Latest / MBA Training Data / DuckDB and text-to-SQL: redesign the analytics team
Transcript
- 0:00Welcome back to Leaders Insights.
- 0:07Duck DB just made the analyst job description obsolete.
- 0:10Here is what CDOs should do about it and why it matters this week.
- 0:14If a business manager can type, show me last quarter's churn by region, and get a working
- 0:19answer in 10 seconds, what exactly am I still paying my analytics team for?
- 0:23That's the question keeping half the chief data officers I talk to awake.
- 0:27And most of them are answering it wrong.
- 0:29They're treating it as a headcount cut.
- 0:32It's a job redefinition.
- 0:33Walk me through what actually changed.
- 0:36Because natural language BI sounds like a slide from 2019 that never delivered.
- 0:41It didn't deliver, and here's why it does now.
- 0:44Two things collided.
- 0:45First, Duck DB.
- 0:47A free, tiny analytics database engine that runs inside your laptop or your browser.
- 0:53No server farm required.
- 0:55Second, language models that can turn plain English into SQL.
- 0:59The query language you use to interrogate a database.
- 1:02So the model writes the query and Duck DB runs it.
- 1:05Where's the leap?
- 1:06People have bolted chatbots onto dashboards for years.
- 1:10The leap is the query executes on real data, locally, in under a second.
- 1:14And you can see the SQL it wrote.
- 1:17Old text-to-query tools gave you a black box and a plausible looking number.
- 1:21This gives you the number and the receipt.
- 1:23You can check its work.
- 1:25Let's tear something apart.
- 1:26Pull up the standard analytics teamwork chart most companies are running right now.
- 1:31Fine.
- 1:32Bottom layer, three or four junior analysts whose whole day is turning Slack messages
- 1:36into queries and pasting results into decks.
- 1:40Middle, a couple of senior analysts building dashboards.
- 1:44Top, an analytics engineer maintaining the data models.
- 1:48That's the shape at maybe 70% of mid-sized firms.
- 1:51Which of those layers is already dead?
- 1:53The bottom one as a job description.
- 1:56Not the people.
- 1:57The description.
- 1:58DBT Labs publish survey data suggesting analysts spend over half their time on ad hoc requests.
- 2:04Though they sell the tooling that promises to fix exactly that.
- 2:07So I'd cross-check it.
- 2:09O'Reilly's developer surveys land in a similar zone independently.
- 2:13Either way, the ad hoc request queue is the first thing a language interface eats.
- 2:18And the senior layer?
- 2:19The dashboard builders?
- 2:21Half dead.
- 2:22And this is where CDOs get it wrong.
- 2:24They think dashboards survive because they're pretty.
- 2:27Dashboards survive because someone decided which metric matters and how it's defined.
- 2:32A model will happily compute churn three different ways depending on how you phrase the question
- 2:38and hand you three confident, contradictory numbers.
- 2:41That sounds less like a helpful intern and more like a very fast liar.
- 2:45It's a fast intern who never asks whether the question makes sense.
- 2:49If you ask for revenue by region and your data has customers build in one region but
- 2:54serviced in another, it'll pick one silently.
- 2:57The machine writes the skewl.
- 2:59It does not write the definition of what region means to your business.
- 3:03So the analytics engineer at the top, the one nobody could explain to the CEO, suddenly
- 3:08matters most.
- 3:09They become the whole game.
- 3:11MIT Sloan Management Review has been making this point for a couple of years now.
- 3:17The value moves from producing answers to governing definitions.
- 3:21Sloan has to own the semantic layer.
- 3:23The dictionary that says churn means this, revenue means that, and here's the one blessed way
- 3:28to calculate it.
- 3:30Without it, natural language BI just industrializes disagreement.
- 3:37You stop hiring query writers and start funding definition owners.
- 3:42Take a team of eight.
- 3:43Historically, five juniors, two seniors, one engineer.
- 3:47Redesigned, two people fielding genuinely hard analysis, and four building and policing that
- 3:52shared dictionary of metrics.
- 3:55Same money, completely different job postings.
- 3:58The mistake is keeping the old five and handing them a chatbot.
- 4:01What happens to the CDO who buys the tool and calls it a transformation?
- 4:05In 18 months, they've got a team optimized for a bottleneck that no longer exists, and
- 4:10a business making faster decisions off inconsistent numbers.
- 4:15Need without agreed definitions is just quicker arguments.
- 4:18One thing they should do Monday morning.
- 4:20Take your five most quoted metrics.
- 4:23Revenue, churn, active user, whatever, and check how many distinct definitions are live
- 4:29across your teams right now.
- 4:31If it's more than one per metric, you don't have a tooling problem.
- 4:34You have a dictionary you haven't written yet.
- 4:37Sources for today's episode.
- 4:38Towards Data Science, Cade Nuggets, The New Stack, DBT Labs, Vendor, Data Tooling, O'Reilly
- 4:45Radar, MIT Sloan Management Review.
- 4:49That's a wrap.
- 4:50Fresh CDO briefings drop daily at MBA-training.com.