learn-dataops-with-phoebe / Leader session 3 of 6
Learn DataOps with Phoebe · Leader track · Session 3 of 6

The maturity model and the metrics that matter

"Are we good at DataOps?" is a fair question with a terrible answer if all you have is a feeling. This session gives you two honest instruments: a five-level maturity ladder that tells you where you stand, and four metrics - borrowed from the best software research on Earth - that tell you the truth about your data delivery. This session, no code: what each rung of the ladder unlocks, what the four numbers mean, and how to read them without lying to yourself.

🤝 Leader track C-level · managers · data leads No code, ever Medium depth
0-3 · Recap 3-30 · Ladder and metrics 30-42 · Boardroom exercise 42-45 · Q&A
Part 0

Where we are

Session 1 named the reliability tax. Session 2 mapped the four pillars to four business risks. Now the natural leadership question: how mature are we, and how would we prove it? This session hands you two instruments a board will respect - a maturity ladder to locate yourself, and four metrics to measure yourself against the best. Both are plain-English, and both are meant for improvement, not for beating people up.

Live - discussed in session Self-study - read after ★ Questions for your team Strategy, not tooling
★ What you walk out with today A five-level maturity ladder you can place your organization on honestly, a clear sense of the single leap each level unlocks, the four DORA metrics translated into their data equivalents, an honest framing of what elite versus low performance looks like, and the discipline to read these numbers as a compass rather than a scoreboard.
Part 1 · locate yourself

The DataOps maturity ladder 8 min live

Maturity is not a grade; it is a location. Most organizations sit on one of five rungs, and each rung is defined by the single capability it unlocks. You climb by earning the next leap, not by buying the next tool. Find your rung honestly and the path up becomes obvious.

1 · Ad hoc / heroic it works because someone remembers 2 · Repeatable / version-controlled steps are written down and tracked 3 · Automated CI/CD changes ship through automated gates 4 · Measured / observed you can see quality and speed live 5 · Self-service / governed Leap unlocked: from 1 to 2 - the process stops living in one person's head and survives when they leave Leap: from 2 to 3 - releases become fast and safe because machines test and ship each change Leap: from 3 to 4 - you can finally see quality and speed, so you manage by evidence, not anecdote Leap: from 4 to 5 - trusted teams serve themselves safely inside guardrails, and the data team scales Top rung: governed self-service across the business Climb one rung at a time. Each leap unlocks a capability the rung below simply cannot reach.
🔍 Click to zoom - the five-rung DataOps maturity ladder and the leap each rung unlocks
LiveThe five levels, plainly4 min

Read these as honest self-descriptions. Most organizations are lower than they would like to admit, and that is fine - naming the rung is how you climb it:

  • Level 1 - Ad hoc / heroic. Things work because talented people remember the steps and fix breakage by hand, often at night. Nothing is written down where it counts. It feels productive and is quietly fragile.
  • Level 2 - Repeatable / version-controlled. The steps are captured and tracked, so the same process runs the same way twice and survives someone leaving. This is the first rung where the system, not the person, holds the knowledge.
  • Level 3 - Automated CI/CD. Changes flow through automated checks and ship on a repeatable path. Releases stop being scary because machines test each change before it reaches anyone.
  • Level 4 - Measured / observed. You can see quality and speed as they happen - you know when data is late or wrong before a user tells you. Now you manage by evidence instead of anecdote.
  • Level 5 - Self-service / governed. Trusted teams across the business serve themselves inside clear guardrails. The central data team stops being a bottleneck and starts being a platform. This is the summit, and few reach it.
The honest-placement rule You are on the highest rung you can meet for every pipeline, not the best one you have ever managed once. If any core pipeline is still heroic, part of you is still on level 1.
Part 2 · measure yourself

DORA for data - the four numbers that tell the truth 9 min live

Software has the best-researched delivery metrics in existence - the four DORA metrics, drawn from years of studying thousands of teams. They translate cleanly to data, and together they cut through every "we are doing fine" with four honest numbers: how often you ship, how fast, how often it breaks, and how quickly you recover.

LiveThe four metrics, translated4 min
DORA metricData equivalentWhat a good answer looks like
Deployment frequencyHow often you ship data or pipeline changesFrequent, small, low-drama changes rather than rare, giant, risky ones
Lead time for changesIdea to production for a data changeHours or a day, not weeks stuck in a queue
Change failure ratePercent of data changes that cause an incident or bad dataLow and falling - most changes land cleanly
Time to restore (MTTR)How fast you detect and fix a bad-data incidentMinutes to hours, because you saw it before the users did
Why these four and not a dashboard of fifty Two measure speed (frequency, lead time) and two measure stability (failure rate, restore time). The whole point of the DORA research is that the best teams are not forced to trade one for the other - they are fast and stable at once. Four numbers is enough to know the truth.
SPEED STABILITY Deployment frequency how often you ship a data or pipeline change Lead time for changes idea to production for a single data change Change failure rate percent of changes that cause bad data Time to restore how fast you detect and fix a bad-data incident Elite teams are fast AND stable at once Low performers believe they must trade one for the other Two speed numbers, two stability numbers. The best teams refuse to trade them - they win on all four.
🔍 Click to zoom - the four DORA metrics as speed and stability, for data
Self-studyElite versus low performers, honestly2 min read

The DORA research sorts teams into performance bands, and the gap is enormous - elite performers ship far more often and recover far faster than low performers, while breaking things less. The honest takeaways for a data leader:

  • Speed and stability rise together. The old belief that going faster means breaking more is wrong for high performers. Automation is what lets them have both, which is the entire DataOps thesis.
  • The bands are relative, not absolute. "Elite" for a bank is not "elite" for a startup. Do not chase someone else's numbers; measure your own trend and beat last quarter.
  • You do not need perfect measurement to start. A rough, honest number this month that you improve next month beats a precise number you never collect. Direction matters more than decimals.
Part 3 · use them well

Reading the metrics honestly 5 min live

Metrics can heal a team or wreck it, depending on how the leader holds them. The difference is entirely in your hands, and it is worth getting right before you ever put a number on a slide.

Self-studyCompass, not scoreboard2 min read

Four principles keep these numbers useful instead of toxic:

  • For improvement, not punishment. The moment a metric is used to blame a person, people optimize the metric and hide the truth. Measure the system, coach the people.
  • Watch the trend, not the point. A single month's figure is noise. The direction over a quarter is the signal. Ask "are we getting better?" not "did we hit the number?".
  • Beware vanity metrics. "We ran 4,000 pipelines" sounds impressive and tells you nothing. "Our change failure rate fell from 20% to 6%" is real, because it maps to a risk you named in session 2.
  • Pair speed with stability, always. Never celebrate deployment frequency without showing change failure rate beside it, or you reward recklessness. The four numbers are a set; report them as one.
The one-line test Before you show a metric, ask: would my team feel safe telling me the honest number? If not, you have a scoreboard, and it will lie to you. Turn it back into a compass.
Boardroom exercise

Place yourself on the ladder ★ 10 min · everyone reflects

No instrumentation required - just honesty. Locate your organization on the maturity ladder, pillar by pillar, and put a rough figure against each of the four metrics. The output is a one-slide "where we are" you could show a board next week.

Find your rung. For your most important pipeline, which of the five levels honestly describes it today - heroic, repeatable, automated, measured, or self-service? Resist rounding up.

Check for split levels. Are some pillars higher than others? Many organizations are level 3 on delivery but level 1 on measurement. Name the lowest rung any core pillar sits on - that is your real floor.

Guess the four numbers. Roughly: how often do you ship a data change, how long from idea to production, what share of changes cause an incident, and how fast do you recover? A rough number you can improve beats none.

Pick one leap. Name the single next rung you want to climb this year and the one metric you would watch to know you are climbing. One leap, one number - that is a strategy a board can back.

Real world

The team that was two levels at once. A data group proudly called itself "automated" because releases were smooth - level 3 on delivery. But nobody could say how often data was late, because nothing measured it: level 1 on observation. Naming that split, rather than averaging it away, turned their next year's plan into a single clear leap - get to measured - with time-to-restore as the number they watched. Honesty about the lowest rung is what set the direction.

Take this to work

★ Questions to ask your data team

Five questions that place you and measure you Ask for numbers, not adjectives. "Fast" is a feeling; "under a day" is a metric.
Source material

What this session covers

This session draws on DataOps maturity frameworks and the DORA software-delivery research, translated for a data leader. This page covers:

The DataOps maturity ladderPart 1 · five levels from ad hoc to governed self-service
The four DORA metrics for dataPart 2 · frequency, lead time, failure rate, restore time
Elite versus low performersPart 2 · speed and stability rise together, bands are relative
Reading metrics honestlyPart 3 · compass not scoreboard, trends, vanity metrics
Build, buy, or open-sourcepreviewed here; the tooling decision is the subject of a4
Check yourself

Three questions before you go 🎯 ◐ 90 seconds

1 · The leap from level 1 (ad hoc/heroic) to level 2 (repeatable) mainly means...

Level 2 is the first rung where knowledge lives in the tracked, version-controlled system rather than in a heroic individual's memory - so it runs the same way twice.

2 · "Change failure rate" as a data metric measures...

It is one of the two stability metrics. Paired with time to restore, it tells you whether shipping faster is quietly costing you reliability.

3 · The healthiest way for a leader to hold these metrics is...

Metrics used to blame people get gamed and hidden. Measure the system, watch the trend, and always show speed and stability together so no one rewards recklessness.

Leader session 3 cheat sheet · pin this

The five rungs1 ad hoc, 2 repeatable, 3 automated, 4 measured, 5 self-service/governed.
Honest placementYou are on the lowest rung any core pipeline sits on, not your best day ever.
Deployment frequencyHow often you ship a data change. Frequent and small beats rare and giant.
Lead timeIdea to production for one change. Hours or a day, not weeks in a queue.
Change failure ratePercent of changes that cause bad data. Low and falling is the goal.
Time to restoreHow fast you detect and fix a bad-data incident. Minutes, not days.
Elite vs lowThe best teams are fast AND stable. Automation is what removes the trade-off.
Next sessiona4 tackles build, buy, or open-source - how to acquire the tooling behind the ladder.