How Vatra evaluates an energy decision
Vatra ranks the upgrades a specific home could make by a published, weighted score. Every input is either cited, or marked as an assumption that has no citation yet, and the plan says which.
The short answer
For a given address, Vatra builds a list of candidate upgrades, checks whether each one is possible for that property, estimates its cost and return from cited public data, and ranks the candidates by a weighted score of 7 factors. Rank is not payback alone, and nothing outside those factors feeds the ranking.
The steps
- Eligibility first. Each candidate is checked against the property before it is scored — for example, a renter cannot buy a roof, and an old roof puts solar behind a roofing note. A candidate that fails is shown with its reason, not silently dropped.
- Cost and return. Costs, savings fractions, financing terms and degradation come from a single registry of assumptions, each with its source. Solar production is estimated with the U.S. Department of Energy's PVWatts model.
- Scoring. Each candidate gets a score from the weighted factors below, and the plan is ordered by that score.
- Disclosure. The plan shows what each factor contributed and lists every assumption it used, with its source or its lack of one.
What the ranking weighs
| Factor | Weight | Where its value comes from |
|---|---|---|
| Financial return (10-yr) | 30% | Calculated for this property |
| Monthly cashflow, year 1 | 20% | Calculated for this property |
| Payback period | 15% | Calculated for this property |
| Comfort | 10% | A fixed rubric value per upgrade type, the same for every home |
| Backup / resilience | 10% | A fixed rubric value per upgrade type, the same for every home |
| Your stated preference | 10% | Your answer to the survey |
| Sourcing confidence | 5% | Share of inputs that carry a citation |
Two things in that table deserve saying plainly. Comfort and backup are fixed values per type of upgrade, not measurements of your home. And one factor rewards how much of a candidate's input data is actually cited, so an estimate built on placeholders ranks lower than one built on sources.
Sourced data versus assumptions
Every input the engine uses lives in a registry — one shared file, plus one file per market for local rates — and each entry has one of three statuses:
| Status | Entries | What it means |
|---|---|---|
| Sourced | 13 | Has a citation that can be checked, with the date it was read. |
| Needs a source | 2 | A placeholder. It is shown with a visible “unsourced” badge, and a plan that depends on it is marked as a draft. |
| Design choice | 15 | A product decision, such as where a scoring scale starts and ends. Not a fact about the world, and labelled as one. |
Known limitations
- Roof geometry is assumed, not measured. Solar estimates use a fixed tilt and a due-south orientation for every property. A roof facing any other way will usually produce less. Both values are listed in the registry as needing a source.
- Equipment costs are national estimates. Several installed-cost figures come from published cost surveys rather than local quotes, and the registry says so for each one. A local quote is better evidence than any of them.
- Coverage is limited. Plans are produced for New Jersey and Texas (addresses inside Austin Energy's service area only), and an address elsewhere is told so rather than given a guess. A Texas address gets a plan only if it falls inside Austin Energy's published service-area map and not within 250 metres of its edge, where a neighbouring utility may serve it instead. A ZIP code is never used to decide.
- The utility is estimated, not verified. Which utility serves an address is decided by locating the address with the US Census Bureau geocoder and testing that point against the utility's published service-area map. That is an estimate, not a check of the homeowner's utility account. A geocoded point is interpolated along the street and can sit some distance from where the property's service actually connects, which is why addresses near a boundary are left out rather than guessed. A homeowner's own electric bill is stronger evidence of who serves them than anything Vatra can infer.
- A bill credit is counted only when it is used. Where a utility credits solar production against the bill — Austin Energy's Value of Solar — Vatra runs the credit month by month against the homeowner's reported electric bill, using PVWatts' monthly production. Only credit applied to electric charges counts as a saving. Credit left over carries forward, but it has no cash value, does not pay water, trash or other city services, and ends if the account closes, so it is counted as $0. Without a monthly production profile, no solar figure is produced.
- Estimates, not guarantees. Nothing here is tax, legal or investment advice, and no figure is a forecast of a particular bill.
How the research pages are written
Articles in Vatra Insights are drafted from public primary sources, list every source with the date it was read, separate what the evidence shows from Vatra's reading of it, and are published only after a person has reviewed them.