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.
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.
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.
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 metric | Data equivalent | What a good answer looks like |
|---|---|---|
| Deployment frequency | How often you ship data or pipeline changes | Frequent, small, low-drama changes rather than rare, giant, risky ones |
| Lead time for changes | Idea to production for a data change | Hours or a day, not weeks stuck in a queue |
| Change failure rate | Percent of data changes that cause an incident or bad data | Low and falling - most changes land cleanly |
| Time to restore (MTTR) | How fast you detect and fix a bad-data incident | Minutes to hours, because you saw it before the users did |
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.
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.
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.
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.
★ Questions to ask your data team
- "Which of the five maturity levels would you honestly put our core pipelines on, and why?" The reasoning matters more than the number - listen for whether the process lives in people or in the system.
- "How often do we ship a data change, and how long does one take from idea to production?" These are your deployment frequency and lead time - the speed half of the truth.
- "What share of our changes cause an incident or bad data, and how fast do we recover?" Change failure rate and time to restore - the stability half. If we cannot answer, we are not yet measured.
- "Would you feel safe telling me the real numbers?" If the honest answer is no, our metrics are a scoreboard and they will mislead us. This is the culture question hiding inside the data.
- "What single leap up the ladder should we target this year, and which one metric proves we made it?" This turns maturity from a vague aspiration into a fundable plan.
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:
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.