10 minutes

Your data-interpretation checklist

Six questions to run before a number becomes a decision, short enough that you will actually use them, and written down so they do not depend on who happens to be in the room.

What you’ll be able to do

Draft a six-line data-interpretation checklist for one report you produce or rely on, naming when it runs, who runs it, and what happens when a line cannot be answered.
Compare the responses to this question and explain your choice: What is the most useful correction?
Draft a data-interpretation checklist for your work, including what this number counts, in one sentence, who is missing, and in what pattern, what the categories were built for.

Short enough that you will use it

Most interpretation checks fail for the same reason review stages do, and it is not resistance. It is scope. A unit writes a thorough list, runs it once, discovers it costs an afternoon, and within two months it is being completed after the work it was meant to shape. A check that runs late is a record of what already happened.

Six lines is roughly the limit of what gets used. They are worth attaching to a moment rather than to a document, because the moment decides what the check can still change. Three moments matter. When a question arrives, the check asks whether the available numbers can answer it. Before a finding is written, it asks what the numbers actually support. Before a number is used for a decision, it asks who is missing and who will read it.

The point is not to certify a report. It is to make three or four questions ordinary enough that asking them is not heard as an accusation, and to make the unanswerable ones visible instead of quietly filled in.

A data-interpretation checklist

Six lines to answer before a number becomes a decision, with somewhere to record what you could not answer.

What this number counts, in one sentence Share of support plans opened with a scheduled date that were completed inside the window, for the tracked workflow only. It does not count referrals that never reached a scheduled date.

Who is missing, and in what pattern Referrals closed before scheduling are outside the denominator. How many is unknown; a separate report suggests they concentrate in two regions. Recorded as unknown and raised with the workflow owner rather than estimated.

What the categories were built for Preferred language comes from the intake form and is blank in about a third of records; the template has been reading blank as English, and that default is now stated on the chart. The disability field is the eligibility category and is used here to describe the caseload, not the people.

What the comparison assumes Each region is compared with the statewide target rather than with the largest region, so no region becomes the standard. The window covers four full quarters before and after the change, and the reason for that window is printed beside the chart.

What the data can and cannot support Supports: completion differs by region and by number of access points. Cannot separate: travel time, staffing, notice practice and referral relationships move together in these records. No causal sentence appears in the finding.

Who sees it before it travels, and what changes One colleague who did not build the number, plus compensated community advisors for any finding about a specific community, with their input answered in writing. Change committed: the notice translated, and evening scheduling trialled in two counties, both reviewed at the agreed point.

Copy these six lines, fill them in for one report you produce or rely on, and run them once out loud with someone who did not build the number.

Browse and download only. Staff notes are not typed or saved on this page.

Three ways the checklist fails

It becomes a footnote

The lines get answered and the answers stay in a file while the chart travels alone. What catches it: put the first line, what this number counts, on the chart itself in plain words. If it will not fit in a sentence, you have found something worth knowing before anything else happens.

It becomes a veto

A check with authority to stop work and no service timeline teaches people to route around it. Findings start arriving as decisions already made, and the check hears about them last. What catches it: compare the number of findings that went through the check with the number the division actually published in the same period.

It becomes a rating

The moment a completed checklist is read as a measure of an analyst, a unit or a county, the answers turn careful and the unknown lines disappear first. What catches it: never attach a name to a result, never total the lines, and say out loud and often that unknown is a good answer.

Worth writing down, and worth leaving alone

Write down: what the number counts, in one sentence a person outside your unit could repeat.
Write down: the groups that cannot appear in the count, and whether you know how many there are.
Write down: every line you could not answer, marked as unknown rather than filled in.
Write down: the comparison you chose, and why that reference point rather than another.
Write down: what you promised would change, by when, and who was told.
Leave alone: anything that becomes a rating of a person, a unit or a county. It will not improve the numbers; it will reduce what anyone is willing to write down.
Leave alone: anything a colleague wrote in a private reflection. Reflections in this program are not collected and are nobody else's business.

Make unknown a respectable answer

What you can change

You control how the first unanswerable line is received, and whether it is written down as unknown or quietly filled with something plausible.

What to watch for

Do not reward a complete checklist over an honest one. The first time an unknown is treated as a failure, every later checklist will be complete and none of them will be true.

Your next step

The next time a line cannot be answered, write unknown, name who could answer it, and put that question somewhere it will be picked up.

You run the checklist and cannot answer the line about who is missing from the count. What is the most useful thing to do?

Estimate the missing group from the records you do have, so the line is complete and the finding can go out.

Record the line as unknown, say so in the finding, and name who could answer it.

Set the finding aside until the missing group can be counted properly.

Try a situation

A DSD unit agrees to try a short interpretation check. The first version has fourteen questions, a scoring sheet and a sign-off line. The analysts run it once and it takes most of an afternoon. Two of the questions turn out to be unanswerable for any report the unit produces. By the third month the check is being completed after the finding has already gone out.

What is the most useful correction?

Cut it to six questions, attach it to a specific moment in the work, and drop the scoring sheet and the sign-off line.

Keep the fourteen questions and assign one person to complete the check for the whole unit.

Keep the check as designed but move it earlier, and remove the two questions nobody can answer.

Carry this forward

The checklist works because it runs at a moment, not because it is comprehensive. Three moments are worth naming: when the question is asked, before the finding is written, and before it is used for a decision.
A line that cannot be answered is the most valuable result the checklist produces. Record it as unknown rather than filling it in, because an unknown that is written down gets fixed and a plausible guess never does.
Run it with at least one person who did not build the number. The blind spots in a set of records are invisible from the inside, and the questions that find them tend to sound naive.
Keep records about numbers and processes, never about people. Nothing in this checklist should become a rating of an analyst, a unit or a county, and nothing a colleague writes in a private reflection is collected.

A completed six-line checklist you can copy, a scenario about a check that grew too large to use, and a knowledge check on what to do with a line that cannot be answered.

Fill in the six lines for one report you produce or rely on, run them once with someone who did not build the number, and take the unanswerable lines to whoever could answer them.

Your next step

Where would six questions fit in your own week?

Name one report or one recurring request you produce or rely on.

Name the moment the check would run: when the question arrives, before the finding is written, or before it is used for a decision.

Name the person who did not build the number and would be willing to ask the questions out loud.

Decide now what you will do with the first line that cannot be answered.

Browse and download only. Course notes are not typed or saved on this page.

Mark this lesson complete

Participation and course completion in this program do not count toward DHS-required training credits unless management, a director, or DHS leadership expressly approves an exception.