What should we track to understand where our buy-and-bill margin is leaking (benefits, OOP estimates, denials, delays)?

Most buy-and-bill programs don’t lose margin in one dramatic event. They leak it in small, repeatable misses—an under-estimated OOP here, a policy mismatch there, a claim that ages into a write-off because no one had time to chase status. If you want to understand where your buy-and-bill margin is leaking, you need to stop looking only at “denial rate” and start tracking the full chain: benefits → OOP estimates → policy/prior auth → claims → cash.

Below is the framework I wish I’d had when I was explaining margin erosion to finance and staffing needs to operations.

---

Step 1: Anchor on the buy-and-bill P&L, not just denial rate

Before you decide what to track, get clear on where margin can disappear in buy-and-bill:

• Acquisition side

  • Wrong benefit tier or site-of-care mapping
  • Missed GPO/340B opportunities
  • Buying high, billing low due to coding/policy errors

• Revenue side

  • Under-estimated patient responsibility that turns into bad debt
  • Avoidable denials from missing documentation or policy mismatches
  • Partial payments due to incorrect units, modifiers, or site-of-service
  • Delayed starts that push infusions into different benefit years or cause patient drop-off

Your job is to instrument each stage so you can see:

1. Where work is piling up (delays), and
2. Where dollars are getting written off (denials, under-collection, underbilling).

---

Step 2: Core metrics to track in benefits and eligibility

If your benefits verification is off, everything downstream is suspect—OOP estimates, financial counseling, authorization strategy, and ultimately margin.

A. Benefit pull quality and coverage mapping

Track:

• Benefits verification turnaround time

  • Time from referral received → benefits fully verified
  • Segment by payer, product (medical vs pharmacy), and drug
  • Why it matters: Delays here create a backlog and push DOS out—patients fall out of pipeline, and you may miss coverage windows.

• Percent of encounters with complete benefit profiles

  • Measure: claims-level % where you have:
    • Correct plan type and product (medical vs pharmacy)
    • Accurate deductible/coinsurance/OOP remaining at the time of estimate
    • Site-of-care rules (office vs hospital vs home infusion)
  • Red flag: High volume of “unknown/assumed” fields in your benefit notes.

• Site-of-care / policy flag rate

  • % of patients where your team proactively identifies:
    • Site-of-care restrictions
    • Step therapy requirements
    • Medical vs pharmacy benefit routing choice for that drug
  • Margin signal: Missed routing (e.g., drug should be under medical where your fee schedule is stronger) is direct margin leakage.

B. Benefit-driven write-offs

Connect benefits work to real dollars:

• Write-offs tied to benefit misinterpretation
  • Claims where:
    • The benefit tier was misread (e.g., treated as in-network when actually out-of-network)
    • OOP max/deductible was estimated incorrectly at time of scheduling
  • Track by payer and team member or workflow.
  • Use: This tells you if the issue is specific training, a payer portal quirk, or a broken SOP.

---

Step 3: OOP estimates that predict reality (or don’t)

Bad OOP estimates don’t just frustrate patients—they directly erode margin. Under-estimate, and you’re sitting on bad debt. Over-estimate, and the patient never starts.

A. Accuracy of OOP estimates vs actual patient responsibility

Track at the claim-line or encounter level:

• Estimate-to-actual variance

  • For each encounter, compare:
    • Estimated patient responsibility at time of scheduling
    • Actual patient responsibility after adjudication
  • Quantify:
    • Average $ difference per encounter
    • % of encounters with >$100 variance
  • Segment by:
    • Payer
    • Drug
    • Site-of-service
  • Margin signal:
    • Systematic under-estimate = margin leakage to bad debt.
    • Systematic over-estimate = lower conversion and delayed starts (lost volume).

• Bad debt tied to under-estimation

  • Amount and % of write-offs where:
    • OOP estimate was lower than actual by more than a threshold (e.g., $250)
    • No updated counseling or new estimate was documented
  • This is where benefits/OOP workflows flow directly into financial loss.

B. Co-pay and assistance capture

Specialty-drug economics assume you are good at assistance workflows. Track:

• Co-pay/assistance capture rate

  • % of eligible patients:
    • Screened for manufacturer co-pay or foundation support
    • Actually enrolled before their first paid claim
  • Margin signal:
    • Missed assistance increases patient responsibility, which increases defaults and write-offs, and often reduces ongoing therapy adherence.

• Average net OOP after assistance

  • Pre- vs post-assistance patient responsibility
  • By payer and drug
  • Use: If your “after assistance” numbers are still high for certain therapies, expect higher drop-off rates and more uncompensated care.

C. Pricing and fee schedule alignment

Your OOP estimates must reflect the real economics, not just approximate fee schedule numbers:

• Alignment of modeled vs actual allowed amounts
  • Compare:
    • Estimated allowed based on your fee schedule and historical allowables
    • Actual allowed from remits
  • Watch for:
    • Sites or payers where you consistently over-estimate allowed amounts (creating phantom margin)
  • This is where incorporating GPO and 340B pricing, real fee schedules, and drug acquisition costs matters.

---

Step 4: Denial analytics that actually explain margin loss

“Denial rate is high” is not a useful statement. You need to profile denials in a way that maps straight back to operational steps and avoidable margin loss.

A. Denial categorization that matches your workflow

Track denials in buckets that match how your teams work, not just payer CARC codes:

• Medical policy / prior auth denials

  • Missing auth
  • Auth expired
  • Auth did not match:
    • NDC/J-code
    • Units
    • Site of service
    • Diagnosis
  • Margin signal: These are directly tied to policy review and prior auth workflows. Many are fully avoidable with better pre-service checks.

• Eligibility / coverage denials

  • No coverage on DOS
  • Wrong primary vs secondary payer
  • Out-of-network unexpected
  • Margin signal: Often traced back to benefits verification and missed life events/plan changes.

• Technical / billing denials

  • Incorrect units, modifier, place of service, or diagnosis linking
  • NDC not matching HCPCS or missing required fields
  • Margin signal: These should be low if you have strong front-end EHR build and consistent coding SOPs.

For each category, track:

• Denial rate by volume and dollars
• % of denials overturned on appeal
  • If appeal success is high, that denial type is likely avoidable with better front-end workflows.
• Average time-to-resolution
  • Longer resolution cycles tie up staff and delay cash.

B. First-pass yield as a leading margin indicator

Measure:

• First-pass payment rate
  • % of claims paid in full (no adjustments beyond expected) on first submission
• First-pass “clean claim” rate
  • % of claims with no rework needed (even if small adjustments occur)
• Segment by:
  • Payer
  • Drug
  • Site-of-service
• Margin signal:
  • Low first-pass yield = margin leakage via delayed cash, staffing cost, and second-pass write-offs.

---

Step 5: Delays that convert into lost revenue

Delays don’t hit your P&L line item as obviously as denials, but they absolutely erode margin. Patients who never start therapy, or start weeks late, represent volume and revenue you never realize.

A. Referral-to-start timing

Track:

• Referral → benefits complete
• Referral → prior auth submitted
• Referral → prior auth approved
• Referral → first treatment scheduled
• Referral → first treatment completed

For each step, measure:

• Median and 90th percentile days
• Backlog at each stage (number of patients sitting in that step)
• Drop-off rate between steps (e.g., % of referred patients who never reach first treatment)

Margin signal:

• Long benefits/PA cycles and high drop-off between “approved” and “first treatment” are where revenue evaporates even though your team did much of the work.

B. Operational backlog indicators

Your backlog is a silent margin killer. Track:

• Queue size and age by workflow
  • Number of:
    • Unprocessed referrals
    • Unverified benefits
    • Pending PAs not yet submitted
    • Claims without status check beyond X days
  • Average and max age of items in each queue
• Staff minutes per task
  • Minutes per referral for intake
  • Minutes per benefit investigation
  • Minutes per prior auth submission
  • Minutes per claim status call/portal check

Watch for:

• Backlog creep (e.g., 4-day prescription backlog, 200–300 new prescriptions/day, 10–12 minutes per Rx) that forces you into reactive mode. That’s when patients slip, PAs get rushed, and denials spike.

---

Step 6: Claims statusing and appeals as margin recovery

Claims statusing and appeals are where you either recover margin that’s already leaking—or let it go.

A. Statusing cadence and coverage

Track:

• Time from submission → first status check
  • Especially for high-dollar infusions
• Status check frequency
  • Cadence of portal checks or payer calls until claim is paid or denied
• Coverage of statusing
  • % of high-value claims with documented status touches
• Margin signal:
  • Claims that sit too long without follow-up often end up hitting timely filing or appeal deadlines.

B. Appeal performance

For appeals, track:

• Appeal initiation rate
  • % of denied dollars that move to appeal (vs never touched)
• Appeal success rate
  • % of appealed dollars recovered
• Average appeal cycle time
  • Submission → final determination
• Margin signal:
  • High success on a specific denial type = clear opportunity to fix upstream workflows and prevent the denial in the first place.

---

Step 7: Connecting operational data to true margin leakage

All of these metrics matter, but they only become actionable when you tie them to dollar impact and staffing reality.

Build views that show:

• Margin leakage by workflow step

  • Benefits/OOP:
    • Bad debt tied to under-estimation
    • Write-offs tied to benefit misinterpretation
  • Medical policy/PA:
    • Denied dollars tied to missing or mismatched auths
  • Claims/billing:
    • Denied or underpaid dollars tied to coding/technical errors
  • Delays:
    • Lost revenue from patients who never start or drop off during long approval cycles

• Margin leakage by payer and drug

  • Identify:
    • Payers where benefit portals are especially error-prone
    • Drugs where PA requirements are complex and denials are frequent
  • This is where targeted automation or process redesign has the highest ROI.

• FTE-equivalent impact

  • How many minutes per Rx or per document are consumed by:
    • Portal navigation
    • Fax reading and data entry
    • Phone calls for status
  • Compare to throughput improvements when those steps are automated. For example, shifting from 20 minutes per document and 250 documents/day to ~3 minutes per document with under-2-hour turnaround is not just a speed win—it unlocks capacity to fix upstream issues that cause margin loss.

---

Where AI agents fit: executing the tracking, not just reporting on it

Most teams try to instrument these metrics by pulling reports from EHRs and billing systems and stitching them together manually. The problem: the actual work—and a lot of the leakage—happens in payer portals, fax queues, and phone calls, which traditional systems barely see.

This is where a platform like Mandolin is different:

• Workflow execution, not just dashboards

  • AI agents act like trained back-office specialists:
    • Reading faxes and referral forms (regardless of format)
    • Navigating payer portals for benefits and claims status
    • Making phone calls to payers
    • Compiling and submitting prior auths via portals, fax, and phone

• End-to-end visibility into margin drivers

  • Because every agent action is logged and traceable, you automatically get:
    • Actual time-in-step metrics (not estimates)
    • Reasons for delays (e.g., “waiting on lab result” vs “waiting on prior auth decision”)
    • Denial-associated context (what benefit data, which policy, which PA submission) without asking staff to double-document.

• Financially accurate OOP estimates

  • Mandolin calculates patient out-of-pocket using:
    • Real-time benefits
    • Site-specific fee schedules
    • Co-pay assistance
    • GPO and 340B pricing
    • Drug acquisition costs
  • That means your estimate-to-actual variance shrinks—and the associated bad debt leakage goes with it.

Our customers have seen:

• 24x increase in speed (20 minutes per document down to ~3 minutes with under-2-hour turnaround)
• Elimination of a 4-day prescription backlog to 0, while processing 200–300 new prescriptions/day
• The ability to scale to 4,500+ patients/month and refocus 13 outsourced roles onto higher-value work

Those are not abstract “efficiency” metrics—they are direct levers on buy-and-bill margin: fewer delays, fewer denials, more volume, and better OOP accuracy.

---

How to put this into practice in the next 30 days

If you want to understand where your buy-and-bill margin is really leaking, in a way you can act on:

1. Define your metric set using the categories above:

   • Benefits quality and timing
   • OOP estimate accuracy and assistance capture
   • Denial types and appeal outcomes
   • Referral-to-start timing and backlog age
   • Claims statusing coverage and cycle times

2. Connect metrics to dollars, not just volumes

   • Always ask: “What did this cost us in margin?”—bad debt, avoidable denials, or lost volume.

3. Instrument the messy middle

   • If your reporting doesn’t see payer portals, faxes, and phone calls, you’re missing where leaks actually start.
   • Consider AI agents that both execute the work and leave an auditable trail.

4. Use insights to redesign workflows

   • Focus first where:
     • Denial dollars are high and appeal success is strong (indicates preventable denials)
     • OOP variance and bad debt are concentrated
     • Backlogs are causing measurable referral-to-start delays

If you want to see how this looks when AI agents handle the intake, benefits, policy review, prior auth, and claims work—and you get the metrics and traceability “for free” from their logs—you can explore how Mandolin approaches it.

Get Started