ROI on fax triage automation comes down to one equation: daily document volume times minutes per document times loaded staff cost, minus the exception work that remains, minus the cost of the software. Most multi-location dermatology groups model payback inside a year, because manual handling of an inbound document — open, read, match the patient, file, route — runs several minutes and drops to a short exception review after automation. The number that decides it is minutes per document, and almost every practice underestimates it until someone times it.
The model, in four inputs
Skip the vendor's ROI calculator. Build your own from numbers you can collect in an afternoon, then hand a vendor your inputs and make them work against your reality rather than their averages.
1. Daily inbound document volume. Count across every fax line and every location. Practices with three or more sites routinely find lines nobody remembered — an old number at a location that moved, a dedicated path line that bypasses the main queue.
2. Current minutes per document. Have whoever works the queue time themselves honestly for a few hours across a normal day. Don't average a guess. Separate two buckets if you can: simple documents (records requests, insurance correspondence) and complex ones (referrals, prior auth responses, path with an unclear patient match). The complex bucket usually runs two to three times the simple one.
3. Loaded hourly cost. Wages plus benefits, payroll taxes, and overhead — not the base hourly rate. For most front-office roles the loaded figure is meaningfully higher than the number on the offer letter, and using the base rate understates your current spend by a third or more.
4. Post-automation exception rate. The share of documents that will still need a human. Get the vendor to estimate this against a sample of your own documents, not their marketing average. Then discount their number, because week-one performance is not steady state.
Current annual cost is volume × minutes ÷ 60 × loaded rate × working days. Post-automation cost is the same equation applied to the exception share at a shorter handling time, plus subscription, implementation, and any EHR interface fee.
A worked example, with the assumptions labeled
Assumptions stated plainly so you can swap in your own:
| Input | Assumption |
|---|---|
| Locations | 4 |
| Inbound documents per day, all sites | 180 |
| Weighted average minutes per document | 6 |
| Loaded hourly cost | $28 |
| Working days per year | 250 |
| Post-automation exception rate | 15% |
| Exception handling time | 1 minute |
Today: 180 × 6 minutes = 1,080 minutes/day = 18 hours/day. At $28 loaded, that's $504/day, or roughly $126,000/year — a bit over two FTEs doing nothing but inbound document handling.
After: 15% of 180 = 27 documents/day at 1 minute each = 27 minutes/day. That's about $13/day, or roughly $3,200/year in remaining labor.
Labor delta: about $122,000/year. Net that against the software subscription, implementation, and any EHR interface fee, and the payback question becomes arithmetic rather than debate.
Two honest caveats. First, that labor doesn't disappear as cash unless you actually reduce headcount, and most groups don't — they redeploy. Second, your minutes-per-document number is doing enormous work in this model. If your real figure is 3 rather than 6, the whole thing halves. Measure it.
For external anchoring: the 2025 CAQH Index puts the remaining annual savings available from automating manual and partially manual administrative transactions at roughly $21 billion industry-wide, and inbound document handling sits inside that. The MGMA 2026 Regulatory Burden Report found nearly 95% of surveyed practice leaders reporting increased regulatory burden over the prior three years. The direction of travel is more paperwork, not less.
Returns that are real but harder to put on a spreadsheet
The labor delta is the defensible core of the business case. These are the things that show up in operations and rarely in the model.
Faster path-result turnaround. If your lab signs out within the standard two business days and your office adds another day getting the report to the physician, that third day is yours. Automation removes most of it. Hard to monetize, easy for a physician partner to feel.
Fewer missed prior auth responses. Dermatology carries a heavy authorization load — dermatology-specific research has put the weekly burden at roughly 13 hours per practice with about a third of requests denied. A denial or request-for-information sitting unread in a shared queue past a payer deadline converts into a re-submission or a write-off. Even a handful of those a year is real money.
Reduced overtime and turnover. Document backlogs create end-of-day catch-up. Front-office turnover is expensive to replace and disproportionately triggered by work that feels endless and invisible.
Multi-site standardization. This one is specific to groups and often worth as much as the labor. Four offices that each developed their own filing conventions produce charts that only make sense to the person who filed them. Centralized, rule-driven routing gives every site the same taxonomy — which matters enormously if you're acquiring practices or preparing for a transaction.
Referral capture. A referral that sits three days in a queue is a patient who may have gone elsewhere. This is the highest-value item on this list and the hardest to attribute, because you never see the appointment you didn't book.
How the case differs for a PE-backed derm platform
If your group sits inside a private-equity-backed platform, the business case gets evaluated on a second axis, and it's worth building the model to answer both.
The operating case is the one above — labor, capacity, turnaround. The platform case is about what the asset looks like at exit. Three things matter there.
Integration cost per acquisition. A platform that acquires two or three practices a year pays a repeated onboarding cost every time a new site arrives with its own filing conventions and its own fax lines. Rule-driven routing turns that from a months-long normalization project into a configuration task. That reduction is real and recurring, and it compounds with acquisition pace.
EBITDA quality. Labor savings that show up as redeployed capacity read differently to a buyer than labor savings that show up as reduced headcount. Neither is wrong, but decide which story you're telling before you build the model, because it changes what you measure.
Operational diligence. A buyer's diligence team will ask how documents move across sites and how consistent the charts are. "Every location does it their own way" is an answer that costs money at the table. Standardized routing with an auditable log is a materially better one.
None of this replaces the labor math. It means a platform CFO should run the model twice — once on operating savings, once on what standardization is worth across the acquisition pipeline.
The costs to net out
Model these explicitly rather than discovering them in month two.
- Subscription. Usually priced per document or per volume rather than per user. Ask how it scales as you add locations, and whether there's a volume tier cliff.
- Implementation and configuration. Sometimes bundled, sometimes a separate line. Ask what's included and what's billed hourly.
- EHR interface fees. The most common budget surprise in this category. Many EHR vendors charge a one-time build fee plus annual maintenance per interface. Get the number in writing from your EHR vendor, not from the automation vendor's estimate.
- Internal change management. Someone on your team owns the taxonomy design, the shadow-mode comparison, and staff retraining. Budget real hours for it — this is where thin implementations go wrong.
- Parallel-run period. During shadow mode you're paying for both the software and the existing manual process. Several weeks of overlap, by design.
Honey Health's fax triage agent is built around the post-automation shape this model assumes — documents filing into the EHR the practice already runs, with the residual work arriving as a single exception queue rather than a second system staff work alongside the chart.
Where the ROI is genuinely thin
A business case that doesn't name its own weak cases won't survive a skeptical CFO.
Low volume per site. A four-location group averaging fifteen documents a day per site has a much weaker case than one averaging fifty. Run the math per site, not just in aggregate.
Highly variable document quality. If a large share of your inbound volume is multi-generation faxes and handwritten forms, your exception rate stays high and the savings compress.
Mid-EHR-migration. Adding an integration layer while the underlying chart system is changing is a sequencing problem. Wait.
No slack to redeploy into. If the recovered hours have nowhere productive to go and you're not reducing headcount, the savings are theoretical. Know what the freed capacity will do before you buy.
How to measure it after go-live
Set the baseline before you turn anything on, or you'll spend the renewal conversation arguing about counterfactuals.
Track four numbers monthly:
- Documents processed, total and by auto-file versus exception. The exception share should fall over the first several weeks as the system learns your recurring senders. Flat after four to six weeks is a month-two vendor conversation.
- Staff hours on document handling. The direct labor line. Sample it periodically rather than trying to track continuously.
- Receipt-to-filed interval. Should collapse to near-zero on auto-filed documents.
- Misfile and correction rate. The quality guardrail. Speed that comes with more errors isn't savings, and this is the number a skeptical partner will ask for.
Report those four to whoever approved the spend on a fixed cadence. Groups that instrument the rollout from day one have a straightforward renewal conversation. Groups that don't end up relitigating the original assumptions a year later with no data.
Frequently Asked Questions
What's a realistic payback period for a multi-location dermatology group?
Most groups with meaningful daily volume across several sites model payback within a year on labor alone, before counting referral capture or faster path turnaround. The variables that move it most are documents per day and current minutes per document. Build the model on your measured numbers, not a vendor's benchmark.
Do we have to cut staff to realize the ROI?
No, and most practices don't. The common pattern is redeploying recovered hours into work that's harder to automate — scheduling, prior auth follow-up, patient callbacks, denial work. Frame the case as recovered capacity rather than headcount reduction; it's both more honest and easier to defend to a partner group.
How do we account for EHR interface fees in the model?
Get the number directly from your EHR vendor before you sign with the automation vendor, and include both the one-time build fee and any annual maintenance. This is the most common budget surprise in the category, and it's entirely avoidable by asking early.
Does ROI improve as we add locations?
Generally yes, on two fronts. Volume scales while much of the configuration work is one-time, and you gain standardized filing conventions across sites — worth real money if you're acquiring practices or preparing for a transaction. Still run the per-site volume check; aggregate numbers can hide a site that's too small to matter.
What if our exception rate stays higher than the vendor projected?
Ask for the diagnosis rather than accepting the number. High persistent exception rates usually trace to one of three causes: a taxonomy with too many categories, an integration that can't reach the right EHR module, or genuinely poor source document quality from specific senders. The first two are fixable. The third is a reason to renegotiate your assumptions.
