Quick answer
Data maturity is how far an organization has moved from collecting numbers to deciding with them: data that is trusted, findable and actually used. It is usually graded in five levels, and the level is set by the weakest area, not by how many tools you own.
Formula: overall level = min(trust, find, use)
What data maturity means
Data maturity is how far an organization has moved from collecting numbers to making decisions with them. The usual path runs from "we can pull a report", to "we can explain why a number moved", to "we can forecast it", to "routine decisions run on data without someone pulling a report each time".
The idea of grading an organization on a ladder of levels is older than analytics. The Software Engineering Institute's Capability Maturity Model, published in 1993, described software processes in five levels, and most data maturity models borrow that shape. UK government's Data Maturity Assessment for Government, a framework built by the Government Digital Service and the Central Digital and Data Office, also uses five levels (Beginning, Emerging, Learning, Developing, Mastering), but applies them to ten topics rather than one overall score.
Two points hold across almost every model, and they are the core of this entry:
- Maturity is about whether data is trusted, findable and used for decisions. It is not about how much data you hold or how many tools you have bought.
- You are only as mature as your weakest area. A polished dashboard sitting on numbers nobody agrees about is still a low-maturity setup.
A related but different question is how to build the full roadmap from one level to the next; that is covered in the data maturity model guide. This page stays with the definition and a quick self-check.
The five levels, with one test question each
The levels below are a working scale for self-assessment, not a standard. Each level has one thing you can observe and one question you can answer yes or no.
| Level | Name | What it looks like | Test question |
|---|---|---|---|
| 1 | Ad hoc | Numbers live in personal spreadsheets and one-off exports. Two people asking the same question get two answers. | If two people pull last month's active users, do they get the same number? |
| 2 | Reported | Standard reports exist. Each key metric has a written definition and a named owner. | Can you name the owner and written definition of your five most-used metrics? |
| 3 | Explained | When a metric moves, the team can split it by channel, segment or step and find where it moved. | When signups drop 20%, can you say where the drop happened within one working day? |
| 4 | Predictive | Forecasts or controlled tests guide plans, and the forecast error is tracked. | Do you compare last quarter's forecast with what happened and keep the error figure? |
| 5 | Automated | Routine decisions fire from rules or alerts, with a person reviewing the rules. | Which decisions run without anyone pulling a report, and who reviews them? |
Most small teams get more value from making level 3 solid than from reaching for level 4 or 5. Forecasts and automation on top of numbers nobody trusts only produce wrong answers faster.
Three questions that sit under every level
Rather than asking "what level are we?" in the abstract, score three things separately, each from 1 to 5 using the table above.
- Trust. Can people agree on the number? This depends on written definitions, a record of which events you collect and why, and a way to notice when collection breaks. See data quality and tracking plan.
- Find. Can a new analyst work out which data exists, what each field means and who owns it, without asking five people? This is the work of data governance in its simplest form.
- Use. Do decisions change because of the data? Look for a recent decision, a named metric and a before-and-after, not for the number of dashboards.
Then take the lowest of the three as your overall level:
overall level = min(trust, find, use)
The three questions match the UK framework's topics only loosely: its "Knowing the data you have" topic lines up with Find, "Making decisions with data" with Use, and several of its governance topics with Trust. The point of using the minimum rather than the average is that the weakest area decides what you can safely do next.
A worked self-assessment
Sample data throughout this section.
Step 1: test trust with a spread check. Ask five analysts to compute last month's active users independently. Suppose they return 8,120, 8,400, 7,950, 9,100 and 8,310. The median is 8,310. The spread is (9,100 − 7,950) ÷ 8,310 = 1,150 ÷ 8,310 = 13.8%. As a rule of thumb, a spread above about 5% means people are counting differently, and the team is at level 1 for Trust until definitions are written down. The 5% line is our own cut-off, not a published standard.
Step 2: score each area. Two teams in the sample score themselves like this:
| Team | Tools in use | Trust | Find | Use | Average | Overall (minimum) |
|---|---|---|---|---|---|---|
| A | 14 | 2 | 2 | 4 | (2 + 2 + 4) ÷ 3 = 2.7 | 2 |
| B | 4 | 3 | 3 | 3 | (3 + 3 + 3) ÷ 3 = 3.0 | 3 |
Team A has more than three times the tools and the more advanced use case (it runs a forecast), yet it is a level 2 team: its forecast rests on numbers that analysts compute differently and on data nobody can find documentation for. Team B has a plain setup but its definitions, inventory and decisions are all consistent, so it is level 3.
Step 3: pick the next action from the lowest score. Team A should not buy a new tool. It should fix Trust and Find: write down definitions and owners, and list the events it collects. Team B should look at Use: pick one decision that currently waits on a manual report and set up a way to check it automatically.
Common mistakes when measuring data maturity
- Counting tools. In the example, Team A has 14 tools and Team B has four. The UK framework lists Tools as only one of six themes, next to Uses, Data, Leadership, Culture and Skills, which is a useful reminder that tooling is one slice.
- Equating maturity with machine learning. A team with an accurate, trusted weekly report that drives real decisions is more mature than one running models on disputed inputs.
- Scoring once and filing it. Maturity moves when people leave, tracking changes or the product ships a new flow. Re-score every quarter and keep the history.
- Averaging the areas. An average of 2.7 hides the fact that Trust is at 2. Use the minimum, then look at the lowest area.
- Self-assessing without evidence. Every score needs an observable fact: the spread check, a named owner, a decision log entry. "We feel pretty good" is not an input.
- Treating the levels as a race. Level 5 is not the goal for every team; the goal is the level at which your decisions are made on numbers you can defend.
How to read data maturity from TapCub
Take Team A from the example: Trust 2, Find 2, Use 4. TapCub does not calculate a maturity score for you, and it does not make decisions on your behalf; it supplies the evidence for the three questions.
- Trust: write the plan, then check it against reality. List the events you intend to collect in All-platform analytics › Data › Tracking plan: name, owner, trigger and status (Proposed, Developing, Accepting, Live, Deprecated). Turn on Live verification (24 h) to see how many times each event arrived and how complete its required properties are. A low completeness figure is a prompt to fix the code; it does not block events.
- Find: clear the unnamed events. All-platform analytics › Data › Event inbox collects events that arrived with no definition and autocapture candidates; accept them with a proper name or ignore them. All-platform analytics › Data › Events & properties then shows how often each event fires, for how many visitors and its seven-day trend. The step-by-step version is in create a tracking plan.
- Use: explain a change and get told about the next one. In All-platform analytics › Behavior › Behavior analysis, split a drop by funnel step or segment (the level 3 test). All-platform analytics › Operations › Monitors checks a metric hourly, daily, weekly or monthly and notifies you when a threshold is met; the dry run shows how many times it would have fired in the last 30 days. Monitors only notify; they never act.
- Re-score each quarter. Keep the spread check and the three scores in your own notes. TapCub is where the evidence lives, not where the grade is kept.
Sources
Was this explanation clear?
We use your answers to fix and expand this page.
Couldn’t send. Please try again in a minute.
Thanks for the feedback. We’ll use it to improve this page.You’ve already sent feedback on this page. Thanks.

Customer Experience Consultant
Writes about live chat, visitor conversations, and self-service, sharing practical ways to make customer communication clearer and more effective.