How the Probative benchmarks are made and protected

In short. The Probative benchmarks, 2025–2026 are Probative’s own data, shown as bands, tiers and rounded values, not exact counts. Each value is aggregated across many observations, is shown as a band, a named tier or a rounded value, and is left out when it rests on too few events or too few months. This page explains each step and defines every term the data pages use.

Probative benchmarks, 2025–2026: aggregated, rounded, and small groups suppressed.

What do the terms on the data pages mean?

• Inquiry. A call, form, chat or text that reaches a law firm.
• Qualified. An inquiry that was screened against a firm’s written case criteria and accepted as a possible case.
• Consult. A qualified inquiry with a consultation or signing appointment booked.
• Signed client. A person who signed an engagement agreement. A signed client counts as a signed case.
• Dropped. A signed matter that later ended before it was resolved.
• Action rate. The share of visitors who take an action on the page, such as a call, a form, a chat or a booking.
• Outside business hours. Any time other than a firm’s posted weekday opening hours, including evenings, nights and weekends. All other times are business hours.
• Weekday and weekend day. A weekday is a day on which a firm posts opening hours. Every other day is a weekend day.
• Early reading and final figure. An early reading is a figure taken soon after the period ends. The final figure is the same figure once late records have arrived. The multiplier is the final figure divided by the early reading.
• Home page, practice pages, location pages and blog posts. The home page is a site’s front page. A practice page covers one area of law, a location page covers one place, and a blog post is a dated article.
• Cost per click. Ad cost divided by the clicks it bought. Click-through rate is clicks divided by the times an ad was shown.
• Ad cost per signed client. Ad cost divided by the signed clients that came from those ads. It leaves out staff time and agency fees.
• Calendar quarter. The first quarter is the first three months of the year, and so on. Quarters are compared after the overall growth trend is taken out.
• Lead credit. A Local Services Ads lead that the advertiser disputed and the platform credited back.
• Band. A range of values on a fixed scale. The value column of a file holds the middle of the band, and the note holds the band itself.
• Tier. A named range on a fixed ladder, always shown with its range in brackets. See the ladders below.

What do the numbers measure?

Each table says what it measures. Together, the tables measure:

• How quickly inquiries become signed clients. The share of signed clients who had signed by a given point after the inquiry, and the typical (median) time to sign.
• Phone calls and signed-client likelihood. How likely a call is to become a signed client, by call length and for calls outside business hours, compared with the average call.
• AI answers and organic search clicks. The drop in click-through where ranking held, and the action rate of visitors who arrive from AI assistants compared with visitors from organic search.
• Pace of signed clients. Signed clients per weekend day compared with a weekday, and the busiest and quietest points compared with the average.
• Signed by the next day, by source. The share of signed clients who signed by the next day, for each source, compared with the average across all sources.
• Cost per conversion by search type, relative to average. Cost per conversion for all searches divided by cost per conversion for one type of search. Above one is cheaper than average.
• Action rate by page type. The action rate of one page type divided by the action rate of another.
• Share of inquiries that became signed clients, by source. Signed clients divided by inquiries, before any are later dropped. Shown as a band.
• Inquiries outside business hours. How many calls and web forms arrive outside business hours, and how those calls compare on missed calls and signing.
• Signed-client likelihood by source, relative to average. How likely an inquiry from one source is to become a signed client, compared with the average across all sources.
• Ad-platform conversions per signed client. Conversions an ad platform reports, divided by signed clients from that platform. Shown as a tier.
• Local Services Ads compared with search ads. Cost per signed client and lead credits, as ratios.
• Signed matters later dropped, by source. The share of signed matters later dropped, for each source, compared with the average across all sources.
• Early readings compared with final figures. How far an early reading of signed clients sits below the final figure.
• What tracking can and cannot see. How many signed clients can be matched back to an ad click.
• What search-ad clicks cost. Cost per click for injury search ads and for all law-firm search ads, the click-through rate and the share of clicks made on a phone.
• What people type. The share of search-ad clicks from searches that say “near me”, are written in Spanish, ask for the best lawyer or a free consultation, are phrased as a question, or name a case type.
• Calls from ads. How many calls from ads are missed, how long answered calls last, and how missed calls outside business hours compare.
• Local Services Ads leads. The mix of phone and message leads, and the share charged and credited back.
• Inquiries by calendar quarter. Inquiries in each quarter compared with the average quarter.
• Ad cost per signed client, by channel. For search and video ads, and for Local Services Ads, as ranges.

What do we publish, and what do we never publish?

We publish bands, tiers and rounded values. Each table is its own file.

We never publish counts of any kind, total spend, fees, settlements or case values, dates (time appears only as parts of the year or calendar quarters), places below the national level, or names of any kind. Costs appear only as bands or ranges: cost per click and ad cost per signed client.

How are the values protected?

1. Aggregation. Every value is aggregated across many observations. No value is shown that rests on too few of them or mostly on one.
2. Suppression. A value that rests on fewer than 30 events, or on fewer than 3 months of data, is withheld. An event is a call, a lead, a click or a signed client. The value is left out of the tables and files, and out of any combined value. Some tables therefore have fewer rows than you might expect.
3. Banding. Many values are shown as a band. The value column holds the middle of the band and the note holds the band itself. A band is a range, not a point.
4. Tiers. The most specific values are shown only as a tier. A tier is a named range on a fixed ladder, and its range is always shown in brackets beside its name. The ladder is set before any value is read and is the same for every row of a table. Every tier is wide enough that the value inside it cannot be read back out.
5. Rounding. Costs use $25 or $50 bands and a cost per signed client a $500 range; shares use whole percents or 5-point bands. Where a rounded value would still be too precise, it is shown as a wider band instead.
6. Close values. Values that sit close together may be shown level, or in either order. Two rows in the same tier are level. Do not read a rank from two values in the same band or tier.
7. Grouping. Some rows combine several smaller groups into one broader group, such as search and map advertising or unpaid web and map listings. Group boundaries follow the standard groupings on this site.
8. Release. A page carries one table and at most three figures. Tables that break results down by source are never put on the same page or in the same file.
9. Tables kept from version 3. The speed-to-sign, calls-and-hours, AI-answers, action-rate-by-page and calls-outside-hours tables, and the first rows of the early-readings and tracking tables, are kept as first published. Before they were banded or tiered, their values were moved by a small bounded amount. Values added or recomputed in version 4 are not moved.

Which tiers do the tables use?

The ladders are set in advance. The tier word compares a row with the reference named in its note: the average across all sources, the average call, or the group the row is divided by. On a share ladder the word only places the range from lowest to highest, and it is not a comparison with a published average.

Conversions an ad platform reports per signed client are shown as ranges, such as under 1.5×, 6–12× and 12× or more. The top range is open.

Tiers are approximate. Treat every tier as plus or minus one step: a row may sit one tier above or one tier below the tier printed. Tables kept from version 3 had their values moved within a bounded window before they were tiered, so the printed tier is a guide to size, not a measurement.

A worked example. A band of 10–15% means the published figure sits in that range. It is not a measured 12.5%. A tier of above average with the range 1.3–1.7× means the row sits in that range compared with its reference. It is not a measured 1.5×.

What can the numbers tell you, and what can’t they?

• They are observational. They are not a controlled test, and they do not show cause.
• Compare like with like. Practice area and case mix move results more than channel does.
• Ad platforms change their definitions. A platform’s conversion is not a signed client.
• Recent signings are understated in any figure about dropped matters, because drops take time to appear.
• The figures are not forecasts, and your own numbers may differ. Use them to see where yours sit, not to promise a result. The scorecard tool does this for you without storing what you type.

How are the benchmarks versioned?

Versions are numbered by meaning: a major change when a measure or a protection rule changes, a minor change when rows are added, a patch change for corrections.

• Version 4.0.1 is the current release. It widens one value to a band for privacy.
• The previous release added six tables as free files: search-ad click costs, what people type, calls from ads, Local Services Ads leads, inquiries by quarter and ad cost per signed client. It recomputed the by-source, pace, search-type, platform, Local Services Ads and tracking tables under the aggregation rule, and withdrew three tables on case value.
• The release before that published the first three tables as free files.

How do I cite them, and what is the licence?

Probative Growth (2026). Probative benchmarks, 2025–2026 (v4.0.1) [Data set]. CC BY 4.0.

The licence is Creative Commons Attribution (CC BY 4.0). You may copy, share and adapt the tables with credit to Probative Growth. The files are on the benchmarks page.

What if I spot an error?

Tell us through the corrections page. Probative is a marketing company, not a law firm or a lawyer referral service. Nothing here is legal advice.