Why Data Is the Missing Layer in Self-Funded Health Plans

Self-funded employers generate more claims data than they have ever had access to. Most of them are not using it. Data without analysis is just noise, and noise does not bend a cost curve.

The irony of self-funded health plans is hard to overstate. Employers switch to self-funding partly to gain visibility into claims data. They get that data. Then the monthly reports sit in a shared drive, skimmed once by someone in HR, and forgotten until renewal season forces a scramble.

According to the 2025 KFF Employer Health Benefits Survey, 67% of covered workers are enrolled in self-funded plans (KFF, January 2026). These employers receive detailed claims files every month containing diagnosis codes, procedure codes, provider charges, pharmacy utilization, member demographics, cost-sharing information, and more. That data represents one of the most powerful cost-management assets in the entire benefits ecosystem.

And most of it goes unused.

The data gap is not about access

The problem is rarely that employers cannot get claims data. Third-party administrators provide monthly and quarterly reporting as a standard service. Stop-loss carriers generate large-claimant reports. PBMs produce pharmacy utilization files. The data flows in from multiple directions.

The problem is that raw claims data is complex, fragmented, and useless without analytical frameworks. A CSV file with 50,000 rows of claim lines does not tell an employer which chronic conditions are driving the most spend, which providers are billing at outlier rates, which members are trending toward high-cost events, where pharmacy waste is concentrated, which plan design features are influencing behavior, or whether the DPC investment is producing measurable returns.

Turning that raw data into actionable intelligence requires cleaning, normalization, risk adjustment, benchmarking, and interpretation. Most employers lack the internal capability to do this. And most TPAs, while competent at claims administration, provide basic summary reporting rather than strategic analytics.

The result is a structural gap: employers sitting on valuable data assets with no mechanism to extract value from them.

What claims data can actually reveal

When properly analyzed, self-funded claims data answers questions that are otherwise invisible.

Where is the money going? Claims data can break total plan spending into granular categories: inpatient, outpatient, ER, physician, pharmacy, behavioral health, and ancillary.

Within each category, the data reveals which specific diagnoses, procedures, and providers are driving cost. An employer might discover that 40% of its hospital spend comes from three facilities, or that musculoskeletal claims represent the single largest cost category, not because of frequency, but because of a handful of surgical cases at high-cost providers.

Who is driving the spend? In most employer health plans, approximately 5% of members generate 50% or more of total claims. Identifying these high-cost claimants is not about penalizing them. It is about understanding the clinical conditions driving their costs and intervening with care management, case coordination, or alternative treatment pathways before costs escalate further. Predictive analytics can flag members trending toward high-cost events based on medication adherence patterns, rising lab values, or utilization trajectories (Taylor Benefits Insurance, "Predictive Analytics for Chronic Condition Management," March 2026).

Are vendors performing? Claims data allows employers to evaluate whether their DPC investment is actually reducing downstream costs, whether their PBM's formulary is optimizing for plan cost or PBM revenue, whether care navigation is diverting members from high-cost to high-value providers, and whether wellness programs are producing measurable health improvements or just participation numbers.

How does the plan compare to benchmarks? Without external benchmarks, an employer has no way to know whether spending $18,000 per employee is good, bad, or average for its industry, region, and workforce demographics. Analytics platforms and consultants can benchmark a plan's performance against comparable employer populations, revealing where the plan outperforms and where opportunities exist. One case study found that Quest Diagnostics' self-insured plan was 11% more efficient than the mean for a comparable benefit after implementing data-driven interventions, and 1% more efficient than the top quartile of the study pool (PMC, "Self-Insured Employer Health Benefits Strategy," 2019).

The four levels of health plan analytics

Analytics professionals describe four progressive levels of capability, and most employers are stuck at level one.

Level 1: Descriptive analytics. This is the baseline that most TPAs provide. Summaries of past data, including counts, averages, and percentages. Claims paid by category. High-cost claimant lists. Monthly trend reports. Descriptive analytics answers the question: "What happened?" (W3 Insurance, "The Importance of Health Plan Analytics," July 2024).

Level 2: Diagnostic analytics. This digs into why costs are rising. Why did pharmacy spend increase 14% last quarter? Because three members started GLP-1 medications and one member began a specialty oncology drug. Why did inpatient costs spike in Q3? Because two members had unplanned surgical admissions at high-cost facilities. Diagnostic analytics moves from "what happened" to "why it happened."

Level 3: Predictive analytics. This is where data starts generating proactive value. Predictive models analyze medical and pharmacy claims, demographic information, utilization patterns, chronic condition prevalence, and provider trends to forecast future healthcare needs and costs (Taylor Benefits Insurance, "Machine Learning Models for Forecasting Health Plan Costs," April 2026). Machine learning algorithms can flag members at risk of developing chronic conditions, identify probable high-cost claimants before claims materialize, and forecast plan costs with greater accuracy than traditional actuarial methods.

Predictive models depend on multiple data sources working together. In employer health plans, these may include medical and pharmacy claims, eligibility files, lab results, biometric screening data, health assessments, and information from care management platforms. The strongest insights come from integrated data rather than isolated data points (Taylor Benefits Insurance, March 2026).

Level 4: Prescriptive analytics. The most advanced level goes beyond forecasting to recommend specific actions. If a predictive model identifies 15 members trending toward diabetic complications, prescriptive analytics suggests the optimal intervention: which members should be referred to the DPC physician for medication adjustment, which should receive outreach from a nurse navigator, and which might benefit from a specialty care management program. This level integrates data with clinical protocols and plan design to generate tailored recommendations.

Most employers operate at Level 1, with some reaching Level 2. Employers who reach Levels 3 and 4 operate their health plans with a fundamentally different level of precision, and their cost trajectories reflect it.

Why most employers stay stuck at Level 1

Three barriers keep employers from advancing their analytics capability.

The TPA limitation. Most TPAs are built for claims administration, not strategic analytics. They process claims efficiently, produce compliance reports, and generate standard utilization summaries. But they are not analytics companies. Their reports answer the questions the TPA thinks are important, which may not align with the questions the employer needs answered.

Self-funded plans increasingly expect real-time visibility into claims data, faster reporting cycles, and clearer insights into cost drivers (MagnaCare, "4 Key Trends Impacting Self-Funded Health Plans in 2026," February 2026), but many TPAs have not invested in the platforms to deliver this.

The data fragmentation problem. Medical claims come from the TPA. Pharmacy data comes from the PBM. Biometric screening data comes from the wellness vendor. DPC utilization data comes from the DPC provider. Disability and workers' compensation data sits in separate systems entirely. Integrating these data streams into a unified analytical environment requires technical capability and ongoing maintenance that most mid-size employers do not have in-house.

The interpretation gap. Even when data is aggregated and analyzed, someone has to translate the findings into strategic action. A report showing that musculoskeletal claims represent 22% of total spend is useful. A recommendation to implement a musculoskeletal management program, steer surgical cases to centers of excellence, and add physical therapy benefits with $0 copays is actionable. The gap between data and decision is where most self-funded plans lose momentum.

How to close the gap

Employers do not need to build an internal data science team. They need the right partners and a commitment to using data as a management tool rather than a compliance artifact.

Invest in an analytics platform or partner. Several platforms specialize in employer health plan analytics, integrating claims data from multiple sources and providing dashboards, benchmarking, and predictive modeling. Some TPAs have built or acquired these capabilities.

Others can be layered on top of existing TPA reporting. The key requirement is access to detailed claims data through the employer's claims administrators (Reclaim Health, 2025). HR teams and brokers can now access real-time data on claims trends, pharmacy spend, high-cost conditions, and member engagement, and use it to make smarter, more strategic benefit design decisions (PAI, "2026 Outlook," August 2025).

Demand more from existing vendors. Most TPA contracts include data access provisions. Employers should request monthly claims files in a usable format (not just PDF summaries), pharmacy utilization reports with NDC-level detail, high-cost claimant trending with clinical context, provider cost and quality benchmarking, and year-over-year trend analysis by category. If the TPA cannot provide this level of detail, it may be time to evaluate alternative administrators.

Hire or engage analytical talent. A benefits consultant with strong analytical capabilities can serve as the translation layer between raw data and strategic action. The best consultants review data monthly, present findings quarterly, and recommend plan design adjustments annually based on actual claims experience.

Participate in collaborative data initiatives. In November 2025, the National Alliance of Healthcare Purchaser Coalitions and the Health Care Cost Institute announced the Employer Health Claims Collaborative, a purchaser-driven database designed to give self-funded employers a clearer view of health care spending, utilization, and benefit performance. Set to go live in January 2026, the Collaborative combines the National Alliance's network of business health coalitions (representing employers covering more than 90 million Americans) with HCCI's expertise in data analytics and translation. Initiatives like this reshape how employers use claims data to drive smarter decisions.

What to measure (and how often)

A self-funded employer committed to data-driven management should track these metrics at minimum:

Monthly: Total claims paid per employee per month (PEPM). Pharmacy cost PEPM. High-cost claimant count and trending. ER visit rate per thousand members.

Quarterly: Claims by category (inpatient, outpatient, pharmacy, behavioral health). Top 10 diagnosis categories by spend. Provider cost benchmarking for high-volume services. DPC and care navigation program utilization and engagement.

Annually: Total cost of care per member versus industry benchmarks. Year-over-year trend by category (compared against Mercer, KFF, and Business Group on Health national averages). Vendor performance reviews (TPA, PBM, stop-loss, DPC, navigation). Plan design effectiveness (are incentives driving the intended behavior?). Predictive risk assessment for the upcoming year.

The 2025 KFF survey reported that the most common employer priorities for managing health programs over the next few years were "greater focus on managing high-cost claims" and "measuring the performance of health programs to ensure they provide value" (KFF, 2025). Both priorities are impossible to execute without robust analytics.

Data as competitive advantage

The employers who build a data infrastructure today are creating an advantage that compounds over time. In year one, analytics identifies the biggest cost drivers. In year two, targeted interventions begin bending the curve. In year three, predictive models enable proactive management rather than reactive response.

Meanwhile, employers without analytics are flying blind. They react to renewal increases without understanding the underlying drivers. They invest in wellness programs without measuring ROI. They maintain PBM contracts without benchmarking drug costs. They accept TPA reporting at face value without questioning whether the data tells the full story.

Mercer's national survey found that health benefit costs per employee are expected to rise 6.7% in 2026, the highest increase since 2010, with average costs exceeding $18,500 per employee (Mercer, November 2025). Employers using data to manage their plans proactively can hold trend to 2-4%. Employers managing passively will absorb the full market increase, year after year.

The difference between 3% trend and 8% trend on a $5 million health plan is $250,000 per year. Over five years, that gap exceeds $1.5 million. That is the cost of not using data.

The bottom line

Self-funded employers have the raw material for sophisticated cost management sitting in their claims files. What most of them lack is the analytical layer that transforms data into decisions. That layer, whether it comes from a dedicated analytics platform, a strong benefits consultant, or an advanced TPA, is what separates self-funded employers who merely shift funding risk from those who actively manage it.

Data does not reduce costs on its own. But without data, every other cost containment strategy is a guess. And in a market where healthcare costs are rising 7-10% annually, guessing is an expensive habit.