Building Exhibits That Reach the Jury

What the 1 December 2024 split changed about the work

Two rules took effect on 1 December 2024, adopted by the Supreme Court on 2 April 2024, and between them they decide what an exhibit in this discipline is worth. Rule 1006, as amended, admits a summary, chart, or calculation offered to prove the content of voluminous admissible material as substantive evidence, whether or not that material has been introduced. Rule 107 governs the illustrative aid, which is not evidence and must not be provided to the jury during deliberations unless all parties consent or the court orders otherwise for good cause. Rule 1006(c) sends anything functioning only as an aid to Rule 107; Rule 107(d) sends anything admitted to prove the content of voluminous evidence back to Rule 1006.

The consequence for craft is direct. One exhibit sits on the table while twelve people argue about your case for two days. The other was on a screen for four minutes and then went away. Nothing about the visual design decides which one you get. What decides it is the source of the data, whether that source was produced, whether the chart is faithful to it, and whether the caption states a fact or an opinion. Both rules are short: Rule 1006 and Rule 107.

Build from a produced source, or do not build

The pipeline that produces a Rule 1006 exhibit has five stages, and the order is not negotiable.

1. The source. A native export, a raw crawl, or a log file, collected with its parameters and hashed at collection.
2. Production. That source goes to the other side in native form, at a reasonable time — in practice, with the report rather than with the exhibit list.
3. Derivation. A script or stated query transforms the source into the plotted values, saved and produced with it.
4. The chart. Generated from the derivation output, never hand-adjusted afterward.
5. The caption and legend. Describing what is plotted, its source, and its date range.

Two failure modes account for most losses. The first is the hand-edited spreadsheet, where someone pastes an export into a workbook, filters it, fixes a few rows, and charts the result. That workbook is not the underlying material — it is a second, undocumented summary with no audit trail, and nobody can reproduce it six months later. The second is the chart built from a source that was never produced. Rule 1006(b) requires the proponent to make the originals or duplicates available for examination or copying at a reasonable time and place. An objection on that ground has nothing to do with the quality of the analysis and still costs you the exhibit.

Recipe: rankings over time

This is the exhibit clients ask for and the hardest to build properly. The usual source is a rank tracker — a commercial service running a fixed list of queries on a schedule from a configured location and device, recording where a given URL appeared. Its output is a vendor's measurement rather than the client's record, and the vendor will not certify anything. As the sole basis for a damages exhibit that is a real weakness, and I would say so about my own work.

Build it this way. Where the client's own Search Console data covers the period, prefer it and say why: it records what the engine logged for that property. Plot average position alongside impressions and clicks, and state in the legend that average position is averaged across impressions and is a property-level figure, not a ranking anyone saw. Where a rank tracker is the only source, produce the full raw export with the tracking configuration — query list, location, device, language, and the date each query was added or removed — because the configuration defines what was measured, and changes to it explain movements that had nothing to do with the site.

Do not smooth. A moving average is a modeled series, and plotting it on the same line as observed values without labeling it is the fastest way to turn a summary into something an opponent can call a construct.

Recipe: sessions by channel

The underlying material is the client's own analytics export, a system record of what its measurement platform collected. That is a good Rule 1006 base: authenticable as a process or system under Rule 901(b)(9), certifiable under Rule 902(13), and producible by the party that owns it.

The trap is that “organic search” is a configuration rather than a fact. Channel groupings are definitions set inside the property, they can be customized, and they can change mid-period without anyone recording it. So produce the channel definitions with the export, state them on the exhibit, and check whether they changed during the window. State the date range, the timezone, and whether the platform sampled or thresholded any part of the series.

Keep the interval constant — weekly throughout or monthly throughout, never weekly before an event and monthly after it. Where a measurement change occurred inside the window, such as a platform migration or a tagging change, break the series and label the break. A bridged series across a measurement change compares two different things, and once that is shown on cross-examination the rest of the exhibit inherits the doubt.

Recipe: link acquisition timelines

A link timeline plots when inbound links to a site appeared. The source is almost always a third-party index built by a vendor's own crawler, with coverage nobody outside the vendor can characterize, and that belongs on the face of the exhibit rather than in a footnote.

The critical detail is what the dates mean. A first-seen date records when the vendor's crawler observed a link, not when the link was published, and the gap can be weeks. So the axis label reads “links first seen by [named index]” rather than “links acquired,” which is both accurate and, in my experience, more persuasive than the version that overstates.

Produce the complete raw export — source URL, target URL, anchor text, first-seen and last-seen dates, and whatever attributes the index records — not a filtered subset. Where the analysis excludes links, state the exclusion rule and apply it in the derivation script, so the exclusions are reproducible rather than editorial. And keep the counting unit explicit: links, referring domains, and referring IP ranges are three different series, and mixing them in one line is a defect an opponent will find.

Recipe: crawl error rollups and redirect tables

This is the strongest Rule 1006 material in the discipline, because the underlying data is voluminous, reproducible, and describable as a process. A crawl of forty thousand URLs cannot be conveniently examined in court, and a rollup of status codes by site section is what the rule describes.

Produce the raw crawl, not the tidy table. State the crawl parameters on the exhibit or in a note: date and timezone, seed URL, user agent, whether JavaScript was rendered, depth and concurrency, and whether the crawl respected the site's robots.txt file — the file at a site's root instructing crawlers which paths not to request, which changes what any crawler can see.

Keep the observation separate from the conclusion. What the server returned for each URL is an observed fact: a 200, a 301, a 404, a 500. That a given 404 is an “error” caused by a migration is an opinion, and it belongs in testimony rather than a column header. Count URLs rather than pages, since one page routinely has several, and disclose how duplicates and parameterized URLs were treated. Redirect tables should show the full chain per URL with each hop's status code, because three chained 301s and a single 301 are different facts.

The build review that keeps a summary a summary

Before an exhibit leaves the office it gets a review pass against a fixed list. That is mechanical on purpose: the choices that drop a chart into Rule 107 are usually made unconsciously, by people trying to be helpful.

• The caption states what is plotted, not what it means. “Organic sessions, weekly, 1 January 2023 to 31 December 2024” is a summary. “Traffic collapse following the migration” is a conclusion printed on an exhibit.
• The vertical axis starts at zero, or the truncation is stated on the chart. A truncated axis makes a nine percent change look like a cliff, which is exactly what Rule 403 reaches.
• The interval is constant, and the date range is set by the data question rather than by where the line looks best.
• Every series names its source, date range, and collection method in the legend, because each is a separate admissibility question.
• No event shading, no arrows, no callouts. An annotation asserting that an update or a migration occurred on a date is testimony rendered as graphics.
• Observed and modeled values are never on one line, and modeled series are labeled as modeled.
• The chart regenerates from the script with no manual steps, so faithfulness can be demonstrated rather than described.

A workable test at the end of the pass: if a competent opponent could show the jury that the chart communicates something the underlying data does not, it is doing argument. Rebuild it, or offer it as an aid and make the argument in testimony instead.

The production packet Rule 1006(b) expects

Rule 1006(b) is satisfied by making the originals or duplicates available for examination or copying at a reasonable time and place. Satisfying it well means producing a self-contained packet per exhibit, with the report rather than later.

• The raw source files, in native format, with their collection hashes and the manifest.
• The export parameters — account and property identifiers, date range, timezone, filters, segments, sampling, and configuration definitions such as channel groupings or tracking settings.
• The derivation — the script or stated query, in a form that runs, plus any mapping tables it uses.
• A data dictionary defining every column and every derived field.
• An index mapping each exhibit number to the files it was built from.

The timing point is where otherwise good work is lost. Production during discovery, alongside the report, comfortably satisfies “a reasonable time and place.” Production on the eve of trial invites a fight on grounds that have nothing to do with the analysis. And an exhibit first appearing at trial has a disclosure problem under Rule 26 as well as an evidentiary one, since the summary and its data are part of the basis of the opinion.

Sequencing, and when an aid is the right answer

The order is: collect and hash, produce natively, derive by script, disclose the exhibit and its basis in the report, produce the packet, list the exhibit, and argue Rule 1006 at the pretrial conference if it is contested. Every step except the last happens before anyone is thinking about trial, which is why exhibit quality here is a workflow question wearing a rule number.

Some exhibits are not summaries and never can be. A diagram of how a search engine crawls, indexes, and ranks pages. A schematic of a redirect chain — the sequence of automatic forwards a browser follows from an old URL to a current one. A cutaway of a results page. None of those summarizes voluminous material; each exists to help a jury follow testimony. Offer them as Rule 107 aids, ask that they be entered into the record under Rule 107(c), and plan on their not going back with the jury.

For everything else, build to Rule 1006 by default. A faithful, sourced, reproducible summary of produced admissible material is evidence if the court agrees, and is still usable as an aid if it does not. An exhibit built as argument has nowhere to fall back to.

Frequently Asked Questions

Published 2026-08-11