Latest / MBA Training Data / Decision intelligence: engineer the decision, not the dashboard
Transcript
- 0:00Welcome to the Leaders Insights Podcast.
- 0:08Today's episode, Decision Intelligence and Embedded Analytics, Making the Decision in
- 0:12the Unit of Design.
- 0:13Every company I talk to has dashboards coming out of their ears.
- 0:17So why are they all still making terrible decisions?
- 0:21Because a dashboard is a museum.
- 0:23It's a beautiful glass case where data goes to be admired and then completely ignored.
- 0:28The problem isn't that people can't see the numbers.
- 0:31It's that the numbers show up nowhere near the moment someone actually decides something.
- 0:35That sounds like a slogan.
- 0:37Give me the mechanism.
- 0:38Fine.
- 0:39Picture a support agent deciding whether to refund an angry customer.
- 0:43The data that should inform that—lifetime value, refund history, churn risk—lives
- 0:49in a dashboard the agent has never opened, in a tool they don't have a login for.
- 0:54So they guess.
- 0:55The decision happens in the support ticket, but the intelligence lives two departments
- 1:00away.
- 1:01That gap is where value quietly dies.
- 1:04So decision intelligence is just putting the chart inside the app?
- 1:08That's embedded analytics.
- 1:10The plumbing that pushes data into the tool where work happens.
- 1:14Decision intelligence is the bigger idea.
- 1:17You treat the decision itself as the thing you design around, not the report.
- 1:21You ask, what recurring choice does this person make, and what would make it better?
- 1:26Then work backwards.
- 1:28Most firms do the opposite.
- 1:30They build the dashboard first and pray someone consults it.
- 1:34You're being a bit smug.
- 1:35Plenty of dashboards get used.
- 1:37Used?
- 1:38Sure.
- 1:39Acted on?
- 1:40O'Reilly's radar surveys have been consistent for years.
- 1:43A large share of analytics work never changes a single operational decision.
- 1:49It gets built, presented, admired, forgotten.
- 1:53And MIT Sloan Management Review has been hammering this since before the AI gold rush.
- 1:58The bottleneck was never the technology.
- 2:01It's that nobody redesigned the actual moment of choice.
- 2:05We spent a decade optimizing the museum, and none on the visitor.
- 2:09Alright, doesn't decision intelligence just mean let the algorithm decide and call it
- 2:13a day?
- 2:14No.
- 2:15And that's the trap everyone falls into.
- 2:18The point isn't to automate the human out.
- 2:20It's to shrink the decision to its real shape.
- 2:23Most operational decisions are boring and repeatable.
- 2:26Approve or reject, prioritize or park.
- 2:30For those, you can encode the logic and let the human handle the weird 5%.
- 2:35The mistake is automating the messy judgment calls and micromanaging the trivial ones.
- 2:41People get that exactly backwards.
- 2:43Give me a number that shows this actually moves the needle.
- 2:46GBT Labs, and I'll flag it.
- 2:49They sell the data tooling, so read their figures with one eyebrow up.
- 2:53Reports that teams embedding analytics into workflows cut the lag between question and
- 2:57action dramatically.
- 2:59Cross-check it against something independent before you quote it in a board meeting.
- 3:04But the direction is right.
- 3:06The win isn't a prettier chart.
- 3:08It's compressing the distance between I wonder and I did.
- 3:12What's the most common way companies get this wrong?
- 3:14They hire a data team.
- 3:16The data team builds what it finds interesting, and nobody ever asks the person making the
- 3:21decision what they actually need at 2 p.m. on a Tuesday.
- 3:25It's like a chef cooking whatever they feel like and being shocked the customers left
- 3:29hungry.
- 3:30So how do you know a decision is worth engineering around?
- 3:33Three tests.
- 3:34Is it frequent?
- 3:35Does it happen constantly?
- 3:37Is it consequential?
- 3:39Does getting it wrong cost real money?
- 3:41And is it currently made blind or no data at the point of choice?
- 3:45Get all three and you found gold.
- 3:48A pricing decision made 200 times a day with zero data in the moment beats any executive
- 3:53dashboard nobody scrolls past.
- 3:56And the ones that fail the test?
- 3:58Leave them alone.
- 3:59Not every decision deserves a data pipeline.
- 4:02The rare one-off strategic bet.
- 4:05You're not embedding analytics into that.
- 4:07You're arguing in a room.
- 4:08Know the difference or you'll waste a year instrumenting a decision that happens twice
- 4:12a decade.
- 4:13Give me the one thing a listener should do Monday morning.
- 4:17Pick your most expensive recurring decision.
- 4:20Sit next to the person making it and watch.
- 4:22If the data they need isn't on the screen in front of them at that exact second, your
- 4:27dashboard has already failed.
- 4:29And no amount of polishing it will help.
- 4:31The museum stays empty.
- 4:33The museum stays empty.
- 4:35Sources for today's episode.
- 4:37Cade Nuggets, The New Stack, O'Reilly Radar, Towards Data Science, MIT Sloan Management
- 4:42Review, DBT Labs vendor, Data Tooling.
- 4:46Done.
- 4:47Want to know where you actually stand?
- 4:49Take the CDO self-assessment at MBA-training.com.