Editorial standards · v1.0

How we vet roofing contractors

By George Davis · Published May 3, 2026

Why this exists

Roofing is one of the most heavily-arbitraged categories in local SEO. Search “best roofers in [any small town]” and you will get the same recycled listicle, weighted toward whichever contractor has the most ad budget or the cleverest review-padding firm. The trade is also unusual in how often a homeowner buys under duress — after a storm, with damage they can’t leave alone, talking to people they didn’t choose.

Eaveside exists for one reason: to make sure that when a homeowner in a small town searches for roofing contractors, the first result they read is something written by a person who actually called the contractor, looked up the permits, and confirmed the insurance certificate. We do that against a published methodology so our work is verifiable and so the standard we hold every contractor to is the same one.

The three-tier framework

Every contractor we feature passes through three layers. Tier 1 is binary — fail any item and you are excluded entirely. Tier 2 is a weighted score across six research dimensions. Tier 3 is a set of context signals we publish for reader judgment but do not roll into the score. Every layer is the same for every contractor, in every market.

Tier 1 · Hard filters

A contractor that fails any one of these is not featured. There is no “close enough.” This filter, by itself, excludes the great majority of contractors targeting tier-3 markets after storms.

1. Valid state license for the size and type of work being performed. Some states (including Arkansas) require a contractor’s license at $50,000 of project value; some states have no statewide license. We document what the state requires, then verify the contractor meets it for the work they advertise.
2. Active general liability insurance, $1,000,000 minimum, verified via Certificate of Insurance with a callable agent. We do not accept a screenshot. We call the agent listed on the COI to confirm coverage is in force.
3. Active workers’ compensation insurance where state law requires it. In most states, this triggers at three or more employees; lower thresholds apply in some states. We verify both the carrier and the policy status, again by calling the agent.
4. No recent serious court judgments or unresolved mechanics liens against the contractor or its principals. We check state court systems, the county recorder, and PACER for federal matters. Single-incident small-claims activity does not exclude; pattern litigation or unresolved liens does.
5. Physical office in or near the market. No virtual offices, no UPS Store mailbox addresses, no recently-relocated (within six months) entries from out of state into a storm-affected market. Of every Tier-1 filter, this is the one that excludes the most contractors in small towns after a storm.

Tier 2 · Weighted scoring

Contractors that clear Tier 1 are then scored across six research dimensions, weighted as follows. Scores are normalized to a 100-point scale and translated to a letter grade. We feature only A and B graded contractors in the article body; C-grade contractors are listed below the fold with their score disclosed; below C is excluded.

Online reputation, in detail

This is the single largest input to the score, and the one most often misread. A 4.9 average across 800 reviews is not the same as a 4.9 average across 800 reviews if half of those reviews were posted in a 30-day burst by accounts with no other history. We score this dimension across six sub-criteria:

• Velocity — steady accumulation over years vs. a spike in any one quarter.
• Reviewer-account history — accounts with one review and no other activity are treated as suspect; long-tenured accounts with reviews of unrelated businesses count more.
• Language patterns — specific to the job (“chimney flashing,” “ridge vent”) vs. templated marketing language (“professional, on time, would recommend”).
• Negative-review handling — professional engagement, problem acknowledgment, attempted resolution, vs. silence or hostility.
• Distribution shape — a normal review distribution has some 1- and 2-star reviews. A perfectly uniform 5-star distribution is statistically improbable for any business with real volume.
• Cross-platform consistency — Google, Yelp, Facebook, and Nextdoor should agree directionally. Sharp divergence (5.0 on Google, 1.5 on Yelp) is a flag.

The phone-call rubric

We call every contractor we feature, identify ourselves, and run a standard call. The rubric is six dimensions of five points each, normalized to 100. We publish the duration, the date, and the resulting score on every contractor profile.

1. Communication clarity — answers the question that was asked.
2. Transparency — gives a real answer to a hard question.
3. Sales pressure — scored inversely; less is better.
4. Technical knowledge — discusses the trade with specifics, not slogans.
5. Willingness to provide COI — sends within one business day, with a callable agent.
6. Willingness to pull permits — clear, immediate yes; will do the work in their own name.

Tier 3 · Context flags (disclosed, not scored)

These signals matter for reader judgment but we deliberately exclude them from the composite score. Some are too lossy to weight fairly (BBB grades correlate with paid accreditation more than with quality); some require homeowner context to interpret (a sub-contracted crew model is fine for some jobs and a deal-breaker for others).

• Years in business, cross-referenced for prior entities under the same owner. A “new” LLC with five dead LLCs behind it under the same Secretary of State filings is disclosed as such.
• BBB record, with explicit caveat that BBB grades are partly pay-to-play. We weight complaints-per-year over the headline letter grade.
• Subcontracting model — W-2 crews vs. subcontracted vs. hybrid. Disclosed for transparency.
• Storm-event correlation — was the LLC formed within twelve months of a major local hail or wind event? Flagged for reader judgment.
• Peer reputation — once two or three contractors are vetted in a market, we ask each one which other contractors in the market do good work. Strong positive recommendations are tracked but sources are protected.
• Prior entity history — Secretary of State filings showing the same operator behind multiple successive LLCs.

The storm-chaser problem

The single highest-impact failure mode for a homeowner in a tier-3 market is hiring a storm-chasing contractor. The pattern is well-known inside the trade and almost completely unknown to the homeowners on the receiving end of it. After a hail or wind event, contractors from out-of-state form temporary LLCs, lease short-term office space (or list a UPS Store as their address), hire door-to-door canvassers, and disappear within twelve to eighteen months — well before workmanship-warranty claims come due.

Our framework filters for this pattern explicitly. The Tier-1 physical-office filter is the single most powerful exclusion: most storm-chasers cannot meet it. Of the contractors that get past it, the Tier-3 storm-event correlation flag and the prior-entity check catch most of the rest. The result, in practice, is that contractors featured on Eaveside are almost without exception local operators with multi-year roots in the market.

We’re explicit about this in every market article. The storm-chaser warning is not a generic disclaimer — it is the single most useful thing we can tell a homeowner who has just been knocked on by a stranger in a branded shirt.

What we cannot verify

We are deliberate about the limits of this work. Our framework is a screen, not a guarantee. The following are out of scope for us, and a homeowner should treat them as their own due diligence:

• The quality of any specific job once it is in progress. We do not visit job sites.
• Whether a long-term workmanship warranty will be honored 10 years from now. We have no way to predict business continuity that far out.
• The character of the specific crew that arrives at the homeowner’s house. Even contractors with strong rubric scores have variance in crew quality.
• Whether the homeowner’s specific situation (insurance terms, deductible structure, local code requirements) maps cleanly to a contractor’s typical job.
• The actuarial value of any given warranty. We document what is offered; we do not value it.

When a homeowner reads Eaveside they should treat our research as a screen that eliminates the most common bad actors and surfaces a short list of legitimate operators. They should still call references, ask for an itemized estimate, confirm their insurance terms with their carrier, and pull at least one second opinion on scope.

Conflicts of interest

Eaveside is operated by a roofing-industry insider with a related business interest. The detailed disclosure, including how we handle articles that mention contractors connected to the operator, is on the disclosure page. The short version: contractors connected to George Davis or his family are evaluated under the same framework, are reviewed by an independent reviewer when featured, and the connection is disclosed prominently on the relevant article and contractor profile.

How we make money

As of May 3, 2026: zero revenue. Eaveside is funded by the founder’s separate operating business. The future business model — to be transparent — is two-fold: (1) consumer-side lead generation paid by featured contractors on a fixed-rate basis, and (2) referrals to RoofingLogic, a separate CRM product the founder operates. The bright lines are these:

• We do not accept payment for inclusion. A contractor cannot pay to be featured.
• We do not accept payment for ranking position. A contractor cannot pay to move up.
• Lead-gen pricing, when it exists, will be flat per-lead and identical for every featured contractor in a market.
• We do not surface contractors who refuse the lead-gen offer at lower positions. The offer has no impact on the article order.

Corrections policy

If anything on Eaveside is factually incorrect, we want to know. Email editor@eaveside.com with the URL, the specific claim, and any documentation you have. Substantive corrections are made within seven days, are noted on the article with the date and nature of the correction, and the article’s Last updated metadata is refreshed accordingly. We do not silently edit articles.

Methodology changelog

• v1.0 · May 3, 2026 — Initial publication. Three-tier framework (5 hard filters, 6 weighted criteria, 6 context flags), letter grades A/B/C, phone-call rubric.

Last updated May 3, 2026. Methodology v1.0.