The shape of a production is evidence about the party that produced it.

Not the contents — the shape. How the documents distribute across months and custodians, what proportion carry attachments, where the redactions cluster. All of that is knowable before anyone opens the first file, and most of it can be counted in a day.

It is the only stage of discovery where the other side’s choices are visible as choices. Once you are reading, you are looking at what they sent. Before you are reading, you can still see the shape of what they sent, and compare it against the shape you would expect.

Seven measurements, in the order I would take them.

1. Date distribution

Count documents per month across the whole date range. Then look for the months that are thin.

A company does not stop making documents. If the production has forty thousand documents in March and eleven hundred in April, something happened in April — a retention decision, a system change, a custodian who left, or a collection that did not reach a source. Any of those is worth knowing, and the thin month is where you ask.

2. Custodian distribution

Count per custodian, then set the counts against the organizational chart.

The person who ran the department should not have produced less than their direct report. When they have, the usual explanations are that their mail was collected from a different system, that they used a device nobody imaged, or that their date range was cut differently. Each of those is a question with a document-shaped answer.

3. File-type mix

Email dominates most productions. Look at what came with it.

If ninety percent of the set is email and almost none of it has attachments, the attachments were either not collected or not produced. Real correspondence carries spreadsheets, decks and drafts. A set without them has been through something.

4. Family completeness

Count parents that arrived without their children.

An email produced without its attachment is a document that means less than it did when it was sent, and sometimes means the opposite. Some of this is ordinary — privilege withholding is legitimate and should show on the log. What matters is the rate, and whether the log accounts for it.

5. Duplication and near-duplication

Cluster the set and look at what circulated.

The near-duplicate rate tells you how widely a document traveled and how many times it was revised. A memo existing in eleven near-identical versions across six custodians is a different fact from the same memo appearing once. Both the circulation and the revision history are arguments, and neither is visible document by document.

6. Redaction density

Map redactions by custodian and by month.

Redactions are legitimate. Their distribution is still informative. A quiet production with one month of heavy redaction, or one custodian carrying most of them, tells you where the producing party thought the sensitive material was — which is a useful thing to learn from the people who read it first.

7. Gaps against known events

This is the one that decides motions.

List the dates you already know matter: the recall, the board meeting, the regulatory letter, the day the complaint arrived. Then count what the production holds for each of those weeks.

A company in the week of a recall generates more documents than usual, not fewer. When the production says otherwise, you have a specific, dated, arguable gap — and a motion to compel that names a week rather than complaining about a volume.

What to do with the answers

Two things.

The measurements that come back normal tell you where to spend your reading. Density and custodian concentration point at the parts of the set most likely to carry the case.

The measurements that come back strange are your discovery plan for the next sixty days. They are also the only version of a compel motion that judges respond well to: not “they produced too much” or “we think something is missing,” but “the production contains eleven hundred documents for the month of the recall and forty thousand for the month before it.”

None of this requires reading a single document, and all of it is gone the moment you start.