When Manual and Digital Collide: The Reconciliation Problem Nobody Budgets For

Every health plan I talk to has a retrieval calendar and a digital data program. They have vendors for chart pull, feed agreements for EMR and lab data, and a quality team managing both.

What almost none of them have is a formal answer to this question: what happens when those two programs touch the same member — and tell two different stories?

That question is the reconciliation problem. And it is one of the most expensive operational gaps in payer quality today — not because it is difficult to solve, but because it has never been formally named, owned, or budgeted for.

What Reconciliation Actually Means

Reconciliation, in this context, is not de-duplication. De-duplication is identifying that the same member appears in two systems. That is the easy part and most platforms handle it adequately.

Real reconciliation is this: for a given member, on a given measure, data is arriving from multiple sources — a retrieved chart, a CCD feed, a lab result, a claims record — and those sources do not tell the same story. They overlap in some places, contradict in others, and leave gaps that neither source covers completely. Reconciliation is the work of determining what is true, what is actionable, and what gets submitted.

That is a governance problem, not a technology problem.

Here is why that distinction matters. When health plans encounter reconciliation challenges, the first instinct is to reach for a technology solution. Better integration layer, smarter matching logic, more sophisticated platform. Technology plays a role. But technology can execute a decision. It cannot make one.

When your chart retrieval vendor's data and your digital feed vendor's data conflict on the same member — when the chart says a procedure was completed and the feed shows no corresponding billing — a system can flag that conflict. It cannot resolve it. Resolution requires a decision about which source is authoritative, under what conditions, and who owns that call. That is governance. And governance is not in either vendor's statement of work.

Why Nobody Budgets For It

Health plans contract for data services in silos.

Chart retrieval: a vendor covers chase list management, provider outreach, record collection, and delivery to the abstraction team. Defined scope. Clear deliverables. Specific SLAs.

Digital feeds: different vendors cover feed configuration, data normalization, ingest validation, and delivery to the warehouse. Also defined. Also clear.

What is in neither contract is what happens when a member has data in both programs that tells two different stories. The chart retrieval vendor delivers what they collected. The digital feed vendor delivers what they ingested. The question of what to do when those two deliveries disagree about the same member lands on the internal quality team.

Both vendors have done their job. Both will tell you they've done their job. Both are correct. And nobody has done the job.

The internal quality team — operating at or above capacity during HEDIS season — handles reconciliation at the margin. By whoever has time. After everything else. This is why reconciliation does not appear on any vendor invoice and does not appear as a line item in any quality operations budget. It shows up as overtime, delayed submissions, and abstraction errors that make it through to the final file.

Because reconciliation is not budgeted, it is not measured. And because it is not measured, plans have no visibility into how much of their Stars performance gap is attributable to reconciliation failures versus other causes.

The Unresolved Member

There is a specific population that every plan has and almost none have a protocol for.

An unresolved member is a member where chart retrieval has partial information about a gap, a digital feed has different partial information about the same gap, neither source alone is sufficient to close the measure, and no one has formally reconciled the two. The member is not in the closed column. They are not in the active outreach queue. They exist in a data limbo that most quality management systems do not have a category for.

They exist because the two programs that touch them do not share a status layer. The chart retrieval vendor marks the chart as delivered and moves on. The digital feed vendor marks the record as ingested and moves on. Neither vendor has visibility into what the other delivered. So neither knows that together, their two partial contributions might be sufficient to close the measure — if someone would sit down and reconcile them.

What makes this population financially significant is where they concentrate. Unresolved members cluster in the hardest-to-close segment — members with complex clinical histories, multiple providers across multiple systems, fragmented documentation spread across inpatient, outpatient, and specialist encounters. These are the members most likely to have partial data in both retrieval and digital, precisely because their care events are documented in more places than any single data stream captures completely.

They are also the members with the highest potential Stars impact. The hardest-to-close members have typically been open the longest. Closing an unresolved member is not just a current-year win. It is a correction that can shift a plan's performance trend line.

Why Your Stars Forecast May Already Be Wrong

Here is the argument that most plans have not confronted directly.

Your Stars performance forecast — the model used to project end-of-year rates, allocate resources, and set targets — is built on historical closure data. That is standard practice. It is also where the reconciliation problem becomes a financial model problem.

Historical closure data contains two systematic errors introduced by unmanaged reconciliation. They run in opposite directions and compound.

First: double-counted closures. When a measure closes, the system records it. But it may not record which data source produced it. If chart retrieval and a digital feed both touched the same member for the same measure, and the measure eventually closed, the system may attribute the closure to whichever source processed it last. A plan may have paid for two programs to close a gap that one program would have caught alone. Nobody knows, because attribution was never tracked at the source level.

Second: undercounted closures. Unresolved members — the ones where both sources have partial data and neither alone is sufficient — did not close. They stayed open. But the reason was not that the care was not provided. It was that the reconciliation never happened to connect the data that was already there. So historical closure rates for those measures are artificially low. The model assumes a certain percentage of the population cannot be closed based on history. But some of those members can be closed. They are waiting for reconciliation work that has never been funded.

When a Stars forecast is built on a baseline with both errors present simultaneously, the model is wrong in two directions at once. Plans overspend on some measures and underspend on others, based on rates that were never accurate to begin with.

Three Things That Actually Fix It

The reconciliation problem is not complicated once it is named. Here is what the plans that have solved it actually do.

First: they make reconciliation a defined function with an owner and a budget. Not a task that falls to whoever is available. A named function with someone explicitly accountable for identifying members with data in multiple streams, diagnosing why gaps remain open, and initiating the follow-up work to close them. Without ownership, it does not get done consistently. And consistency is what survives an audit.

Second: they build a shared member-level status layer above both programs. Chart retrieval and digital feeds need to be visible to each other at the member level, in real time. Not a weekly sync between vendor portals. A shared status layer that tracks which sources have touched each member for each measure, what each source contributed, and what remains unresolved. This is what makes the reconciliation function operationally possible.

Third: they measure reconciliation yield. How many unresolved members were identified in a measurement cycle, how many were successfully reconciled, and how many closures came from reconciliation work that neither program alone would have produced. That number is the return on investment for reconciliation infrastructure — and the number that makes the internal business case for funding it properly next year.

Where This Sits in the Transition to ECDS

A note on timing that matters for planning.

NCQA is transitioning hybrid measures to ECDS-only reporting by MY2029. COL, CCS, CIS, and IMA are already ECDS-only. For measures that have fully transitioned, chart retrieval is no longer part of the workflow. One fewer data stream to reconcile.

But the transition period — from now through MY2029 — is when reconciliation complexity is at its highest. Plans are operating ECDS for measures that have transitioned while running hybrid programs for the eight measures still in traditional reporting. More streams. More potential for conflict. More unresolved members.

The plans that invest in reconciliation infrastructure now will arrive at the ECDS era with clean historical baselines and accurate attribution data. The plans that do not will carry corrupted baselines into a new reporting environment and spend the first two ECDS cycles trying to explain why their models keep being wrong.

The Question to Take Back

Pull your open gap list for any measure where you run both chart retrieval and digital feeds.

Ask how many of those members have data in both programs.

Then ask: what is the documented protocol for what happens to those members?

If the honest answer is somewhere between "the team handles it" and a long pause — that is your reconciliation gap. That is where Stars performance is leaking in a way that no vendor's report will surface for you.

Go Deeper — This article is based on Episode 5 of The Execution Gap Podcast: "When Manual and Digital Collide: The Reconciliation Problem Nobody Budgets For." 16 minutes on the governance gap that lives between your vendor programs, the unresolved member population every plan has but nobody tracks, and why your Stars forecast may already be built on historical data that was never clean. Listen to Episode 5 →

How Podero Health Approaches This

Podero Health gives plans a single execution layer that validates data at the measure level, enforces source hierarchy, and integrates digital collection and manual retrieval into one pipeline — so the latency and conflict that drive reconciliation failures do not exist in the workflow.

If you want to see what that looks like against your actual gap population — your feeds, your measures, your member data — that is what a pilot conversation is designed to surface.

Come with your open measures and your current vendor setup. That is exactly where we start.