Modeled savings is what a solar-and-storage project was projected to save at underwriting — production model times tariff times time. Verified savings is what it actually saved, measured from the real utility bill against a reconstructed counterfactual, line item by line item. The two are the same number on the day the deal closes and drift apart from there, because the tariff and the load the model assumed both change. A quarterly report that still carries the modeled number — often with a footnote — is reporting the underwriting case, not the result. That footnote is the gap this article is about.

This guide explains the difference between modeled and verified savings, why modeled savings was the right tool at underwriting but stops being reliable over time, and what it takes to report a savings number you can defend.

#Modeled vs. verified savings: the core distinction

Both numbers answer "how much did this system save," but they're built from opposite directions.

Modeled savings Verified savings
Source The underwriting case, run forward The actual utility bill, reconciled
Inputs Production model × assumed tariff × time Extracted bill vs. reconstructed counterfactual
Tariff Fixed at the assumption made at close The tariff actually in effect that month
What it proves What the system should save on paper What the system did save in dollars
Holds up in audit "We trust our model" "Here is the bill, here is the delta"
Drifts over time Yes — the moment tariff or load changes No — recomputed against reality each month

Modeled savings is a projection. Verified savings is a reconciliation. The distinction matters because the number in most quarterly reports is the first kind presented as if it were the second.

#Why modeled savings was the right tool — at underwriting

It's worth being clear: modeling the savings was correct. At underwriting, there is no bill to verify against — the system isn't built yet. The modeled case is how the deal got sized, financed, and sold to the host and the LPs, and it was the best available instrument for the job. The problem isn't that anyone modeled badly. The problem is that the model was built for a moment that has passed.

A model is a snapshot of assumptions: this production, this tariff, these rates, this load shape. Every one of those was reasonable at close. Each is a hostage to time. The tariff gets revised. The host customer switches plans. Demand charges re-weight against energy charges. Net metering credits step down. None of this means the model was wrong when it was made — it means the report is still quoting a number that was accurate eighteen months ago and hasn't been re-grounded since.

Moving from modeled to verified savings isn't admitting the underwriting was flawed. It's maturing the reporting from "what we projected" to "what we can prove" — the same way a maturing asset class moves from pro-forma to audited results everywhere else in finance.

#Why the footnote fails

The footnote — "savings figures based on modeled estimates" — does its job right up until someone has a reason to look harder. Three moments reliably supply that reason:

  • The audit. An auditor's first question is where the number came from. "Our model projected it" is not a provenance trail. A line-item reconciliation against the actual bill is.
  • The refinancing. A refinancing committee re-underwrites the portfolio on current performance, not the original case. A modeled savings figure that has drifted from reality gets discovered here, at the worst possible time, as a surprise.
  • The dispute. When a host customer or counterparty contests the savings, the modeled number has no standing. Only the bill-grounded number does.

In each case the footnote converts from a routine disclosure into the thing that undermines the report. And it hides a second-order problem: a modeled number can't tell you the shape of a shortfall.

#The question modeled savings can't answer

Ask a portfolio of forty sites whether it's tracking to the underwriting case, and a modeled number says "yes, by construction" — because the model is the underwriting case. It can't report a variance against itself. So three questions an asset manager has to answer go unanswered:

  1. Is the portfolio actually tracking the underwriting case? Modeled savings assumes the answer is yes. Verified savings measures it.
  2. Which sites are off, and by how much? A 5% portfolio shortfall is twenty sites at -2% (a margin issue) or two sites at -50% (an operations failure). These demand opposite responses, and modeled savings can't distinguish them.
  3. What's the audit trail? "We trust our model" versus a per-site, per-month reconciliation traceable to each source bill.

The first framing is a number to report. The second is a number to defend. The difference between them is whether the savings were verified against the bill or assumed from the model.

#What verified savings actually requires

Verified savings is harder to produce than modeled savings, which is why most portfolios still report the model. It requires, for every site, every month:

  1. The actual bill, extracted and line-itemized — energy by TOU period, demand, fixed charges, credits, taxes.
  2. A counterfactual bill, reconstructed against the tariff actually in effect that month — what the site would have been billed with no solar or storage.
  3. The line-item delta between them, broken out by charge type, rolled up to a savings figure.
  4. Variance against the underwriting case, per site, so a portfolio shortfall resolves into which sites and which charge types.
  5. Full provenance — every value traceable to the source PDF line for the actual bill and to the tariff component and load input for the counterfactual.

Done by hand across a portfolio, this is a full-time job, which is why it doesn't get done and the footnote persists. Done with infrastructure built for it, it runs monthly as the bills arrive.

#What Tariform does

Verify produces the verified number. For each site in a C&I solar-and-storage portfolio, every month, it extracts the actual utility bill, reconstructs the counterfactual against the tariff in effect using a US and Canadian tariff catalog and rate engine, and reports the line-item delta — energy, demand, TOU, NEM credits — rolled up to a savings figure with per-site variance against the underwriting case. Portfolio KPIs roll up for the LP report; site-level drill-down answers which sites are off and why. Every number traces to its source: the PDF line for the actual bill, the tariff component and load input for the counterfactual.

If the savings figure in your last quarterly report carried a footnote, the verified number is what replaces it — and what survives the audit, the refinancing, and the dispute the footnote doesn't. Book a demo — twenty minutes, your portfolio, you see the variance breakdown.

#FAQ

Will verified savings make my portfolio look worse than the modeled number?
Sometimes lower, sometimes higher — verification isn't biased in one direction; it replaces an assumption with a measurement. The value isn't the number's size, it's its defensibility and the fact that any variance is located by site and charge type. A shortfall surfaced as a managed line item costs far less than the same shortfall discovered at a refinancing.
Can't I just true up the model once a year instead of verifying every bill?
An annual true-up still runs on modeled assumptions for the eleven months in between, and a single reconciled point can't tell you which sites or charge types drifted, or when. Per-bill verification catches a tariff change or billing error on the first bill it appears on, while it's still cheap to act on — an annual check finds it long after the fact.
Doesn't my monitoring platform's savings number already count as verified?
No. Monitoring derives savings from production and an assumed rate — it confirms the system generated power, but it never compares the actual bill against a counterfactual. That's a modeled figure in a different wrapper. Verified savings is computed from the bill itself; see why a site can produce on plan yet save less.
What do I need to start producing verified savings?
The actual utility bills, the tariff in effect for each site and period, and interval production and battery-dispatch data to reconstruct the load for the counterfactual. With those, each month resolves into an actual bill, a counterfactual, and a line-item delta — the work scales with how cleanly those inputs arrive.
How often should verified savings be produced?
As the bills arrive — monthly. Savings is a per-bill reconciliation, so producing it each month is what keeps variance visible while it's still actionable. Rolling the monthly results up to a quarterly LP figure is then just aggregation, not a re-computation.