How to Use Analyst

Overview

Analyst answers questions about your firm the way you'd ask them out loud. You type "which case types settle highest?" and get a chart and a table back, computed across your whole book of matters — no report request, no export, no waiting on someone to build it.

That simplicity is the point, and it's also the thing that trips people up. Pre-built apps only cover some of your firm’s most likely asks, so the first question is always: what do I actually want to know? This guide covers how to ask well, how to read what comes back, how to turn a good question into something your team reuses, and what to do when an answer looks wrong.

If you're looking for what Analyst is and whether your firm is set up for it, start with the Analyst overview. This one assumes it's on and you're sitting in front of it.

Before Your First Question

Two things worth knowing up front, because they explain most early surprises.

Analyst only knows what it can reach. It answers from your matters in Eve, your case management system(if an integration is in place), Atlas, and Eve's own data including Auditor findings. It does not know about matters that never made it into Eve, and it has no outside or third-party data. If your firm doesn't have a supported CMS connected, Analyst can still answer — it's just working from a narrower slice, and it's worth knowing which slice before you draw conclusions from it.

Analyst is read-only. It reports, it surfaces, it hands off. It won't change a case, update your CMS, or assign anything. When an answer tells you six matters have been sitting in pre-lit for four months, the next move is a person or an Eve Agent — Analyst has done its part.

Asking a Question

Once on the Analyst page, type your question the way you'd ask a senior partner who happened to have every case file memorized. Full sentences are fine. You don't need field names, filters, or any particular format.

Analyst has a set of starter prompts to build apps that can get you going. They're useful for seeing the shape of what it can do, but they're not a menu — if none of them is what you're wondering about, ignore them and type your own. Freeform is the design, not the fallback.

Writing a Question That Actually Works

Most weak answers trace back to a vague question rather than a data problem. A few habits that help:

Name the time window. "How are we doing on settlements" and "how are we doing on settlements this quarter versus last" produce very different answers, and only one of them is actionable.

Say how you want it cut. By attorney, by case type, by referral source, by stage. Analyst will pick something reasonable if you don't, but you know which breakdown you actually run the firm on.

Ask one thing at a time. A question with three clauses tends to return an answer that half-covers all three. Ask the first one, look at the answer, then follow up with a question to filter those results.

Use your own vocabulary for the thing you want to find, not for the data model. "Cases that have been sitting too long without a demand" works. You don't need to know what field a demand date lives in.

Questions that land well look like these:

• Which practice areas are driving the most fees year to date?

• Which cases have been in pre-lit for more than 90 days without a demand?

• What's our average time from intake to demand, broken out by case type?

• How is each attorney pacing against their monthly goal?

• Show total settlement value per week, broken down by attorney.

• Which referral sources are sending us the highest-value cases?

If a question comes back thin, rephrase and ask again rather than assuming Analyst can't answer it. 

Checking an Answer Before You Act on It

Analyst is grounded in your firm's own data. Treat early answers the way you'd treat a report from a smart new hire: probably right, worth spot-checking, and definitely worth checking before it drives a decision with money attached.

The fastest check is the data table. Open it, pick a matter you actually know, and see whether the row matches reality. The best results in using Eve come from firms who trust, but verify. If a question doesn't pull the necessary data, you've learned something more useful than the answer itself — a data gap you can point your Eve team at.

Three things that make an answer look wrong when it isn't:

• Matters not in Eve. Anything your firm never activated in Eve won't be in the count. A total that reads low is often a scope question, not a math error.

• Sparse CMS fields. If the field a question depends on is inconsistently filled in your CMS, the answer inherits that. Analyst reports what's there.

• Definitions that differ from your firm's. "Open case" or "settled" may not be cut exactly the way your firm cuts it. Worth confirming once, early, rather than rediscovering it in a partner meeting.

Who Should Be Using It

Analyst is built for cross-case, money-and-performance questions — firm owners, managing partners, firm leadership, legal ops. Right now, Analyst is only accessible by firm admins. 

When Something Looks Off

Flag these to your Eve team rather than working around them:

• A number that doesn't match what you know to be true, especially where the data table disagrees with a matter you can verify

• A question Analyst consistently can't answer that feels like it should be in scope

• A breakdown that doesn't match how your firm defines the category — case types, stages, or practice areas cut differently than you'd cut them

• Coverage that looks incomplete — a whole attorney, office, or case type missing from a result