CPR benchmark methodology

This document describes how the cost-per-signed-retainer figures published at /benchmarks/ are produced. It covers the current release, Q1 2026, current as of March 31, 2026 and running under method version 1.0. It is published on its own URL so it can be cited independently of any article, and any change to the calculation is logged with a date in the changelog at the bottom of this page.

The sample

The benchmark panel is Mass Tort Marketing Agency’s active client book: roughly 140 active plaintiff-side firms representing approximately $180M of annual mass tort lead spend across 16 active tort categories, of which 14 are published in the current table. The underlying performance data is anonymized.

The figures are not drawn from the panel alone. As originally disclosed, the benchmark ranges are drawn from the panel’s performance data cross-referenced against published industry surveys and practitioner-disclosed benchmarks. Those external sources are not named in the published methodology, and how much weight they carry relative to panel data is not specified. This is listed again under limitations.

The time window for the current release is Q1 2026 — 2026-01-01 through 2026-03-31 — and figures are stamped as of March 31, 2026, published April 24, 2026.

How firms enter and leave the panel is not currently documented. The panel is the client book, so a firm enters when it becomes a client and leaves when it stops being one, but there is no published inclusion rule — no minimum spend, no minimum tenure, no minimum number of signed retainers before a firm’s data counts toward a band. That is a real gap and it is listed again under limitations. Until an inclusion rule is published and applied, readers should treat panel composition as varying between releases in ways this document does not describe.

The calculation

Cost per signed retainer is cost divided by signed retainers that survive the 30-day falloff window (see definitions below).

The method, exactly as disclosed: benchmarks are computed as interquartile ranges after removing outliers above the 95th percentile and below the 5th percentile, to avoid distortion from either catastrophically underperforming or anomalously high-performing campaigns. An interquartile range runs from the 25th percentile (the published low bound) to the 75th (the published high bound).

What the published method does not specify. It refers to trimming campaigns but does not otherwise define the unit of observation — whether a data point is a campaign, a firm, or a firm-tort pairing. It does not say how the national median is computed from the per-tort data, or how the external surveys and practitioner benchmarks it is cross-referenced against enter the calculation. None of these are described here, because describing them would mean guessing.

One consequence follows from the definition alone and is worth stating: a published band is not a min-to-max range. Roughly a quarter of observations sit above its high bound, so operating above it does not make a firm an outlier. And because the bands are quartile boundaries rather than a dispersion estimate, they are not error bars and must not be read as such.

Definitions

Signed retainer

A retainer that has been signed, has cleared medical-record verification, and where the claimant remains engaged with the firm 30 days later. A raw signed count taken before the falloff window closes overstates real performance by 15–28% on average; that inflated figure is not what any number on this platform measures.

Qualified lead

Each tort’s qualifying criteria are published in plain language on its own benchmark page. A general definition of a qualified lead — including whether consent verification is a condition — is not defined in the published methodology.

Qualifying spend

Not defined in the published methodology. The numerator is described only as spend within the panel’s approximately $180M of annual mass tort lead spend. Which cost lines it includes (media, lead purchases, intake staffing, record retrieval) and how spend is attributed across quarters are not stated.

Falloff

A signed retainer that does not survive the 30-day window. Because the denominator counts only retainers that survive it, falloff is already netted out of every published CPR rather than reported beside it. Note that for some torts the source figure reported as falloff is a qualification-failure rate at medical-record review; where that is so, the tort’s page and exports say which.

Exclusions

The trimmed tails above the 95th and below the 5th percentile. Separately, the panel spans 16 active tort categories and the table publishes 14; 2 are not published, and which ones and why is not disclosed. No other exclusions are stated in the published methodology.

Status

Whether a litigation is active, paused, settled, or closed to new claimants, as of the release date. Every tort in the Q1 2026 release is active. Status is published as stated in the release and is not independently verified against MDL dockets.

Known limitations

Read these before citing anything on this platform. They are not boilerplate.

1. The publisher sells the product being priced. Mass Tort Marketing Agency sells mass tort leads. Every incentive that normally makes a vendor’s own pricing benchmark suspect applies here, and nothing in this methodology neutralises that. It is the reason third-party verification matters more for this dataset than for most.
2. The panel is a client base, not a sample. These 140 firms were not drawn randomly from the population of plaintiff-side firms. They chose to work with this agency, which means the panel likely over-represents firms that buy leads from agencies at all, and under-represents firms that generate their own or buy exclusively direct. No weighting is applied to correct for that.
3. Unnamed external sources are blended in. The ranges are cross-referenced against published industry surveys and practitioner-disclosed benchmarks that the methodology does not name, in a way it does not specify. A reader therefore cannot fully separate what the panel measured from what those sources contributed.
4. The data is self-reported by the firms. Signed counts, falloff, and spend come from panel firms’ own reporting. It is not audited against their case-management systems or bank records, and firms differ in how rigorously they classify a retainer as signed.
5. No third party has reviewed this. The method has not been reviewed and the underlying data has not been audited by anyone outside the agency. See verification.
6. Per-row sample sizes are not published. The panel total is published, but how many firms stand behind any single tort’s band is not. A band computed from eight firms and one computed from eighty are presented identically. This is the single most consequential gap on the platform.
7. No confidence intervals, and no per-tort medians. Both for the reason given above: the published dataset retains quartile boundaries, not the observations behind them.
8. Panel entry and exit rules are undocumented. There is no published minimum spend, tenure, or volume threshold for a firm’s data to count toward a band, so composition may shift between releases in ways this document does not describe.
9. The time series is two points, five quarters apart. See historical gaps.
10. 2 of the 16 panel categories are not published. 16 active tort categories are in the panel and 14 appear in the table. The unpublished categories are not named and the basis for leaving them out is not disclosed. A reader cannot currently rule out that the published set is the more favourable one.
11. One published figure conflicts with another. The Roundup signed rate is published as a 12–17% band in the table while the accompanying analysis states it declined to 11% by Q1 2026. Both are reproduced as published and neither has been adjusted to fit the other. The conflict is recorded on the trends page.

What is missing from the historical record

Listed rather than smoothed. Nothing below has been reconstructed, estimated, or interpolated to fill a gap.

• Q1 2025 through Q4 2025 are not in the dataset. Five quarters separate the two points held and the intervening four are absent. The 24% change between Q4 2024 and Q1 2026 is therefore a five-quarter change and must not be described as quarter-over-quarter.
• Q4 2024 has no per-tort breakdown in the dataset. Only the national median is held for that quarter, which is why every per-tort series starts at Q1 2026.
• Q4 2024 panel composition is unknown. The 140-firm, $180M description applies to the Q1 2026 panel. Whether the Q4 2024 median was computed on a comparable panel is not documented, so the two points may not be strictly like-for-like.
• Several stated movements have no published baseline. Camp Lejeune is described as having compressed 18–25% from a 2023 peak, Depo Provera as having risen 12–18% from 2024, and Roundup’s signed rate as having fallen from 13% in Q1 2023. None of those prior values is published as a data point, so each is recorded as a stated change rather than as a second point in a series.

Snapshots, exports, and licence

Every table on this platform is generated from one data file, so an export can never disagree with a rendered page. latest.json and latest.csv always carry the current release and will change when a new quarter publishes.

Dated snapshots do not change. If you are citing a figure in something durable, cite a snapshot URL:

• /benchmarks/data/2026-q1.json · .csv
• /benchmarks/data/2024-q4.json · .csv

Every file embeds its release, its data-as-of date, its methodology version, a link back to this page, and the vendor disclosure. Null values in JSON and empty cells in CSV are never zeros and never estimates — each is paired with a reason field explaining why the value is unavailable.

The data is published under CC BY 4.0: reuse it, including commercially, with attribution to Mass Tort Marketing Agency CPR Benchmark, https://www.masstortmarketingagency.com/benchmarks/.

Third-party verification

The current status is unverified. No academic, bar-association research arm, or accounting firm has reviewed this methodology or examined a sample of the underlying data. Saying so is not a formality: a vendor publishing benchmarks for the product it sells has a credibility problem that only an outside reviewer can solve, and the platform is built to show that honestly until one exists.

The verification status is a field in the dataset, not a line of copy. It takes one of three values — unverified, internally audited, or third-party reviewed — and it appears in latest.json, in every CSV header, and beside the headline stat on every data page. When an attestation exists it will name the reviewer and appear next to the national median, not in a footer.

A one-page proposal for an independent review — what would be shared, what would be attested, and what the attestation would say on the page — is drafted and held in the repository at docs/benchmarks-third-party-verification-proposal.md for the team to act on. No reviewer has been engaged and no target date is committed here yet — a date will be published on this page once a reviewer is approached, rather than announced in advance of one.

Changelog

Every change to the calculation, the definitions, or the disclosed figures is logged here with a date. Current method version: 1.0.

1. 2026-09-20·method v1.0

   Methodology moved onto its own URL and versioned

   The methodology was previously a closing section inside the Q1 2026 benchmark article. It now lives at this URL so it can be cited independently of any one article, and carries a version number. No figure and no calculation changed in this move. Per-row sample sizes, per-tort medians, and confidence intervals were disclosed as unavailable for the first time, having previously been absent without explanation.

2. 2026-06-10·method v1.0

   Q1 2026 figures reviewed, unchanged

   Scheduled review of the Q1 2026 release. No figures changed.

3. 2026-04-24·method v1.0

   Q1 2026 release published

   National median cost per signed retainer of $3,850 across 14 active tort categories, computed on a panel of roughly 140 plaintiff-side firms representing approximately $180M of annual mass tort lead spend across 16 active tort categories. Benchmarks published as interquartile ranges after trimming observations above the 95th and below the 5th percentile. This is the first release under version 1.0 of the method.