How we research

This page describes how the material on this site is put together, so that a reader can judge how much weight any given page deserves. It is written to be checkable rather than reassuring.

Where the facts come from#

Three kinds of source, in order of weight:

• The rules themselves. Where a page states what a Federal Rule of Civil Procedure or Federal Rule of Evidence requires, that is taken from the rule text and its committee notes, not from a summary of a summary.
• Published decisions. Where a page says a court held something, it names the case. A small number of discovery decisions do most of the work in this field, and they are cited by name so a reader can go and read them.
• Vendor documentation. Where a page describes what a product does, that comes from the vendor’s own published documentation, which is the only source that is both authoritative and checkable. It is also marketing, and is read accordingly.

What gets left out on purpose#

• Prices. No page quotes a price for any product or service — principally because, as at November 2026, the major eDiscovery vendors do not publish one. Every pricing page checked routed to a sales conversation. Pages here explain the pricing unit each vendor charges on, which is published and which is the part that transfers between quotes.
• Statistics without a source. No “studies show”, no cost-saving percentages, no market-share claims. The circulating numbers in this field are mostly vendor surveys, recycled past the point where the original can be found.
• Predictions. Particularly about artificial intelligence. What a tool does today is checkable; what it will do in two years is not.

How vendor pages are sourced#

A page about a named product states only what that vendor publishes about itself, read off the vendor’s own live pages on a stated date. Nothing is inferred, nothing is carried over from a competitor’s comparison page, and nothing is tested.

Those pages change often, so each one carries the month it was checked and tells you to verify with the vendor before relying on it. Where a product has been renamed or is mid-rebrand, the page says so rather than picking a name and sounding confident.

How decisions are described#

Case law is easy to overstate, and overstating it is the most common failure in writing about discovery technology. Two distinctions are held to throughout:

• Permitted is not required. A court approving a method in one matter has not made it mandatory in yours.
• Dicta is not holding. Some of the most quoted language about technology-assisted review is commentary rather than decision. Where that is true, the pages say so, because it changes how much the passage is worth.

The limits of this method#

Stated plainly, because a methodology page that lists only strengths is marketing:

• Nothing here is tested hands-on. Capability descriptions are read off documentation, and documentation and behaviour diverge.
• The focus is the United States federal system. State practice and other jurisdictions differ, sometimes fundamentally, and pages that generalise beyond federal practice say so.
• Product capabilities are described at the level of category rather than feature, because feature-level claims are wrong within a release cycle. This makes pages less immediately useful and less often wrong, which is the trade deliberately chosen.
• No page is a substitute for reading the rule, the order, or the protocol that governs your matter.

Review and dating#

Every guide carries its publication date, and a review date when it has been revisited. A page that has not been reviewed recently is not thereby wrong, but the date is shown so the reader can decide. Where a page rests on a fact that is expected to move, it says what would have to change.