Skip to content
Step 1 of 9

Put someone in charge

Name one leader who can settle disagreements about numbers, one person with protected time to do the work, and write a single page saying what you are measuring and why.

How long
2 to 3 weeks
Who owns it
The presiding judge, court administrator, chief, or sheriff

What you get

A decision-maker who can end arguments about numbers, and a person whose job actually includes this work.

Do this first

Ask one question at your next leadership meeting: if two people disagree about what counts as a case, who decides? If nobody can answer, that is the gap.

Most failed data projects failed in the first month, before anyone touched a record. The agency skipped to software, formed a large committee with no way to decide anything, and gave the work to someone who already had a full-time job.

This phase produces no data. It produces the authority to make decisions later, when those decisions get uncomfortable.

Why agencies get stuck here

The work gets handed to IT as a technical project. IT can move data between systems, but only leadership can settle what a case is, and that is the decision everything else rests on.

What to do

  1. 1

    Name one leader, not a committee

    One person with budget authority, whose job here is narrow: settle disagreements about what a number means, protect the analyst’s time when it gets squeezed, and stand behind the first uncomfortable result. Test your choice with one question. Can this person tell the bench, the clerk, and IT that a decision is final, and have it stick?

  2. 2

    Give one person real time to do the work

    Someone has to draft the definitions, check the data, and answer where each number came from. In most agencies this is not a new hire. It is an analyst, deputy clerk, or research coordinator whose hours are formally reassigned. Be honest about the size: roughly a quarter to half of one person in a small agency, a full job at scale.

    • Get their supervisor to approve the time in writing. If it is not written down, it is the first thing dropped in a busy month.
    • Name a backup who sits in the same meetings, so one departure does not end the program.
  3. 3

    Keep the working group to five to eight people

    Not thirty. A large group can review a definition but cannot settle one, and settling is the whole job. You need agency administration, IT or your software vendor contact, and the unit whose work is being measured. Add prosecution, defense, pretrial, or the jail when a measure crosses their work.

    • Include one person likely to challenge your numbers in public. Better they do it in the room first.
    • Agree now what the group decides on its own, and what goes to the leader.
  4. 4

    Write one page

    One page, because longer documents do not get read or revisited. This is what you point at later when someone proposes a measure nobody needs, or when a result comes back badly.

    • The three to five specific decisions this data should inform. "Better decision making" is not one. "Whether to add a second criminal docket" is.
    • Who sees the numbers, in order: your own staff, leadership, partner agencies, funders, then the public.
    • What you are explicitly not doing this year. This is the most useful line on the page.
    • Who breaks a tie, by name.
    • What you will do if the first numbers look bad. Agree this before you see them.
  5. 5

    Aim for a defensible number in 90 days, not a dashboard

    A dashboard built in 90 days on definitions nobody agreed to will be wrong in public, and that makes the second attempt much harder. Promise something better: agreed definitions, one automatic data feed, and a starting number you can defend.

Decide these before you move on

  • Are you measuring mainly to improve operations, to satisfy a reporting requirement, or to answer critics? You will do all three eventually, but the order shapes what you build first.
  • Who signs off before a number goes outside the agency?
  • What happens if the first numbers look bad?

What you will have

  • One-page charter, signed
  • Named roles with agreed time commitments
  • Working group and meeting schedule

You are done when

  • One leader is named in writing and has agreed to attend the definition sessions.
  • The analyst’s time is approved by their supervisor and written into their workplan.
  • The charter fits on one page and names three to five specific decisions.
  • The working group has met once and knows who breaks ties.

What goes wrong

A committee of thirty

Everyone adds an exception and nobody can close the discussion. Keep the group small and brief everyone else quarterly.

An analyst with no protected time

The single most common reason this work stops. It loses every scheduling conflict unless the hours are formally approved.

A charter full of good intentions

If it says the goal is to become data-driven, there is no way to tell whether you succeeded. Name the decisions and you get a scope limit and a success test in one line.

Booking a software demo first

It turns the problem into a purchase. You end up with a tool answering questions nobody agreed to ask, and the hard work still waits for you afterward.

Standards you can borrow definitions from

See what you could measure today

We will map your systems to the measures your state already reports, and show you the gaps before you commit to anything.