is not whether quality-of-energy analysis pays for itself — it is which side of the table commissions it first.

Who does this asymmetry favor?

Today, the prepared party — usually a buyer, occasionally a seller sophisticated enough to run the analysis first and cure or disclose on their own terms. In the medium term, it favors no one, because it becomes standard: the way quality-of-earnings analysis became non-negotiable in corporate M&A, quality-of-energy analysis is becoming non-negotiable in renewable transactions. IEs are beginning to request interval data as a matter of scope; lenders are beginning to ask what the degradation assumption is evidenced by. The only open question is which participants learn this from experience and which learn it from a write-down.

Frequently Asked Questions

Will sellers actually provide raw telemetry?: Increasingly, yes — and refusal is itself diligence information. The practical friction is more often format and export mechanics than willingness; requesting the data early in exclusivity, with a specified schema, is the single best process improvement available.
Does this replace the independent engineer?: No — it arms them. IE scopes have historically been document-bound; telemetry analysis gives the IE measured evidence for the assumptions they were previously asked to bless on reputation. Several IEs now incorporate reconciliation outputs directly into their reports.
Is two years of data enough to measure degradation?: It bounds it, weakly — three or more years of weather-corrected trend is where the confidence interval becomes decision-grade. With less history, the honest move is to widen the assumption range and price the uncertainty explicitly.
Can this be done on wind and storage assets too?: Yes — the method is technology-agnostic: expected output from measured conditions, deviations attributed. Wind adds power-curve and yaw analysis; storage adds state-of-health measurement against the augmentation schedule, where OEM aggregate reporting hides at least as much as solar availability does.
What if the target has no string-level telemetry?: Inverter-level analysis still recovers most of the valuation signal — peer deviation, degradation slope, thermal trends, fault archaeology. String granularity sharpens the recoverable-loss inventory but is not a precondition for the analysis.