From 72 Hours to 5: What a Real Data Platform Does for a Healthcare Business

There is a report that goes out every quarter to every one of our client's customers. Thousands of decks, each one tailored to that customer. It is contractually required. It ran on deprecated technology. It took 48 to 72 hours of hand-holding to produce. And the person who built it had left the company.

That is where a lot of data platform conversations actually start. Not with a strategy deck. With a report that is one bad quarter away from becoming 3,000 PowerPoints built by hand.

This is the story of what we built for a healthcare benefits company, and what a working data platform changed for them. Real numbers, told with their program lead's blessing.

The starting point

The environment will sound familiar to anyone running a mid-market healthcare business. Transactional data lived in separate systems: SQL Server, Salesforce, NetSuite, files arriving from outside partners. Teams pulled data by hand and shaped it in Excel. Business definitions were embedded in code across 487 individual Tableau workbooks, so two reports could answer the same question differently. Storage costs kept climbing because there was nowhere to offload anything.

None of that is a technology problem yet. It is hundreds of small decisions that made sense at the time, compounding.

How the engagement started

The client's program lead evaluated eight firms. Here is how she describes what made this one different:

You were the only one in the pile of eight that wasn't trying to talk about solutioning upfront. Everyone else had done Databricks a million times, or came in saying "I've got a Snowflake, this is what we're going to do." You were actually asking, what's the problem?

Jac Carrier, Senior Data Architect and independent program lead

We ran a discovery. We gathered requirements across teams, not just the BI team that owned the immediate project. We set up working sessions with the product vendors, debated the options together, and wrote up a recommendation. The answer turned out to be Microsoft Fabric, a data warehouse and lakehouse in one platform. We did not walk in the door knowing that. That is the point of discovery.

The build that followed connected an on-premises SQL Server, Salesforce, two NetSuite environments, Amazon S3, Azure Blob storage, network file shares, and Exchange mailboxes into one governed platform, with development, UAT, and production environments and every change in source control. Full implementation took under six months. The platform now holds three terabytes and runs on reserved capacity for about $2,500 a month.

Here is what that foundation made possible.

Story one: the quarterly customer report

The report from the opening of this post. We automated the entire pipeline: ingesting NetSuite data, applying the business logic and period calculations, calling the Power BI API, and dropping a customized three-slide deck for every customer onto the share drive.

The old process took 48 to 72 hours per quarter, plus a person to escort it through every step. The new one finishes in 3 to 5 hours with nobody touching it. The step that processes raw data covers every customer, every quarter, back to the beginning of their history, in about six minutes. The client gave us three months. It shipped in two.

One more thing that build surfaced: when the client audited the new output line by line, almost every discrepancy they found traced back to upstream data issues the old process had been carrying silently for years. They now have monitoring that catches those the day they appear.

Story two: six weeks to 45 minutes

The company owner needed every claim connected to a specific list of drugs. Simple question, brutal answer: the systems had no mapping between claims and drugs. Drug names lived in free-text description fields, spelled however the person typing them spelled them that day. The last time a request like this came in, it took four to six weeks.

David, who led the build, had already constructed a fuzzy matching mechanism for exactly this kind of free-text data. He wrote a custom script against the ingested tables, scored the matches, kept everything above a 95 percent confidence threshold, and delivered the list. Elapsed time: about 45 minutes.

That capability now sits in the platform permanently. For a business that works with drug claims every day, that is not a party trick. It is a standing ability the organization never had.

Story three: finding the money

Invoices lived in nested folders. Confirming that every invoice actually went out to the right customer by email was a manual audit, four to five days of one person's time, and the volume meant only a small sample ever got checked. Missed invoices meant missed revenue, discovered months later or never.

The platform now pulls the metadata from those nested folders into a queryable table and reads the accounts receivable mailbox alongside it. The automated audit built on that is rolling out now, and it takes coverage from a small sample to everything. A companion process validates claims between NetSuite and SQL daily, maps statuses across systems, and hands the operations team a specific worklist each morning: these are the records in an invalid state, go fix them. Before, errors surfaced when a customer reported them. Now they surface before the invoice goes out.

Why this worked

Healthcare data raises the stakes on all of this. Here is Jac again:

Digital2DNA comes equipped with solid HIPAA, PHI, and security capabilities just as a base. There was no need to teach them healthcare. They already understood HITRUST, SOC 2, all of it, and it was vetted during discovery. I never worried about it once during the project.

Jac Carrier

And on what the engagement felt like:

I sometimes have to remember, oh yeah, you're a third party. We come through as such a gelled team that I forget.

Jac Carrier

When priorities changed mid-project, and they did, including an urgent pivot to bring call center data under control for a new executive, the platform and the team absorbed it without stalling. That flexibility is the delivery model doing its job.

Her recommendation score, when I asked: 10 out of 10.

What to take from this

If your quarterly reporting is a ritual with a single point of failure, if answering an executive's question takes your BI team a month, if you find revenue problems only when a customer tells you about them, those are not staffing problems. They are platform problems, and they are fixable in months, not years.

We wrote up the full technical build, including the architecture, the integration list, and the cost profile. If you want to talk through what this looks like on your data, Get in Touch. You can also read more about how we work and what we do in healthcare.