Every capital plan, every wildfire mitigation filing, and every resource decision a utility makes rests on a model. And every model rests on a quieter assumption: that the data describing what’s in the field is accurate. For a large share of the distribution system, it isn’t. And that gap doesn’t stay at the bottom of the stack. It propagates upward, into the plans and the dollars that depend on it.
This isn’t a knock on planners. It’s a structural problem with where the data comes from and how rarely it gets verified against the physical asset. And in the current western regulatory environment, it’s becoming an expensive one.
The model is only as good as the field record behind it
A distribution model is an abstraction of copper, steel, and concrete that exists in the real world. The model is only as good as the records it’s built from: the GIS layers, the asset registries, the as-builts. When those records drift from field reality, the model inherits the error and presents it as fact.
The Electric Power Research Institute documented this years ago and it remains true. Utilities’ multi-billion-dollar smart grid investments have, in cases, failed to yield expected returns because their distribution-system data does not accurately represent what is actually in the field.¹ The investment was sound. The data underneath it wasn’t.
The point is uncomfortable but simple. A model can be mathematically perfect and still be wrong, because it’s faithfully computing on inputs that don’t match the poles and conductors they claim to describe.
What “wrong” actually looks like
This isn’t vague. The errors are specific and well-catalogued. The National Renewable Energy Laboratory has identified feeder-model development as the most error-prone stage of distribution analysis, with recurring problems including mislabeled phases, missing or incorrect conductor types, incorrect voltage-regulator settings, and load data that’s estimated rather than measured.²
Read that list again, because each item is a planning landmine:
• A mislabeled phase skews load-balancing and hosting-capacity analysis.
• A missing conductor type means thermal ratings and upgrade triggers are computed against the wrong physical limits.
• Incorrect regulator settings distort voltage modeling across an entire feeder.
• Estimated load carried as if it were measured quietly bakes uncertainty into every downstream number.
None of these are exotic. They’re the ordinary residue of decades of field work that updated the asset but never fully cascaded back into the system of record. The asset got upgraded, re-rated, or replaced, and the database didn’t always hear about it.
The cost compounds as it climbs
Here’s why this matters beyond the engineering team. The error doesn’t stay local. It moves up the planning stack.
Bad field data feeds the distribution model. The distribution model feeds the capital plan and the resource plan. The capital plan feeds the rate case and the wildfire mitigation filing. By the time the number reaches a regulator, the original field discrepancy has been laundered through three layers of analysis and now wears the authority of a planning document.
That’s not a small exposure given the dollars in motion. Industry grid-modernization spending ran to roughly $36.4 billion between 2018 and 2023, growing at a 37% compound annual rate.³ When the data foundation under that spend is unverified, a meaningful fraction of it is being allocated against a picture of the grid that’s partly fictional.
And there’s a hard compliance edge to it. Under NERC’s facility-ratings standard, discrepancies between documented ratings and actual field conditions create direct violation exposure, with statutory penalties of up to $1 million per day, per violation under the Federal Power Act.⁴ A stale asset record isn’t just a planning inconvenience at that point. It’s a regulatory liability sitting in your database waiting to be found.
Ground truth, then anything else
The fix isn’t a better model. It’s better inputs. And there’s only one way to get them. Someone who knows the asset has to go look at it, in the field, and verify the record against reality.
That’s the whole premise of field-verified data: an inspection performed by people who’ve worked the assets they’re inspecting, with findings that trace cleanly back to a specific structure and a specific condition. Not a model’s best guess. Not a desktop estimate. The ground truth, captured once, correctly, and carried through to closeout.
It’s the unglamorous part of the work, and it’s the part everything else depends on. You can’t optimize a plan built on inputs you haven’t verified. You can only compound their error and hope the regulator doesn’t notice. Lately, regulators are noticing.
The sequence matters. Ground truth comes first. Then the model. Then the plan. Then the dollars. Run it in that order and the spend is defensible all the way down. Run it backward, model first and verify never, and you’re betting a capital program on a database nobody has checked against the field.
Which raises the question worth sitting with: when was the last time the data under your largest planned investment was actually verified against the asset it describes, rather than modeled or estimated?
Notes
1. Electric Power Research Institute, Report 1024303 (2012).
2. National Renewable Energy Laboratory / IREC, data-validation methodology for Hosting Capacity Analyses (2022).
3. Wood Mackenzie, U.S. grid modernization spending, 2018–2023 (approximately $36.4B; 37% CAGR).
4. NERC Reliability Standard FAC-008-5; penalty authority under Federal Power Act §215 (up to $1,000,000 per day, per violation).
