This is the part of the discipline nobody writes about. Nine pages currently rank for the search that most likely brought you here, roughly twenty-five thousand words between them, and not one mentions the rule of evidence that decides whether its author may testify at all. Meanwhile every practical failure I have seen in a search matter happens somewhere in the sixteen entries below, and not one is a failure of search expertise.
The data expired before anyone thought to ask for it. The chart was an illustrative aid and never reached the jury room. The archive capture was dated to a year instead of a day. The causation opinion outran the method meant to support it. Four failures, four rules, one thing in common: each was decided long before trial by someone who did not know the rule was there.
It is a sequence, and that is the whole argument
Search evidence has to survive four stages, in order. Fail one and everything spent on the others is wasted.
First, preservation
Search evidence deletes itself on a timer, and no bad faith is required. Google Search Console — the reporting property a site owner verifies, which records the queries, impressions and clicks the engine logged for that site — holds sixteen months of performance data on a rolling basis. Raw server access logs are commonly rotated every thirty to ninety days.
An ordinary litigation hold reaches none of that: nobody's mailbox contains Search Console data. Preservation here takes an affirmative export from a verified property, a change to a retention setting, and an instruction to server administrators to halt log rotation — and the duty attaches when litigation becomes reasonably foreseeable, not when it is filed. FRCP 37(e), amended in 2015, asks whether reasonable steps were taken and whether the information can be restored or replaced. I have not located a reported decision applying it to analytics data or log rotation, so that entry is written from the rule text and says so.
Second, authentication
A screenshot proves nothing until a witness can say what it is, where it came from, and when it was taken. FRE 901(a) sets a lower bar than most people assume — evidence sufficient to support a finding that the item is what its proponent claims — and the same low bar admits the other side's screenshot. What fails is the record behind the exhibit: a capture with no time zone; a capture from one browser in one location offered as proof of what the engine displayed generally, when results vary by location, device and time; an archived page offered for a proposition that needs a specific date, where the archive establishes only the year.
There are routes through this — a witness with knowledge under 901(b)(1), comparison under 901(b)(3), and the process-or-system route in 901(b)(9), the natural home for crawl output and log exports. FRE 902(13) and 902(14), added in 2017, allow certification in place of live testimony, with 902(14) resting on a hash value computed at the moment of capture. That turns a foundation fight into paperwork, and almost nobody here bothers to generate one. Every route depends on a decision made at collection.
Third, admissibility
An expert in this field is a Kumho expert. Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999), extended the gatekeeping obligation beyond scientific testimony to technical and specialized knowledge, and held the Daubert factors to be neither mandatory nor exhaustive. Search opinion is not a peer-reviewed science with error rates; it is experience-based technical analysis, which is what Kumho addressed.
That is not a safe harbor, though it is often mistaken for one. What follows is that a court looks for a protocol underneath the experience. Thirty years of doing search work is a qualification. It is not a methodology. An expert can be qualified and still lose the opinion because the way it was produced cannot be examined. What can be examined is a written protocol — stated sources, date ranges, exclusions, and controls for the obvious confounds, in enough detail that the same protocol on the same data would produce the same result in someone else's hands.
Fourth, method
Then: does the inference actually reach from the data to the opinion? Since the amendment to FRE 702 took effect on 1 December 2023, that is an admissibility question, decided by the judge under Rule 104(a) on a burden the proponent carries by a preponderance. The amendment tightened subsection (b), that the opinion rest on sufficient facts or data, and subsection (d), that it reflect a reliable application of the principles and methods to the facts of the case.
Those two are where search opinions are exposed. Subsection (b) is where a damages figure built on a third-party rank tracker's estimated traffic, rather than the site's own first-party data, comes apart; an estimate as the sole basis for a quantification is a real weakness, not a matter of taste. Subsection (d) is where the two-lines-on-a-chart causation opinion comes apart, because aligned dates are a reason to investigate, not an application of a method. Failing to separate a confirmed algorithm update from the defendant's conduct is now an admissibility problem rather than a cross-examination problem.
Where the four disciplines sit
Sixteen entries divide into four groups of four: admissibility and qualification, disclosure and discovery, evidence and exhibits, and causation and damages. Read in rule order that sequence runs backwards. Read it in the order the evidence has to survive, and it is exactly right.
The exhibit problem people learn about last
Almost everything an expert produces here is a chart or a summary, and since 1 December 2024 two rules govern them with opposite consequences. FRE 1006 as amended admits a summary, chart or calculation of voluminous admissible material as substantive evidence. New FRE 107 governs the illustrative aid, which is not evidence and must not go to the jury during deliberations unless all parties consent or the court orders otherwise for good cause. One exhibit is in the room while a jury argues about your case; the other was on a screen for four minutes. Which one you have is decided months earlier, because Rule 1006 requires that the underlying material be admissible — not admitted, admissible — and that originals or duplicates were made available to the other side.
Getting the numbers right
These pages get read adversarially, so the citations are stated exactly. The FRE 702 amendment is effective 1 December 2023. FRE 1006 as amended and new FRE 107 were adopted 2 April 2024 and are effective 1 December 2024. Under FRCP 26(a)(2)(B), the report lists publications authored in the previous ten years at subsection (iv), and cases in which the witness testified as an expert at trial or by deposition in the previous four years at subsection (v). Those get reversed constantly, and it is the kind of error that gets read aloud.
Nothing here is legal advice. It is a record of what the rules say and where opinions in this field fall apart — written for someone who knows FRCP 26 cold and has never had a reason to learn what a canonical tag is.
Published