Latest / MBA Training Data / How JPMorgan Chase built data contracts across 50+ domains
Transcript
- 0:00Welcome to the Leaders Insights Podcast.
- 0:07Today's episode how JPMorgan Chase built data contracts across 50-plus domains.
- 0:12JPMorgan Chase just published how they rolled out data contracts across more than 50 business
- 0:17domains.
- 0:18For a bank that big, that's either a triumph or an admission of how bad things were.
- 0:23Which is it?
- 0:24Both.
- 0:25Honestly.
- 0:26A contract, a formal agreement between the team producing data and the teams consuming
- 0:30it about what the data means and how it'll behave, is only necessary when you've already
- 0:35got chaos.
- 0:37JPMorgan had hundreds of business lines, all defining customer differently.
- 0:43This is them cleaning up a mess they let fester for a decade.
- 0:47Defined differently.
- 0:48Because every executive says their data is a mess, and it usually just means the dashboard's
- 0:52ugly.
- 0:53No, this is deeper.
- 0:55One team counts a customer at account opening.
- 0:58Another counts them only after the first transaction clears.
- 1:02So the retail division reports 12 million customers.
- 1:05The risk team reports 10 and a half.
- 1:08And both are technically right.
- 1:10When your regulator asks how many customers you have, it depends who you ask is not an
- 1:15answer you want to give.
- 1:16So a data contract fixes that.
- 1:18How?
- 1:19Mechanically?
- 1:20Walk me through what one actually contains.
- 1:22Four things, roughly.
- 1:24The schema, the fields and their types, so account underscore balance, is always a number
- 1:29and never sometimes text.
- 1:32The semantics, what the field actually means in plain English.
- 1:36The quality guarantees, like, this is never null, updated within two hours.
- 1:42And ownership, a named human who answers when it breaks.
- 1:46That last one is what most companies skip, and it's the whole point.
- 1:50Why is a name the whole point?
- 1:51Sounds like the least technical part.
- 1:53Because the real problem isn't technical, it's the ownership vacuum.
- 1:58When a data feed goes bad and 40 downstream teams are affected, everyone assumes someone
- 2:03else owns it.
- 2:05A contract forces one team to sign their name.
- 2:08At JP Morgan, the friction wasn't building the schemas.
- 2:11It was engineers realizing they were now on the hook for data they'd been quietly neglecting
- 2:16for years.
- 2:17Which means they resisted.
- 2:19How do you make an engineer volunteer to be accountable for something painful?
- 2:22You don't make them volunteer, you make it worse to opt out.
- 2:26JP Morgan tied contracts to their data mesh.
- 2:29Why?
- 2:30Yes.
- 2:31That's the architecture where each business domain owns and serves its own data as a product,
- 2:35instead of dumping it into one central lake.
- 2:38If you want to publish data other teams can use, you publish a contract.
- 2:42No contract, no distribution.
- 2:45Suddenly, signing your name is the cost of being visible.
- 2:48The article calls it a practical blueprint.
- 2:51The JP Morgan has a technology budget bigger than some countries' GDP.
- 2:56What can a company with a normal budget actually copy?
- 2:59The sequencing, not the scale.
- 3:01Don't try 50 domains at once.
- 3:03Pick the three or four datasets that everyone fights over, usually customer, product, and
- 3:08revenue definitions, and contract those first.
- 3:12JP Morgan didn't boil the ocean either.
- 3:15They started where the inconsistency was costing them regulatory pain and expanded from
- 3:19there.
- 3:20The tooling is almost irrelevant.
- 3:23You can write a contract in a shared document if the accountability is real.
- 3:27A shared document?
- 3:29Vendors selling data governance platforms would hate hearing that.
- 3:32They would.
- 3:33And I'd flag that most of the you-need-our-catalog numbers come straight from those vendors'
- 3:37own marketing.
- 3:39A tool helps you enforce contracts at scale, chuainain checking automatically that the
- 3:43data still matches the promise.
- 3:45But if nobody's named as owner, the fanciest platform just automates the finger-pointing.
- 3:51The organizational commitment comes first, the software second.
- 3:55People buy it in the wrong order constantly.
- 3:57So the tool is the reward for getting the culture right, not the shortcut around it?
- 4:02Exactly.
- 4:03JP Morgan spent years on the human side before the mechanics clicked.
- 4:07The banks that bought a platform first are still generating beautiful dashboards of data
- 4:11nobody trusts.
- 4:13Give me the one thing a listener does Monday morning.
- 4:15Take your most argued-about metric, the one where two teams show different numbers in every
- 4:20meeting, and write down, on one page, who owns it, and exactly what it means.
- 4:25That single page is a data contract.
- 4:28Everything else is scale.
- 4:29That's it.
- 4:30The CDO Decision Tools are live at MBA-training.com.