Data contracts: how JPMorgan Chase named data owners

September 4, 2026 · 5 min · Free, with full transcript

Who owns a dataset, and what happens when it breaks? JPMorgan Chase could not answer that for years across hundreds of business lines, so it replaced governance policy documents with data contracts covering more than 50 domains, each with a named person who signs off on format, quality, freshness and escalation. The position here: accountability without a name attached is theater. You get the design that avoids organizational paralysis, contracts strict at the core and light at the edges, where consumers subscribe to a published interface instead of renegotiating each request. You also get the failure mode being avoided, silent schema changes, the lineage case regulators care about, a caution on the vendor-quoted 60% cleanup figure from Monte Carlo and Anomalo, and a one-page contract to write this week.

Listen to the podcast

4 min

Chapters

Key takeaways

  • Name a specific person who signs off that a dataset meets its contract, since making the data team responsible means nobody is.
  • Define a contract once at the publishing domain and let consumers subscribe, so you negotiate the interface rather than every table request.
  • Write the contract to cover format, field meanings, update frequency and who gets paged when it breaks.
  • Treat an undeclared schema change, like renaming customer_id, as a breach that is flagged before it ships.
  • Start with the one dataset that already causes fights between finance and marketing, one page and two signatures, instead of launching a 50-domain program.
Read the full transcript

Host:Welcome to the Leaders Insights Podcast. Today's episode, how JPMorgan Chase built data contracts across 50-plus domains. JPMorgan Chase just published how they built data contracts across more than 50 business domains. Before we go further, what's a data contract in plain English?

Expert:It's a formal agreement between the team that produces a dataset and the teams that consume it, spelling out the format, the quality, the update frequency, and who's on the hook when it breaks. Think of it like a service level agreement, a promise with consequences, but for data instead of uptime.

Host:Fine, but JPMorgan is a bank with hundreds of business lines. Why is this news in September 2026 and not something they sorted out a decade ago?

Expert:Because for years they couldn't answer the most basic question. Who actually owns this data? They had hundreds of business lines all touching the same customer records and nobody accountable when the numbers didn't line up. What changed is they stopped treating governance as a policy document and started treating it as a contract with named owners across 50-plus domains, a domain being a coherent slice of the business, like mortgages or card payments.

Host:Named owners sounds like a euphemism for someone to blame.

Expert:It is, and that's the point. Balance without a name attached is theater. If your credit risk data goes stale and 40 downstream teams inherit garbage, the data team being responsible means nobody is. JPMorgan forced the question, which specific person signs off that this dataset meets the contract? That's uncomfortable, which is exactly why most banks avoid it.

Host:So they solved accountability. What did it cost them in speed? Because the classic failure here is you formalize everything and the whole place seizes up.

Expert:That's the trap. Organizational paralysis, where every data request needs three committees and a prayer. Their move was to make contracts lightweight at the edges and strict at the core. A domain publishing a dataset defines its contract once, and consumers subscribe against it. You're not renegotiating every time someone wants a table. You negotiate the interface, then you leave it alone.

Host:Give me the failure mode they were escaping. What actually breaks without this?

Expert:Silent schema changes. Someone in one team renames a column from customer underscore ID to cussed underscore ID on a Tuesday. And by Wednesday, 12 dashboards are wrong and the fraud model is quietly degrading. Nobody told anyone because there was no contract saying you can't do that without notice. With a contract, that change is a breach. It gets flagged before it ships, not discovered three weeks later in a regulatory report.

Host:You mentioned regulators. Is this compliance dressed up? Or does it genuinely move the needle for a CDO, a chief data officer, the person who answers for all of this?

Expert:Both, honestly. In banking, regulators expect you to trace where a number came from. Lineage, they call it. The full paper trail of how data got from source to report. Data analytics give you that almost for free because every data set has a declared owner and a declared shape. But the real win is internal. Your data scientists stop spending 60% of their time cleaning inputs they don't trust. That's the number vendors like Monte Carlo and Anomolo love to quote. Treat it as directional since they sell the fix. But the direction is right.

Host:60% from vendors trying to sell reliability tools. Noted. So what should a CDO who isn't JP Morgan actually do Monday morning? They don't have that budget or headcount.

Expert:Don't start with 50 domains. Start with one data set that already causes fights. The one where finance and marketing show up to the same meeting with different revenue numbers. Write a one-page contract for it. Who owns it? What the fields mean? How fresh it is? And who gets paged when it's wrong? Get the two teams to sign.

Host:One page, one data set, two signatures.

Expert:Write. The mistake is treating governance as a program you launch. It's a habit you build one contract at a time. JP Morgan didn't wake up with 50 domains covered. They earned each one by proving the last contract made someone's life measurably less miserable.

Host:So the takeaway?

Expert:Pick your most argued about data set, write a one-page contract with a named owner, and make two teams sign it this week. Everything else is procrastination.

Host:That's it. Your decision tools are live at MBA-training.com.

Read and go deeper