How to architect revenue operations for a medical billing company in 2027

The outcome you should expect

When the architecture is right, a medical billing company stops describing itself by throughput and starts describing itself by yield. That shift is not cosmetic. A company that reports "we submitted 180,000 claims last quarter" has no way to know whether it is compounding or quietly bleeding, because claim count moves with client mix, seasonality, and specialty acuity — none of which map cleanly to revenue. A company that reports "net collection rate across billings under management held at the high nineties while days in AR fell by nine days" is describing the two numbers that actually determine its own P&L.

The concrete outcome is a small set of durable capabilities. First, you can quote a per-client net collection rate on demand, by specialty and by payer, without a spreadsheet reconciliation project. Second, you can see the charge-to-cash cycle as a measured pipeline with stage-level dwell times, rather than as a vague sense that "AR feels heavy this month." Third, denials stop being a queue that a few specialists grind through and become a feedback signal that changes upstream scrubbing rules. Fourth, client reporting becomes an automatic byproduct of the operating system rather than a monthly manual scramble — which matters more than most owners expect, because provider clients churn on opacity as often as they churn on results.

There is a second-order outcome worth naming: pricing power. A billing company that can demonstrate, with its own instrumented data, that it lifted a cardiology group's net collection rate by several points and cut days in AR meaningfully has an evidence-backed case for its fee percentage. A company that can only show claim counts is negotiating on price against offshore competitors who will always be cheaper per claim. The architecture is, in a real sense, the sales asset.

Expect the transformation to be uneven. Instrumenting charge capture and submission is usually straightforward because the RCM platform already holds that data. Instrumenting denial reasons at a granularity useful for rule-writing is harder, because denial codes arrive in ERA and claim-status transactions that many companies never parse programmatically. Instrumenting client-level profitability — revenue earned per client against the labor hours that client consumes — is hardest of all, and it is the one most owners skip. Skipping it is how a billing company ends up with a flagship account that generates impressive top-line fees and negative contribution margin because its payer mix and documentation quality require triple the follow-up labor of a comparable account.

What drives that outcome

Four variables drive medical billing economics, and the architecture exists to make each of them observable and controllable. The revenue identity is straightforward: billings under management multiplied by net collection rate multiplied by fee percentage. Profit is then governed by clean-claim rate, denial recovery, and cost to serve. Everything in the stack should ladder to one of those terms.

Billings under management is the volume base. It grows two ways — adding provider clients, and expanding charges within existing clients as they add providers, service lines, or locations. The second path is chronically underexploited. A practice that adds a nurse practitioner or opens a second location increases charges without any new business development cost to you, but only if someone is watching for it. Architecturally that means client-level charge volume needs a trend view and an alerting threshold, not just a monthly report nobody opens.

Net collection rate is collections against collectible charges — the honest denominator, excluding contractual adjustments you were never going to collect. This is the number that justifies your existence. A provider comparing you against in-house billing is implicitly asking whether you recover more of what is legitimately owed. If you cannot answer with a per-client figure, you are asking them to take it on faith.

Fee percentage is set at contract time and is the one variable most owners treat as fixed. It should not be. Fee should vary with payer mix complexity, specialty, average charge size, and the services bundled — coding, credentialing, patient collections, prior authorization support. A flat percentage across a book that includes both a straightforward primary-care practice and a behavioral health group with heavy Medicaid exposure and prior-auth burden is a pricing error dressed up as simplicity.

Cost to serve is the term that quietly decides whether growth is good. It is driven by claims per FTE, rework rate, and the amount of manual data chasing required per client. A client whose documentation arrives clean and structured costs a fraction of one whose charges arrive as scanned superbills.

Two loops in that diagram matter more than the linear path. The denial loop — denial to pattern analysis to rule update to scrubbing — is what converts a cost center into a compounding asset. Without it, the same denial reason recurs indefinitely and your specialists rework the same error forever. The reporting loop back to business development is what turns operational performance into client acquisition, because your best sales material is a demonstrated collection lift on a comparable practice in the same specialty.

A note on the adjacent case, because the pattern generalizes: this is structurally the same architecture a dental billing service, a veterinary billing service, or a durable medical equipment biller needs. The payer landscape differs, the codes differ, and the prior-auth burden differs, but the identity — volume under management times yield times fee — is identical. If you build the instrumentation generically rather than hard-coding it to one specialty's quirks, you buy yourself optionality to expand into adjacent verticals without re-architecting.

Benchmarks and realistic ranges

Be careful with benchmarks in this industry, because the published figures vary enormously by specialty, payer mix, and how the denominator is defined. Treat what follows as directional framing for your own baselining, not as targets to import wholesale.

Net collection rate. Well-run operations generally target the mid-to-high nineties against collectible charges. The gap between a strong and weak operation is usually only a few percentage points — but on a large book those points are the entire margin. The critical discipline is defining "collectible" consistently. Companies that quietly exclude aged balances from the denominator can report impressive rates while collecting less cash. Define it once, document it, and report it the same way every month even when the number is unflattering.

First-pass clean-claim rate. This is the percentage of claims paid without any rework. It is the single most leveraged operational metric because rework cost is roughly linear in rejections while revenue is not. A meaningful improvement here drops straight to contribution margin. Track it by payer and by provider, because an aggregate number hides the fact that one payer or one physician's documentation habits are generating a disproportionate share of the rework.

Days in accounts receivable. Report it with aging buckets — current, 31 to 60, 61 to 90, 91 to 120, and over 120 — not as a single average. The average is misleading because a long tail of aged claims that will never be collected drags it upward while telling you nothing actionable. The over-120 bucket deserves a standing weekly review; balances there are usually either recoverable with a specific documented action or should be written off so the metric stops lying.

Denial rate and recovery rate. Two separate numbers that get conflated. Denial rate is what fraction of claims come back unpaid on first submission; recovery rate is what fraction of those you eventually collect. A company can have a mediocre denial rate and excellent economics if recovery is strong and fast, or a decent denial rate and poor economics if denials sit unworked past appeal deadlines. Timely-filing and appeal windows are hard deadlines — a denial that ages past its appeal window converts from a recoverable receivable into a permanent write-off, and that transition should trigger an alert well before it happens.

Onboarding time to first clean claim. The industry commonly runs multiple weeks for a small-to-mid practice, longer for larger groups with complex payer contracts. Every day of delay is charge volume you never earn a fee on. The realistic improvement path is to compress the sequence — data collection, credentialing and payer enrollment verification, fee schedule loading, and test claim submission — by making each stage a tracked pipeline with an owner rather than an informal handoff. Credentialing is usually the long pole and is largely outside your control; the fix is to start it first and in parallel, not to try to accelerate it.

Cost to serve. Measure claims processed per FTE and revenue per FTE, segmented by client. Then build a simple contribution view: fee revenue per client minus loaded labor hours attributed to that client. Do this even roughly. The distribution will surprise you, and it is the input to both repricing conversations and the occasional decision to resign an account.

Client retention. Retention is the compounding variable. A billing company with strong retention grows on a rising base; one with churn runs a treadmill where new logos replace lost ones. Instrument it as gross revenue retention and net revenue retention separately, so expansion within existing accounts is visible as its own motion.

One caution on benchmarking against published industry averages: specialty mix dominates. A practice with a heavy Medicare book and clean documentation will produce metrics that look nothing like an out-of-network behavioral health group. Segment your own benchmarks by specialty and payer mix before comparing yourself to anything external, or you will chase a number that was never achievable given your book.

Risks, edge cases, and failure modes

Counting claims instead of collections. The most common architectural failure. Submission volume is easy to measure and feels productive, which is exactly why it becomes the default dashboard metric. It is nearly uncorrelated with the health of the business. If the executive dashboard leads with claim counts, the organization will optimize claim counts.

Silent underpayments. Explicit denials get worked because they show up in a queue. Underpayments — where a payer pays, just less than the contracted rate — often do not, because nothing flags them. Catching these requires comparing each remittance line against the provider's contracted fee schedule and flagging variances beyond a threshold. This is real work: it means loading and maintaining fee schedules per client per payer, keeping them current as contracts renew, and handling modifier logic. But an unwatched variance is indistinguishable from correct payment, and it recurs on every claim under that contract.

Aged denials past appeal deadlines. Payer appeal windows are finite. A denial that sits unworked past its window is not a slow collection, it is a permanent loss — and it is a loss your client will eventually notice. Build the deadline into the work queue's prioritization, not into a specialist's memory.

Credentialing gaps. A provider who is not enrolled with a payer generates claims that reject in bulk. This surfaces during onboarding if you check, and surfaces as a crisis if you do not. It also recurs: enrollments lapse, providers move between locations and tax IDs, and group NPIs change. A periodic re-verification sweep prevents a slow accumulation of rejections nobody traced to a root cause.

Client concentration. A billing company where one large group represents a large share of revenue is one contract renewal away from a crisis. This risk is amplified because large groups are exactly the clients most likely to eventually bring billing in-house once they reach scale. Track concentration explicitly and factor it into how aggressively you pursue mid-size accounts.

Payer mix shifts inside an existing client. A practice that adds a Medicaid-heavy service line, joins a Medicare Advantage network, or opens to a different population changes the economics of your contract without renegotiating it. Your yield drops, your labor rises, and your fee percentage stays put. Monitor per-client payer mix as a trend, and treat material shifts as a trigger for a pricing conversation rather than a slow margin bleed you discover at year-end.

Compliance and data handling. This is protected health information. Every integration, every export, every reporting tool, and every vendor in the chain sits inside your HIPAA obligations, and business associate agreements need to cover each one. The failure mode is not usually a dramatic breach — it is a well-meaning analyst exporting claim data into an unapproved spreadsheet tool to build a report the system should have produced natively. Good reporting architecture is a compliance control, not just an operational convenience.

Automation applied to the wrong layer. Automating claim submission and scrubbing is high-value and low-risk. Automating coding decisions or appeal narratives without human review is where companies get into trouble, both clinically and from an audit standpoint. Draw the line explicitly: automate the mechanical and the detective work, keep judgment human, and log what the system recommended versus what a person approved.

Over-buying tooling. The temptation to build or buy a bespoke platform is strong, and it is usually the wrong call. Your product is collection performance, not software. An established RCM platform paired with a clearinghouse covers the overwhelming majority of what the architecture needs; the differentiating work is in the analytics layer, the denial feedback loop, and the operating discipline around them.

A practical rollout plan

Sequence this so each phase produces something usable before the next begins. The failure mode of revenue architecture projects is a six-month build that delivers nothing until it delivers everything.

Phase one — define the metrics and their denominators. Before touching a system, write down exactly how you compute net collection rate, clean-claim rate, days in AR, denial rate, and denial recovery rate. Specify the denominators. Get the operations lead and the finance lead to agree in writing. This sounds bureaucratic and it is the single highest-leverage step, because every downstream report inherits these definitions and re-defining them later invalidates your entire trend history.

Phase two — establish the source of truth. Confirm that the practice-management/RCM platform holds claims, payments, and denials authoritatively, and that no material data lives only in spreadsheets or email. Where it does, move it or accept that your reporting will be wrong. Verify that ERA and claim-status transactions are being ingested programmatically, not just eyeballed, because denial reason codes are the raw material for the feedback loop.

Phase three — build the reporting layer. Produce a per-client view showing net collection rate, days in AR with aging buckets, clean-claim rate, and denial rate — refreshed on a fixed cadence, from the source of truth, with no manual assembly. Ship it internally first and let operations attack the numbers for a few weeks before showing a client. The first version will be wrong somewhere; better to find that yourselves.

Phase four — close the denial loop. Start manual. Have someone review denial reason codes weekly, identify the top recurring patterns by payer, and propose scrubbing rule changes. Deploy the rules and measure whether that denial reason's frequency drops. Only automate the pattern detection once you have proven the human version produces rule changes that work — automating a loop you have not validated just generates noise faster.

Phase five — instrument onboarding. Turn onboarding into a tracked pipeline with explicit stages and a named owner, and measure time to first clean claim as a standing metric. Start credentialing verification first and in parallel with everything else, since it is the constraint. Add a test-claim gate before going live so configuration errors surface on ten claims instead of a thousand.

Phase six — add client-level profitability. Attribute labor to clients, even approximately, and build the contribution view. Use it for repricing, for deciding which segments to pursue, and occasionally for resigning an account. This phase is last because it depends on the earlier instrumentation and because it is politically the hardest — it produces conclusions people would rather not have.

Two staffing notes. Give the reporting layer an owner who is neither a biller nor a pure analyst — someone accountable for the numbers being right and used. And resist assigning denial pattern analysis to the denial specialists themselves; they are optimized for clearing the queue, and pattern work always loses to queue work when the same person owns both.

Related questions

Should a medical billing company charge a percentage of collections or a flat fee?

Percentage of collections aligns your incentive with the provider's — you earn more when they collect more. Flat fees can suit very simple, high-volume, low-variance billing, but they decouple your revenue from collection quality, which is the value you actually provide.

What is the difference between gross and net collection rate?

Gross collection rate compares collections to total charges, which is misleading because charges include contractual adjustments you were never entitled to. Net collection rate compares collections to collectible charges after those adjustments — it is the honest measure of recovery performance.

How do you grow billings under management without adding headcount proportionally?

Improve clean-claim rate so fewer claims need rework, and target clients whose documentation arrives structured rather than as scanned superbills. Growth in claims per FTE comes from reducing rework and manual data chasing, not from working faster.

Does this architecture apply to dental or veterinary billing services?

Largely yes. Codes, payers, and prior-auth burdens differ, but the revenue identity — volume under management times yield times fee — and the denial feedback loop are identical. Build the instrumentation generically rather than hard-coded to one specialty's quirks.

When should a billing company resign a client?

When client-level contribution is persistently negative and repricing has been refused. That decision requires a labor-attributed profitability view; without one, unprofitable accounts hide inside a healthy-looking top line indefinitely.

FAQ

What is the most important metric for a medical billing company?

Net collection rate on billings under management. It measures how much of the provider's legitimately collectible revenue you actually recover, and since your fee is a percentage of collections, it drives your own revenue directly. Claim volume is a workload measure, not a performance measure.

Do I need to build custom software to run revenue operations?

No. Established practice-management and RCM platforms paired with a clearinghouse cover the transactional layer. Your differentiation is collection performance, not software. Build only the analytics layer and the denial feedback loop on top — that is where the leverage is and where off-the-shelf tools tend to be weakest.

How do I reduce days in accounts receivable?

Attack it at the source rather than the tail. Faster charge capture, better claim scrubbing before submission, and denial work queues prioritized by appeal deadline all shorten the cycle. Report aging buckets rather than a single average, since the average hides the aged tail that actually needs decisions.

How should denials be handled structurally?

Two separate motions. Recovery — working the denial before its appeal deadline — and prevention — analyzing denial reason codes for patterns and updating scrubbing rules upstream. Assign them to different people, because queue-clearing work always crowds out pattern work when one person owns both.

What causes onboarding delays, and what can actually be fixed?

Credentialing and payer enrollment are usually the constraint and are largely outside your control, so start them first and in parallel. What you can compress is document collection, fee schedule loading, and configuration validation — turn each into a tracked pipeline stage with an owner rather than an informal handoff.

How do I know whether a client is actually profitable?

Attribute labor hours to clients and compare against fee revenue earned. Rough attribution is far better than none. Payer mix, documentation quality, and prior-auth burden vary enormously between accounts, so two clients with identical billings can have opposite contribution margins.

Sources

• https://www.hfma.org/
• https://www.mgma.com/
• https://www.aapc.com/
• https://www.ahima.org/
• https://www.cms.gov/
• https://www.availity.com/
• https://www.hhs.gov/hipaa/index.html
• https://www.ama-assn.org/

Related on PULSE

• [How do you architect revenue operations for a subscription billing company in 2027?](/knowledge/ra0377)
• [Revenue Architecture for Medical Device + SaMD Software in 2027 (Regulatory Outcomes, Reimbursement)](/knowledge/ra0142)
• [Lead-to-Cash Architecture Blueprint: From MQL to Recognized Revenue Across CRM, CPQ, and Billing](/knowledge/ra0607)
• [Revenue Architecture for Professional Services Firms: Project-Based Billing and Retainer Models](/knowledge/ra0518)
• [CPQ to Billing Handoff Architecture in 2027](/knowledge/ra0476)
• [Designing Revenue Operations for Subscription Box Companies: Churn Prevention and Inventory-Linked Billing](/knowledge/ra0530)