How long does it take to unify metric definitions across multiple regions?

Eleven weeks for fourteen regions at Kaiser Permanente, from first workshop to a certified metric layer running live. The bulk of that is not build time. Week one was spent getting all fourteen regional leads to agree one definition with the edge cases written down, because no technical work can proceed while that is open.

What does metric disagreement actually cost an organisation?

At Kaiser Permanente it was $2.1M a year in reconciliation labour across fourteen regional teams, and a stalled AI readiness programme. After the certified layer went live that fell to under $180K a year, which is the cost of running the governance council and maintaining the layer.

Do you have to replace the data platform to fix inconsistent metrics?

No. The certified metric layer at Kaiser Permanente went on top of the existing enterprise data warehouse rather than replacing it, so all fourteen regional teams kept their own tools. Clinical reporting ran uninterrupted for the whole eleven weeks.

Kaiser Permanente · Healthcare · Oakland, CA · 11 weeks

14 regional definitions of patient encounter unified in 11 weeks

Previous efforts failed because teams couldn’t agree on definitions. We approached the problem by focusing on governance before technology. We built a certified metric layer spanning 14 regions that maintained consistency without interrupting ongoing clinical reporting.

Challenge

Kaiser had 14 regional data teams, each tracking patient encounters differently. Finance, clinical operations, and compliance couldn’t agree on encounter counts for the same period. This mismatch caused $2.1 million in yearly reconciliation costs and stalled their AI readiness project. Two previous attempts to fix this internally had failed.

Approach

In week one, we gathered all 14 regional leads to align on a single metric definition, explicitly documenting edge cases. Next, we created a certified metric layer on the existing EDW, adding automated lineage tracking. We set up a governance council to resolve future disputes. We kept the current platform intact throughout.

Outcome

The certified metric layer launched in week 11, cutting reconciliation costs from $2.1M a year to under $180K. We enabled the AI readiness initiative, which kicked off six weeks after delivery. Throughout the project, clinical reporting stayed fully operational without any interruptions.

Frankly, the tech wasn’t the problem. It was how we framed the whole thing from day one.

Over eight years, every region was basically doing its own thing with encounter logic. Same data, but what counted as an encounter? Totally different depending on where you looked. The big problem? No one could agree on what an encounter actually meant. So, we pulled everyone into a workshop, hashed out clear definitions everyone could get behind, and made sure to write them down before even touching the code.

We put together a certified metric layer so we could keep using the same platform without any hassle.

We added a semantic layer on top of the existing EDW. Why? To let the 14 regional teams keep doing their thing without any headaches. They stayed with their usual tools but now had one reliable source for consistent, certified numbers. Plus, we kicked off monthly governance meetings to spot and squash any data issues before they got messy.

Our AI project had been stuck for months. So, we rolled up our sleeves, zeroed in on the problem, and finally got it moving again.

The client was stuck because their metrics didn’t match up. So, we created a certified data layer to clear things up. Six weeks after we delivered it, they picked things back up and launched the project the following quarter.

Results

$2.1M Annual reconciliation labor eliminated
11 wks Delivery, zero disruptions
14 Regions on a single metric definition
$180K Residual annual governance cost

We tried to handle it ourselves twice and couldn’t get it done. Thinklytics stepped in, helped us reach an agreement in the first week, and wrapped up the project in eleven weeks. The quarterly reconciliation work we used to do is pretty much eliminated now.
VP of Clinical Analytics, Kaiser Permanente

In the client's terms

The question

Which number is right, and who gets to decide?

The situation

Fourteen regional teams each counted a patient encounter their own way. Finance, clinical operations and compliance could not agree on the number for the same period, and two internal attempts to settle it had already failed.

What we found

The disagreement was not technical. Eight years of independent regional logic had produced fourteen defensible definitions and no authority to choose between them. Nothing downstream could move until somebody decided, and nobody had been given the standing to.

What it produced

The layer went live in week 11. Reconciliation labour fell from $2.1M a year to under $180K. The AI readiness initiative that had been stalled behind the disagreement started six weeks later. Clinical reporting ran uninterrupted throughout.

The roadmap

Week one put all fourteen regional leads in one room and kept them there until a single definition existed, with the edge cases written down rather than assumed. The certified metric layer went onto the existing EDW rather than replacing it, with automated lineage behind every figure. A governance council was stood up to settle the next dispute without another eleven-week engagement.

The work behind this result

Our AI work is not delivering, the problem this engagement resolved.
AI Readiness Assessment & Consulting
Generative AI Consulting
Managed Data Readiness Services
AI Deployment Approval Pack, for whoever has to approve the spend.
The 30 day Corporate Drag and Risk Diagnostic, findings yours either way.