.png)
Frameworks, core principles and top case studies for SaaS pricing, learnt and refined over 28+ years of SaaS-monetization experience.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
Which Pricing Metric Fits Radiology Groups SaaS Best per Seat per Transaction or per Outcome
Radiology software has an unusually difficult pricing problem. The human user is obvious, but the value is not created by the mere presence of that user. A radiologist can interpret hundreds or thousands of studies in a month, a PACS can serve many more, and workflow software can become more valuable as a practice grows without adding people at the same rate. US Medicare economics reinforce that pattern: radiology services are paid through service-specific fee-schedule mechanics built around relative value units and individual services, rather than a fixed amount per radiologist.
Cloud delivery makes the mismatch harder to ignore. As of 12 August 2026, Sirona Medical sells its unified radiology workflow platform on a per-exam subscription basis, OmniPACS publishes per-study rates, and Candelis describes its cloud PACS as priced per study. Meanwhile, products such as RadPal and PostDICOM demonstrate that seat and subscription pricing remain viable for narrower tools or smaller installations.
Monetizely's position is clear: per transaction should be the primary pricing metric for SaaS sold to radiology groups. More specifically, vendors should charge against completed diagnostic studies, using committed annual volume and a small number of complexity bands. Seats should normally be unrestricted within the purchased package, while clinical outcomes should affect service-level credits or performance incentives rather than determine the core bill.
Study volume follows radiology economics more closely than headcount
The logic starts with Monetizely's 5-Step Pricing Framework, developed in Monetizing Agentic AI. The framework moves through Goals & Segments, Positioning & Packaging, Pricing Metric, Rate-setting, and Operationalization. Goals & Segments establishes which customer and commercial objective matter; Positioning & Packaging determines what capabilities that customer buys; Pricing Metric chooses the unit that causes the bill to grow; Rate-setting assigns prices and volume curves to that unit; Operationalization turns the design into quoting, billing, overage, renewal, and reporting rules. For radiology-group SaaS, the third step carries unusual weight because a poor metric can overwhelm otherwise sound packaging: charge for radiologists and automation suppresses the bill; charge for raw infrastructure and customers cannot forecast spend; charge for outcomes and attribution becomes the argument instead of the product.
A useful ranking should therefore ask which meter grows when the buyer receives more value, remains observable enough to bill without debate, and tracks vendor cost closely enough to protect gross margin. Applying those tests produces a committed ranking rather than an even survey of three choices.
The ranking reflects the underlying economics. A study enters the workflow, consumes storage and compute, creates interpretation work, moves through reporting and communication, and usually corresponds much more closely to the economic activity of a radiology group than another login does.
US payment mechanics make the point tangible. CMS's Physician Fee Schedule covers radiology services and assigns service-level RVUs that are converted into payment amounts; CMS's 2026 rule also explicitly adjusted selected non-time-based services, including diagnostic imaging interpretation, for expected efficiency gains. In other words, the reimbursement system already recognises that radiology's economic unit is the service performed, while the work attached to different services can differ.
Peer-reviewed research reaches the same operational issue from another direction. Studies of radiologist productivity have long used procedure volumes and RVUs rather than raw physician counts, and research has documented variation in reading time across examination types. A 2017 review of academic radiology quality metrics also showed that radiology performance extends beyond simple productivity into dimensions such as reporting quality, communication, safety, and service.
A seat therefore measures who can open the software. A completed study measures something much closer to why the software was purchased.
Commercial evidence already favours transactions when usage carries the value
Radiology vendors are not waiting for pricing theory to settle the question. Several current products already put study volume at or near the centre of the commercial model.
The evidence is strongest when we separate broad enterprise workflow from narrower point tools.
Sources: -, -.
The examples span different software categories, but the common pricing logic matters. When each additional unit of customer activity creates additional customer value and some incremental vendor cost, charging for that activity can let revenue expand without forcing the customer to manufacture more seats.
Sirona provides the closest radiology-specific proof point. Its 2026 product page describes one per-exam fee covering viewer, reporting, worklist, PACS archive, platform AI, and other applications. That design avoids an increasingly awkward question for AI-enabled workflow software: why should a vendor earn less when its automation lets the same radiologists handle more volume?
OmniPACS shows how the same metric can preserve predictability. Its published 2026 plans pre-package case volumes: 30 cases for $99 per month, 100 for $250, 250 for $562, and 700 for $1,400, translating to progressively lower published per-study rates. The buyer sees a monthly commitment; the vendor still expands with imaging activity.
Candelis takes the concept further by describing per-study pricing for both viewing and storage, with no additional ongoing charge beyond the per-study price. Whether every enterprise vendor should copy that exact inclusion policy is another question. The important lesson is that a study can be observable enough to bill and broad enough to absorb several cost components.
Infrastructure vendors confirm why the transaction should be a business transaction, not a raw technical event. As of 12 August 2026, AWS HealthImaging charges separately for frequent-access and archive storage plus API requests, while Google Cloud Healthcare API pricing combines storage, request volume, DICOM storage and retrieval, ETL, de-identification, and other usage components. Those units make sense for infrastructure buyers. Passing the same granular list through to a radiology practice would leave the customer budgeting API calls, retrieval patterns, and gigabytes rather than the clinical work the software supports.
Monetizely's view is therefore not merely "usage pricing is good". The right usage unit is the completed study.
Complexity bands fix the biggest weakness in a flat per-study fee
A naïve per-study price still has a serious flaw: studies are not economically identical.
A radiograph, a large multi-series MRI examination, and a complex CT workflow can require different amounts of storage, viewing activity, compute, and radiologist attention. CMS explicitly assigns different resource values to services, while peer-reviewed radiology research shows that reading time varies across examination types. Cloud infrastructure creates another source of variation because image size, storage class, API activity, and retrieval behaviour affect underlying cost.
Charging every study at the same rate can therefore reproduce the very pricing mismatch we are trying to remove. The answer, however, is not to expose a hundred CPT codes or cloud SKUs on the invoice.
A small weighting system is enough.
| Transaction design | Example billing weight | Commercial purpose | Monetizely view |
|---|---|---|---|
| Standard study | 1.0 unit | Creates the reference unit customers forecast against | Core meter |
| Higher-complexity study | 1.5 units | Recognises heavier workflow or compute without building a CPT tariff | Use selectively |
| Very high-complexity study | 2.5 units | Protects economics where storage, processing, or workflow burden is materially higher | Keep to a small defined set |
| Prior retrieval, routing, amended report | 0 additional units | Avoids billing customers repeatedly for the same clinical episode | Include |
| Optional third-party AI analysis | Separate published study rate where necessary | Keeps externally licensed variable AI cost visible | Add only when cost cannot be absorbed |
The numbers in the exhibit are modelled rather than market quotes; the modelling basis is consolidated in the Assumptions note.
The important design choice is restraint. Three understandable bands are better than turning a SaaS contract into another medical fee schedule.
Operational rules matter just as much. A vendor should define the billable event when the contract is signed: one unique diagnostic study accepted into the production workflow, counted once regardless of how often users open it, route it, fetch priors, or amend its report. Duplicate transfers should not create duplicate revenue.
That definition gives finance a forecastable meter and gives radiology operations an auditable count. More importantly, the unit survives automation. If AI enables 30 radiologists to interpret the workload previously requiring 35, the platform still shares economically in the increased throughput.
Seat pricing becomes less aligned as software makes each radiologist more productive
Seat pricing remains attractive because everybody understands it. It is easy to quote, easy to budget, and easy to reconcile against identity management.
Narrow radiology products show where the model still makes sense. As of 12 August 2026, RadPal prices its Impression plan at $119 per month and its Pro plan at $199 per month, with unlimited usage; institutional licences use multi-seat arrangements. A tool that runs beside an individual radiologist's existing reporting workflow has a natural user boundary, so paying for each active professional can remain reasonable.
PostDICOM offers another version. Its 12 August 2026 pricing combines subscription tiers with included users, storage, sharing allowances, devices, and locations. Its Enterprise tier lists six users and 2,000 GB of storage, while an additional user is listed at €25 per month on annual billing. Here the seat works partly because the product is packaged with several capacity limits rather than pretending that user count captures all usage.
Enterprise workflow is different. A full radiology platform should improve the amount of work each radiologist can handle. Charging mainly by radiologist creates a structural contradiction: the better the product gets at productivity, the weaker its own expansion mechanism becomes.
History elsewhere in SaaS shows what happens when a visible unit stops matching actual usage or value.
| Company | Meter correction | Date | Lesson for radiology SaaS |
|---|---|---|---|
| New Relic | Shifted towards consumption pricing | FY2022 | Its 10-K said the new model was intended to encourage customers to instrument more applications and hosts while reducing unforeseen fees and overages |
| Splunk | Added workload/infrastructure pricing alongside data-ingestion pricing | 2020 10-K | A single raw-ingestion measure was not sufficient for every customer; Splunk offered compute access as an alternative metric |
| HubSpot | Introduced "marketing contacts" | 2020 product change reported in Apr 2021 filing | Customers could pay for contacts they actually wanted to market to rather than treating every database contact as equally billable |
Sources: -.
None of those cases proves that per-seat pricing always fails. They demonstrate a broader pricing truth: vendors eventually face pressure when the charged unit diverges from what customers are actually doing.
Radiology software is moving towards exactly that divergence. Workflow automation, centralised worklists, cloud PACS, speech tools, and AI can allow volume to grow faster than radiologist headcount. A seat meter captures hiring. A study meter captures the workload the software is helping the group process.
The distinction becomes stark in a simple three-year model.
| Metric | Year one | Year two | Year three | Three-year spend |
|---|---|---|---|---|
| Per study | $300,000 | $324,000 | $349,920 | $973,920 |
| Per seat | $300,000 | $310,000 | $320,000 | $930,000 |
| Per outcome | $225,000-$375,000 | $243,000-$405,000 | $262,440-$437,400 | $730,440-$1,217,400 |
The model deliberately starts seat and transaction pricing at the same first-year spend. Study volume then grows faster than radiologist count. Outcome pricing swings far more widely because the proportion of studies producing the contractually defined outcome changes.
For a buyer, seat pricing initially looks more predictable. For a vendor creating real throughput gains, it steadily leaves value uncaptured.
Outcome pricing sounds more sophisticated than transaction pricing. Rather than charging when a study is processed, the vendor might charge when turnaround time improves, a critical finding is communicated, a follow-up is completed, a diagnostic miss is avoided, or some downstream patient result changes.
The problem is not ambition. The problem is attribution.
Radiology quality is multidimensional. Peer-reviewed research published in 2017 documented the range of quality measures used across academic radiology, while CMS's 2026 payment system distinguishes conventional fee-for-service radiology payments from broader alternative payment arrangements that carry accountability for quality and cost. A software supplier operating inside the imaging workflow does not control everything between interpretation and the eventual clinical outcome.
Consider a pulmonary nodule follow-up. Software may identify the recommendation, create a work queue, notify the responsible team, and track closure. Whether the patient eventually completes imaging can also depend on the referring physician, scheduling capacity, insurance authorisation, patient behaviour, and care delivered outside the radiology group's systems.
Making the vendor's core bill hinge on that final event introduces disputes over denominator definitions, exclusions, data quality, and causality. More troublingly, the highest-value outcome may be the least frequent. Pricing software around rare avoided harms can create an invoice that is mathematically clever but commercially hard to audit.
A completed diagnostic study avoids that problem. Both parties can count it.
Outcome measures still belong in the commercial agreement. A vendor promising better operations should accept service-level accountability around system uptime, routing latency, report availability, migration milestones, or other outputs directly under its control. Credits can apply when those commitments are missed, and performance bonuses can apply where a result is exceptionally clear.
The base meter should remain the study.
Committed study volume gives buyers predictability without weakening value alignment
Pure pay-as-you-go is not the end state Monetizely would recommend. Twilio's 2024 and 2025 filings highlight the vendor-side consequence of usage pricing: revenue becomes more variable as customer activity changes. Snowflake similarly distinguishes its consumption model from traditional ratable subscription software and allows customers to consume beyond contracted capacity under its model.
Radiology vendors can capture the alignment of usage without imposing that volatility on either side. The commercial architecture should use an annual committed quantity of weighted studies, billed monthly or quarterly, with a clear unit rate for consumption above the commitment.
A 600,000-study group might therefore buy a 600,000-unit annual commitment rather than receive 600,000 tiny invoices. Growth beyond an agreed collar should trigger the published overage schedule, while unused committed volume can receive limited rollover when the customer renews at an equal or larger commitment.
Packaging should remain separate from the meter. Enterprise features such as multi-site routing, analytics, advanced integration, disaster recovery, administrative controls, and premium support can distinguish packages. Study volume then scales the chosen package.
That separation avoids three recurring errors:
- Do not gate ordinary radiologist access. More readers using the same platform should usually improve adoption rather than trigger an unexpected licence audit.
- Do not bill technical noise as customer value. API calls, prior fetches, individual DICOM instances, and report edits belong beneath the study-level price unless they create an exceptional cost.
- Do not make "outcome" a synonym for transaction. A completed report is a transaction. Calling it an outcome does not create outcome-based pricing.
The architecture also gives procurement a much cleaner benchmark. Buyers can compare effective cost per 100,000 studies across competing systems, adjust for included modules, and model growth without guessing future radiologist hiring.
Monetizely's position for 2026 is therefore specific: radiology-group SaaS should anchor expansion to completed studies, not people. Packaging should govern which capabilities the customer receives; weighted study volume should govern how much the customer pays as activity grows; annual commitments should provide predictability; and outcomes should enforce accountability without becoming the dominant meter.
For operators designing or buying the next contract, five decisions follow.
Make the completed diagnostic study the board-level primary meter. Define exactly when a study becomes billable and count the clinical episode once, regardless of internal routing or repeat access.
Replace one flat unit with a small complexity schedule. Use no more than a few bands, validated against actual compute, storage, workflow activity, and customer economics rather than an elaborate clinical coding table.
Sell committed annual volume rather than uncapped pay-as-you-go. Give customers a known base spend, explicit overage rates, and controlled rollover while preserving the transaction as the unit that drives expansion.
Remove ordinary radiologist seats from the growth equation. Permissions and identity controls still matter operationally, but a workflow platform should not earn less merely because automation lets fewer people handle more studies.
Put outcomes into performance terms instead of the core invoice. Tie credits and selected bonuses to results the vendor can directly observe and influence, while keeping downstream clinical outcomes outside the primary billing calculation.
Assumptions
The three-year model assumes 300,000 studies and 30 radiologists in year one, 8% annual study growth, and one additional radiologist per year. Transaction and seat models are normalised to $300,000 in year-one spend solely to compare metric behaviour. The outcome model uses the same economic starting point at a 20% qualified-outcome rate and shows a 15%-25% range. Complexity weights of 1.0, 1.5, and 2.5 are design examples, not vendor quotes. Public vendor prices and structures are stated as observed on 12 August 2026 unless another date is specified.
Footnotes
https://www.amazon.com/Monetizing-Agentic-AI-Handbook-Transformation/dp/B0H7Z13VKJ/
https://www.cms.gov/medicare/payment/fee-schedules/physician
https://www.cms.gov/newsroom/fact-sheets/calendar-year-cy-2026-medicare-physician-fee-schedule-final-rule-cms-1832-f
https://pmc.ncbi.nlm.nih.gov/articles/PMC5267601/
https://pmc.ncbi.nlm.nih.gov/articles/PMC5601514/
https://sironamedical.com/pro
https://www.omnipacs.com/
https://www.candelis.com/products/imagegrid-pacs/imagegrid-cloud-pacs
https://www.postdicom.com/en/pricing
https://www.radpal.ai/pricing
https://www.twilio.com/en-us/pricing/messaging
https://www.sec.gov/Archives/edgar/data/1447669/000144766926000021/twlo-20251231.htm
https://stripe.com/pricing
https://www.sec.gov/Archives/edgar/data/1640147/000164014725000052/snow-20250131.htm
https://www.sec.gov/Archives/edgar/data/1448056/000144805622000021/newr-20220331.htm
https://www.sec.gov/Archives/edgar/data/1353283/000135328320000008/a01312010k.htm
https://www.sec.gov/Archives/edgar/data/1404655/000156459021020337/hubs-def14a_20210603.htm
https://aws.amazon.com/healthimaging/pricing/
https://cloud.google.com/healthcare-api/pricing