How to Choose an ABA Billing Service: 8 Real Criteria

Every ABA practice eventually hits a point where billing is either eating revenue or eating clinical leadership time. The choice that follows is not binary. Four real operating models compete for the work: an in-house full-time biller, an outsourced full-service revenue cycle management partner, an A/R-only hybrid that keeps front-end billing in house, or a DIY-with-software approach where the owner runs claims themselves. The right choice depends on practice size, payer mix, owner time, and the bench depth a practice can sustain. This guide walks through eight criteria that separate billing services that actually work for ABA practices from those that look fine on paper. Vendor selection is one piece of a larger operational picture, and the ABA Practice Operations Guide covers how billing fits with the other eight operational disciplines that keep a practice sustainable.

TL;DR

• Industry pricing runs 3 to 8% of collected revenue. Most ABA billing services do not publish rates publicly. A vendor that refuses to put a number on the table before a sales call is a meaningful signal.
• ABA-specific coding fluency is the single biggest predictor of clean claims. Generic medical billing teams will lose CO-97 bundling appeals and CO-197 auth disputes that an ABA-native team would win.
• Honest appeal win rates land at 70 to 85% on legitimately appealable claims. Vendors quoting 95%+ without specifying which denial categories they exclude are doing math on a curated subset.
• Real-time dashboard access separates accountable services from statement-based ones. If your service cannot show first-pass rate, denial rate, days in A/R, aged A/R over 90 days, and net collection rate on demand, those numbers are not being tracked.
• Onboarding is faster than the industry says. ABA-native services with disciplined playbooks can clean over in 2 to 3 weeks if the practice is already enrolled. The 6-to-12-week industry default is usually a coordination problem, not a technical one.

The Four Models You're Choosing Between

Before evaluating vendors, the practice has to decide which operating model fits. Each carries a different cost structure and risk profile.

In-house full-time biller. A salaried billing specialist (or two) inside the practice handles eligibility, claims, denials, and posting. Total cost runs roughly $65,000 to $80,000 in salary plus benefits and software, or 13 to 16% of revenue at a $500,000 practice. Works at scale ($2 million+) where the volume keeps a billing team fully utilized. Fails on bench depth: a single biller who leaves takes the institutional knowledge with them.

Full-service outsourced RCM. A specialized partner handles the end-to-end claim lifecycle, priced at 5 to 8% of collected revenue per HFMA and MGMA benchmarks. Works for $500,000 to $2 million practices where the practice cannot keep one biller occupied but the billing complexity exceeds DIY tolerance. Risk: vendor selection is the entire game, because a generic partner without ABA fluency will leak revenue at every CO-97 and CO-197 they touch.

A/R-only hybrid. The practice keeps front-end work (eligibility, authorization, claim submission) in house and outsources back-end work (denial work, A/R follow-up, payment posting reconciliation). Priced at 3 to 5% of collected revenue. Works when the practice has front-end discipline and wants accountability on the back-end without giving up control of the patient-facing workflow. The handoff between in-house front-end and outsourced back-end is the failure mode if it is not tightly governed.

DIY-with-software. Owner or admin runs billing through a practice management system. All-in cost is software ($2,000 to $5,000 per year) plus owner time, which depending on how you value the time runs $3,000 to $50,000 effectively. Works under $500,000 if the owner has the discipline. Stops working as the practice grows past 4 staff because billing expands faster than the owner can absorb.

For deeper analysis of the make-vs-buy decision itself, see the companion piece on ABA billing: in-house versus outsourced.

Criterion 1: Real Pricing (and Why Most Services Won't Publish It)

The first signal is whether a vendor publishes a rate on their public website. The honest industry range, per HFMA revenue-cycle benchmarks and MGMA DataDive performance benchmarks for behavioral health and per the CAQH Index on revenue cycle costs, runs 3 to 5% of collected revenue for A/R management engagements and 5 to 8% for full-service RCM. A few specialty firms quote per-appeal or per-claim flat fees, but the dominant pricing model in ABA is percentage of collected revenue.

Public publication of rates is the exception in this market, not the default. As of May 2026, public-website price audits of the named ABA billing competitors return the following:

• Cube Therapy Billing does not publish rates publicly.
• Missing Piece does not publish rates publicly.
• AnnexMed does not publish rates publicly.
• Plutus Health does not publish rates publicly.
• MBW RCM does not publish rates publicly.
• Operant Billing does not publish rates publicly.
• VG Soft Co (RCM) publishes 3% (A/R management) or 6% (full-service) on the public RCM pricing page.

Non-publication is not by itself a quality signal. Custom pricing can reflect tailored arrangements for different practice profiles, volume tiers, or contract terms, and most ABA billing services operate this way for legitimate reasons. The named competitors above each bring real differentiators worth weighing alongside the rate question. Cube Therapy operates with 16 years of tenure, 650+ providers, and a substantial content footprint that may matter for practices billing ABA alongside speech or occupational therapy. Operant Billing publishes a 98% first-pass clean-claim rate and 18-day A/R with SOC 2 certification. Plutus Health has published a case study documenting a 41.6% denial reduction in 50 days and a 33.2% A/R reduction at one engagement. Missing Piece holds BBB accreditation and 15+ years of ABA-specific experience. The practical implication of non-publication is not that rates are higher, but that a buyer evaluating five vendors has to extract pricing through five separate sales conversations, which is a real cost of evaluation.

The diagnostic question to ask any vendor: "What is your stated rate for a practice my size, and is it consistent across your customer base?" A vendor whose rate moves significantly based on negotiation skill is one whose rate is not anchored in their cost structure, though this is a signal worth probing, not an automatic disqualifier.

Criterion 2: ABA-Specific Coding Expertise

ABA billing is not generic medical billing with a different CPT code set. Three structural features of ABA claims make ABA-specific fluency non-optional.

The 97153 and 97155 bundling rules vary by payer and by year. Whether protocol modification (97155) can be billed concurrently with direct treatment (97153) on the same date is a payer-specific policy that shifts annually. The underlying code descriptors are maintained by the AMA CPT system and bundling guidance flows from the Medicare National Correct Coding Initiative. A generic biller will appeal a CO-97 bundling denial the same way they would appeal a primary-care bundling denial, which usually loses. An ABA-native biller knows which payers permit concurrent billing under which conditions and how to document the protocol-modification justification.

Modifier rules (HM, HN, HO, HP, AF, AH) are highly specific to behavioral health. Missing or wrong modifiers drive a large share of CO-16 denials in ABA. Generic billers often submit claims without ABA-specific modifier validation, which leads to clean-claim rates 10 to 15 percentage points below what an ABA-native team produces.

The 8-minute rule applied to ABA is its own gotcha. Direct ABA service codes (97153, 97155) are time-based and follow Medicare's 8-minute rule for partial-unit billing. ABA practices often submit time documentation that rounds against the rule, and generic billers do not catch it. The denial rate consequence is small per claim and large in aggregate.

Supervision-link denials are uniquely ABA. Sessions billed by an RBT under the supervision of a BCBA require credentialed supervision linkages on the claim. When the supervision link is broken (BCBA credentialing expires, supervision documentation is not on file, or the supervisor changes mid-cycle), the entire session can deny. The supervision requirements themselves are defined in the BACB supervision and training standards and the BACB ethics code (sections 4.0 and 5.0). Experienced generic billers may track supervision linkages under broader credentialing audits, but ABA-native services typically treat it as a daily-cadence check rather than a periodic one.

The diagnostic question: "What is your team's bundling-error rate on 97155 versus 97153 same-date claims?" A vendor that cannot produce a number is either not measuring it or aggregating it under a different category name. The point is to get a specific operational answer, not to disqualify a vendor for lacking a single metric on the spot. The follow-up worth asking: "How does your team appeal a CO-97 bundling denial on 97155 specifically?" An ABA-fluent answer will reference payer policy, documentation requirements, and which payers permit concurrent billing. See the ABA billing codes 2026 reference for the full code-level detail.

Criterion 3: Authorization Lifecycle Management

Authorization-related denials (CO-197 and its variants) are the single largest denial category in ABA, accounting for roughly 20 to 30% of preventable claim losses according to the X12 Remittance Advice Remark Codes reference (the standards body that maintains the CARC and RARC code sets adopted by CMS and commercial payers) and industry surveys. Authorization lifecycle management is the operational discipline that closes this gap.

A real authorization lifecycle has four components:

Expiration tracking with lead-time alerts. Every authorization in the practice has an end date, an authorized unit count, and a renewal cadence specific to the payer. A working system alerts the team 14 days before expiration, not the day after the first denial.

Retro-authorization handling. When a session is rendered before the auth is in place, retro-auth requests have to be filed within payer-specific windows (often 30 to 90 days). A vendor that cannot articulate retro-auth workflow is one that loses these dollars by default.

Payer-specific renewal cadence. Medicaid renewals follow state-specific cycles; commercial renewals are payer-by-payer. ABA-native vendors typically maintain a payer-by-payer renewal calendar; generic vendors more often default to a monthly review cadence that can miss the tighter payer-specific cycles.

Authorization-unit reconciliation. Sessions billed against an exhausted unit count deny under CO-197 regardless of clinical appropriateness. The unit-count reconciliation has to happen before submission, not after the denial.

The diagnostic question: "How many days before expiration does your team flag an authorization, and what happens if a unit count is exhausted mid-session-block?" A vendor that can articulate the process in operational terms (who gets alerted, how the alert escalates, what the team does on receiving it) can be tested against their last 90 days of CO-197 denial volume. A vendor that cannot articulate it probably does not have a defined process yet.

For the in-product detail on how this is operationalized in software, see the authorization management feature page.

Criterion 4: Appeal Win Rates by Denial Category

Vendors will often quote a single appeal win rate ("our appeal win rate is 92%") without specifying which denials are included in the math. The honest framing is per-category, because some denial categories are nearly always recoverable and others are nearly always permanent.

The realistic per-category win rates on legitimately appealable ABA claims, drawn from a 10-code reference set compiled at the VG Soft Co denial management deep-dive:

• CO-16 (claim missing data, often modifiers): 90%+ recoverable. These are correction-and-resubmit, not true appeals.
• CO-197 (authorization absent or expired): 60 to 80% recoverable if retro-auth is filed within payer windows.
• CO-97 (bundling, often 97155 versus 97153): 50 to 75% recoverable with payer-policy-specific appeal language.
• CO-50 (not medically necessary): 40 to 70% recoverable with BCBA-signed medical-necessity letter and session-level documentation.
• CO-29 (timely filing expired): 5 to 20% recoverable, and only on documentable submission-system failures.

Aggregate appeal win rate across all categories on legitimately appealable claims should land at 70 to 85%. The remaining 15 to 30% are structural denials (timely-filing past the appeal window, hard plan exclusions, services rendered without retroactively obtainable auth) that are not realistically recoverable.

The diagnostic question: "Can you produce a per-category appeal win rate report, by reason code, for the last 12 months across your ABA book?" Vendors that report only an aggregate number are either not tracking by category or are excluding the hard categories from the math.

Criterion 5: Real-Time Visibility vs Statement-Based Reporting

The most consequential operational difference between billing services is how much of the work the practice can see in real time. Two models exist, and both are legitimate operating choices.

Statement-based reporting. The vendor handles billing and reports back monthly with summary statements. The practice sees collections, denials, and aged A/R as aggregate numbers. Individual claim status is opaque unless the practice escalates. The implicit promise is "we handle it, you do not worry about it."

Real-time dashboard. The vendor handles billing while exposing the underlying workflow in a shared dashboard. The practice sees, in real time, claim-by-claim status, denial reasons by code, days in A/R by payer, aged A/R buckets, and the same operational KPIs the vendor's team works against. The implicit promise is "you do not have to do the work, but you can see all of it."

Statement-based services optimize for the absence of practice involvement. Real-time-dashboard services optimize for the practice's confidence in the work. Both serve real needs. A high-trust long-term partnership with a vendor that has earned its track record can run fine on monthly statements. A practice rebuilding from a prior bad outsourcing experience usually needs the real-time view to recover that trust. The question is which model serves the practice better at the practice's current stage.

Three practical signals separate the two:

• A live dashboard the practice can log into independently, not a monthly emailed PDF.
• Claim-by-claim audit trail with timestamps, who touched the claim, and what changed.
• Real-time KPI visibility on first-pass rate, denial rate, days in A/R, aged A/R over 90 days, and net collection rate.

If a vendor's answer to "can I see the dashboard" is "we generate a monthly report," the model is statement-based regardless of how it's described. That is a legitimate operating choice, but one the practice should make consciously rather than discover after signing.

Criterion 6: Contract Terms That Actually Protect You

Contract language is where vendor selection often goes wrong, because the contract terms become visible only when the practice is already locked in and trying to leave. Five clauses to inspect before signing.

Notice period. Industry-standard is 30 days. Anything longer (60 or 90 days) is a switching-cost bomb and signals a vendor optimizing for retention rather than performance.

Data ownership and export. The practice owns its claim file, not the vendor. The contract should specify that the practice can export the full claim file in standard 837/835 formats at any time, without charge, with no per-claim fee. A contract that does not specify this, or that charges per-claim for export, is a hostage clause.

Pricing structure. Percentage of collected revenue is the most aligned model: the vendor only gets paid when the practice gets paid. Flat monthly fees are misaligned because the vendor gets the same payment whether collections are $50,000 or $500,000 in a given month. Per-claim fees can work for very high-volume practices but require careful modeling.

Term and renewal. Month-to-month is the most flexible. Annual contracts can work if they include performance-out clauses (the practice can exit without penalty if KPIs fall below stated thresholds). Annual contracts without performance-outs are pure lock-in.

Platform fees and surcharges. Some vendors charge a separate platform fee, software fee, or clearinghouse pass-through on top of the percentage. The all-in cost calculation has to include these. A 5% headline rate with a $400 monthly platform fee is materially different from a 5% all-in rate.

The diagnostic question: "Can I see a sample contract before our discovery call ends?" Vendors who decline to share contract language until after a verbal commitment may have reasons (legal review, customized terms per practice), but the practical effect is that the terms become visible only after lock-in is harder to reverse. A vendor willing to share standard contract language up front makes evaluation faster and signals less concern about how the terms read in isolation.

Criterion 7: Onboarding Time and the First-90-Days Reality

Industry default onboarding for an outsourced billing partner is 6 to 12 weeks. The practical floor for an ABA-native vendor with a disciplined playbook is 2 to 3 weeks. The gap between the two is almost always coordination, not technology.

The activities that consume onboarding time:

• Payer enrollment confirmation (1 to 2 weeks). The new vendor has to verify the practice's enrollment status with every payer in the panel. If NPI or tax ID changes are required, this stretches to 4+ weeks per payer.
• Clearinghouse setup (3 to 5 days). The vendor's clearinghouse account has to be connected to the practice's payer panel.
• Credentialing transfer (variable, 2 to 8 weeks). If credentialed providers are changing or if the vendor needs delegated credentialing access through CAQH ProView, this is the slowest step.
• System integration and data migration (1 to 2 weeks). Patient records, authorization records, and claim history have to be loaded into the vendor's system. ABA-native vendors with prebuilt connectors run this in days; generic vendors run it in weeks.
• Parallel run (2 weeks recommended). The new vendor processes claims alongside the old service to catch coordination errors before the full cutover.

The diagnostic question: "What is your median onboarding time across the last 10 ABA practices, and what is the longest?" Median, not best case. The variance between median and longest indicates how often onboarding goes sideways.

The first 90 days after cutover are where the relationship is actually tested. Three metrics to watch by Day 90:

• First-pass clean-claim rate on new claims submitted under the new service: target 90%+ by Day 30, 95%+ by Day 90.
• Days in A/R trend: the absolute number takes 120 days to stabilize, but the trend should be flat or improving by Day 60.
• Communication cadence: weekly operational reviews for the first 90 days, monthly thereafter. Vendors who drop to monthly cadence by Week 4 are abandoning the relationship early.

Criterion 8: When In-House Still Wins

Outsourcing is not the right choice for every practice. Three profiles favor keeping billing in house.

Solo BCBAs with simple payer mixes. A solo practice with one or two commercial payers, no Medicaid, and 25 billable hours per week can run DIY-with-software at 1 to 3% all-in cost, well below the 5 to 8% full-service outsourcing band. The owner-time cost is real (5 to 10 hours per week) but at low billing volume the math favors absorbing it. See the BCBA income potential calculator for the underlying solo-practice math.

Established practices over $2 million with built billing teams. A practice that has already invested in a billing manager, two specialists, and operational discipline often outperforms outsourcing economically. At $2 million in collections, 5% full-service outsourcing is $100,000 per year, which buys roughly 1.3 billing FTEs in house. If the existing team produces clean claims at 95%+ and runs aged A/R under 8%, the outsourcing math does not improve on it.

Specialty Medicaid markets. Some state Medicaid programs have such specific rule sets (Texas, Indiana, North Carolina, and a few others have major 2026 program changes per the state-of-ABA pillar) that the cost of teaching an outside team the state-specific knowledge exceeds the salary cost of an internal expert who already has it. This is a narrow case but a real one.

For the conceptual framework around make-vs-buy, see ABA billing: in-house versus outsourced and the in-house versus outsourced denial guide.

One more practice profile worth naming: for practices billing across multiple disciplines (ABA combined with speech therapy, occupational therapy, or mental health) or operating at enterprise scale across 10+ locations, a multi-specialty billing partner with breadth (Cube Therapy and similar vendors fit this profile) may genuinely outperform an ABA-only specialist. The operational savings of a single vendor across all service lines can exceed the per-claim accuracy advantage of ABA specialization. The ABA-only argument is strongest when ABA is the practice's exclusive or near-exclusive service line.

The Worked Example: $500K Practice Across All Four Options

The clearest way to see how the four models compare is to run the same practice through each. Assume a $500,000 collected revenue practice with one BCBA owner, two RBTs, and a mixed commercial-plus-Medicaid payer panel.

Two caveats on the math.

The DIY number depends on how owner time is valued. If the owner does not value billing time (treating it as part of the role rather than as opportunity cost), DIY drops to roughly $3,000 per year (software only). If the owner values their time at $150 per clinical hour, the cost is $25,000. Both are real ways to count, and the practice has to decide which applies.

The math does not include the collection-rate delta. Industry benchmarks suggest full-service RCM with an ABA-native team produces 94 to 96% net collection rates versus DIY or generic-vendor rates of 71 to 88%. A 10-percentage-point collection improvement on $500,000 in billable revenue is $50,000 per year, which dominates the cost-side comparison. The honest framing: the cost difference between models is meaningful, but the collection-rate difference between models is often larger.

For the full tier breakdown with deeper line-item analysis, see the RCM pricing page.

What VG Soft Co Built

VG Soft Co provides ABA-specialized outsourced billing with Glass Box visibility, 95%+ collection rates, and 3% or 6% of collected revenue pricing. Tier 1 (A/R management) is 3% and covers denial work, payment posting, A/R follow-up, and operational reporting. Tier 2 (full-service RCM) is 6% and covers the end-to-end claim lifecycle from eligibility to net collection. Both tiers run on a shared real-time dashboard the practice can log into directly, with the same five operational KPIs the RCM team works against (first-pass rate, denial rate, days in A/R, aged A/R over 90 days, net collection rate).

The team is ABA-native, which is the determining factor on every criterion in this guide. Bundling rules, modifier specificity, authorization lifecycle, supervision-link audits, and state-Medicaid variance are the differences between a clean-claim rate above 95% and one below 85%, and a generic billing team does not catch them.

Pricing is published. Contracts are month-to-month. Data is portable in standard 837/835 formats with no per-claim export fees. The denial management sub-page walks through the 10 ABA denial codes and per-category appeal win rates in deeper detail. The RCM pricing page shows the full Tier 1 versus Tier 2 inclusion table and the $500K worked example with line-item math.

Closing Decision Framework

The eight criteria roll up into a one-page decision sketch.

• If pricing is unpublished, ask why before the second call.
• If the vendor cannot articulate their bundling-error rate on 97155 versus 97153, they are not measuring it.
• If authorization lifecycle is not a daily-discipline answer, auth-related denial volume will not be controlled.
• If appeal win rates come as a single aggregate number, the math is hiding a structural weakness.
• If you cannot log into a live dashboard with claim-by-claim status, the model is statement-based, which is a legitimate choice but worth making consciously.
• If the contract has a 90-day notice clause or charges for claim export, the lock-in is structural.
• If onboarding median exceeds 4 weeks for ABA-native practices, the vendor's playbook is generic.
• If your practice profile fits in-house (solo with simple payer mix, $2M+ with built team, specialty Medicaid expertise), the right answer may be to not outsource.

The buyer's job is to compress these eight checks into one or two diagnostic conversations and to verify the answers against references. The vendor's job is to make the answers easy to find before the contract gets signed.