The evidence-grading framework.
Most reports bury methodology at the back. In a research publication it belongs at the front, because it's the thing that makes every number that follows either trustworthy or worthless. Here is exactly how to read this report.
The three inputs
Published 2025–2026 third-party benchmarks (legal PPC/Meta/LSA costs, response-time research), each footnoted to source in §References.
WorkersComp Marketing's real, aggregated, de-identified campaign data: $16.5M spend, 219,000 leads, 21,000 signed cases.
Our interpretation of observable, documented industry shifts (e.g. AI Overviews, zero-click search) — reasoned, and labeled as analysis.
The four evidence grades
Every finding in this report carries one of these grades. Weight each finding accordingly — and notice that a responsible research publication has plenty of the lower two.
| Grade | Meaning | How to weight it |
|---|---|---|
| Established | Backed by cited third-party data and/or large-sample portfolio signal, consistent across sources | Act on it with confidence |
| Emerging | Directionally clear from real signal but still developing; limited or single-source data | Act, but monitor and revisit |
| Contested | Real disagreement or high variance across firms/sources; the honest answer is "it depends" | Test in your own context |
| Data-gap | The industry does not yet measure this well; our statement is a hypothesis, not a fact | Treat as a question, not an answer |
A word of intellectual honesty. A true statistical confidence interval requires a random, representative sample survey — which this inaugural edition is not. Rather than print fabricated intervals to look rigorous, we grade evidence qualitatively and tell you the source of every number. That is the more honest form of rigor. A formal, sampled industry survey — with real sample composition and confidence intervals — is the headline addition planned for the next edition. Until then, the evidence grades are the calibration.