Ask two partners for the firm's realization rate and you may get two numbers. Both can be right. They simply used different definitions.
Conversational BI tools let anyone ask financial questions in plain English. That is useful only if every question resolves to one agreed definition. This guide lists the metrics to settle first and the choices each one hides.
Why Definitions Come Before AI
A dashboard shows a number with its filters visible. A chat answer often does not. If "realization" can mean billed, collected or both, the AI will pick one, and a partner may act on the wrong one.
Agreeing definitions first turns AI BI from a source of debate into a faster way to reach numbers the firm already trusts.
The Ten Metrics to Define
1. Billing Realization
Usually fees billed divided by the standard value of recorded time. Decide which rate counts as standard: rack rate, client rate or agreed matter rate.
● Do pre-bill write-downs count against it?
● How are alternative fee matters treated?
2. Collection Realization
Usually fees collected divided by fees billed. Decide the period. Collections lag billing, so a monthly figure can swing widely.
3. Overall Realization
Collected fees divided by the standard value of time worked. Bar associations, such as the Illinois State Bar Association, describe the three realization types this way.
➤ Pick one as the default answer when someone asks for "realization"
4. Utilization
Billable hours divided by an hours target. Decide the denominator: annual target, available hours after leave, or a fixed standard.
● Are pro bono hours included?
● Do non-billable firm projects count?
5. Work in Progress (WIP)
Unbilled time and costs. Decide the valuation basis and an aging rule, such as days since the work date.
6. Accounts Receivable Aging
Billed but unpaid fees by age bucket. Decide whether buckets run from invoice date or due date, and how partial payments apply.
7. Lockup
The time between doing work and collecting cash, often WIP days plus debtor days. Decide the calculation window and whether to use averages or period-end balances.
8. Write-Offs and Write-Downs
Decide where each is recorded, at pre-bill, on invoice or after billing, and whether the AI should separate them. Mixing them hides where value is lost.
9. Matter Profitability
Revenue less the cost of the time and expenses on a matter. Decide the cost rate for each timekeeper and whether overhead is allocated.
10. Origination and Working Credit
Decide how credit is split when several partners share a client. This affects many partner-level questions, so it needs firm leadership sign-off.
A Definitions Template
Use one row per metric. Get sign-off from the CFO and the managing partner.
| Field | Example for billing realization |
|---|---|
| Name | Billing realization |
| Formula | Fees billed ÷ standard value of time billed |
| Standard rate | Client-approved rate |
| Includes | Hourly matters |
| Excludes | Fixed-fee and contingency matters |
| Period | Month and rolling 12 months |
| Source report | Approved billing report from the financial system |
| Owner | Director of finance |
| Approved | CFO, date |
Decide Who Sees What
Financial data is sensitive inside a firm. Before launching AI BI, agree access rules.
● Partners see their own clients and matters, plus firm totals
● Practice leaders see their group
● Finance sees everything
● Associates see only their own time and targets, if anything
The AI should respect the same permissions as your reports. Test this with a few users before wider release.
Set Answer Rules for the AI
● Always state which definition was used
● Show the period and filters applied
● Name the source report
● Say "not available" rather than estimating a metric that is not defined
● Flag when data is incomplete, such as a month not yet closed
Test With Real Questions
Collect 20 questions partners actually ask. Answer each manually from approved reports. Then ask the AI and compare.
➤ "What was our realization on Acme matters last quarter?"
➤ "Which clients have the oldest unpaid invoices?"
➤ "How is the litigation group's utilization against target this year?"
Any mismatch points to a missing or unclear definition. Fix the definition, not just the answer.
Practical Example
A 200-lawyer firm defines eight metrics and signs them off in two meetings. During testing, the AI's realization for one client differs from finance's figure. The cause: fixed-fee matters were included in one and excluded in the other.
The firm adds an explicit exclusion to the definition. The next test matches. Partners now see the definition name under every answer.
Where to Go Next
For broader finance priorities, see our priorities for law firm CFOs and our comparison of financial analytics platforms for Elite 3E. Industry commentary from Thomson Reuters Legal is also useful for benchmarks.
Common Definition Traps
These are the places where firms most often find two versions of the same number.
● Rates: standard, client-approved and matter-specific rates mixed in one metric
● Timing: work date, bill date and payment date used interchangeably
● Fee types: fixed-fee and contingency matters included in hourly measures
● Credits: write-offs recorded as discounts, or the reverse
● People: lateral partners counted in one period and not another
Write the rule for each trap into the definition itself, not into a separate memo.
Check Data Readiness
Definitions only help if the data supports them. Before launch, confirm a few basics with finance and IT.
● Month-end close status is visible, so the AI can flag open periods
● Timekeeper rates and history are complete
● Matter fee types are coded consistently
● Approved reports exist for each defined metric
Where a report does not exist, create and approve it first. The AI should read from governed reports, not raw tables.
Roll Out in Stages
Start with finance and a few practice leaders. Collect their questions for a month, fix definitions, then extend to all partners. A staged launch keeps early mistakes small and builds trust.
Our Own Product
Disclosure: LuMay offers Metrics Insights, a natural-language assistant across approved 3E reports and measures. Its page describes governed definitions, visible context and review controls. Apply the checks above to any tool, including this one.
Next Step
Fill in the definitions template for realization and utilization this week and get them signed. Then test AI answers against them. To see governed definitions in practice, explore Metrics Insights.
References
● Illinois State Bar Association, The 3 types of realization rates: https://www.isba.org/barnews/2016/07/13/best-practice-3-types-realization-rates
● Thomson Reuters Legal insights: https://www.thomsonreuters.com/en-us/posts/legal/
● Altman Weil, Realization and Compensation: https://www.altmanweil.com/wp-content/uploads/2022/04/Realization-and-Compensation.pdf
