TL;DR: The ROI of fax triage software for a multi-location behavioral health group is daily document volume × minutes of manual handling per document × loaded hourly staff cost, minus the residual exception-queue time and the subscription. Most groups find the labor line alone clears the spend before any revenue effect is counted. The multi-location variable that changes the math is queue fragmentation — per-site fax inboxes make it impossible to load-balance intake staff across locations, and a centralized triage layer makes it possible.
The formula, stated plainly
Three components. Two are countable and one is an estimate, and mixing them up is how these models lose credibility with a partner group.
Labor recovered (countable). Documents per day × minutes handled per document × 21 working days × loaded hourly rate ÷ 60. "Loaded" means salary plus benefits and payroll taxes — typically 1.25 to 1.4 times base pay. Use the rate of the person actually doing the work.
Revenue effect (estimated). Referrals that reach intake same-day convert better than referrals worked two days later, because the patient hasn't already called somewhere else. Real, but practice-specific. Any vendor quoting you a universal conversion lift is guessing.
Cost (countable, once you ask). Subscription plus implementation plus the internal hours to configure and supervise the ramp. That third item is the one buyers leave out.
If the labor number alone doesn't clear the subscription, be skeptical of a model that only works once soft benefits are added. That's the test worth applying before you take anything to your partners.
Working the labor math across a range of volumes
Rather than one invented practice, here's how the number moves. All three scenarios assume three minutes of average handling per document — open, read, identify, search the chart, categorize, file, route — and a loaded staff rate of $28 an hour. Substitute your own figures; the shape holds.
- 100 documents/day — 5.0 staff hours daily, about 105 hours a month, roughly $2,940 in monthly labor
- 250 documents/day — 12.5 staff hours daily, about 263 hours a month, roughly $7,350 in monthly labor
- 500 documents/day — 25.0 staff hours daily, about 525 hours a month, roughly $14,700 in monthly labor
Two things to notice. This covers sorting and filing only; it excludes the downstream work of acting on documents, which automation doesn't remove. And automation doesn't take it to zero — a realistic steady-state target is 60 to 80% of documents filing without a human touch, so you recover most of the hours above, not all.
At 250 documents a day and 70% straight-through processing, you're recovering roughly 184 hours a month, about $5,150 in loaded labor. That's the figure to hold against a subscription quote.
The three-minute assumption is the one most worth checking. Practices that measure often find the true average is closer to two minutes on clean payer correspondence and eight on a multi-page agency packet, and in behavioral health the hard documents are a larger share of the mix than in general medical.
The multi-location wrinkle
This is what separates a group's math from a single-site practice's, and it usually gets left out entirely.
A multi-location behavioral health group typically ends up with a fax number per site, because that's how each location was set up and nobody consolidated. The consequences compound:
- Staff can't be load-balanced. Your Riverside office is buried on Monday and your downtown office has slack, but the work is stuck in separate inboxes. You staff each site for its own peak rather than for the group's average.
- Nobody can see total volume or backlog. "How many unworked referrals does the group have right now" is a question that requires five people to answer, so it doesn't get asked.
- Documents get manually forwarded between sites. A referral for a patient who'll be seen at another location gets opened, identified, and re-sent. That's the manual sorting problem plus an extra step.
- Coverage breaks on absence. When the coordinator at a two-person site is out, that site's queue simply doesn't move.
A centralized triage layer changes the unit of work from "each site's inbox" to "the group's document stream." Routing by location still happens — it just happens by rule after classification, based on the requested location or the referring provider's geography, rather than by which number the fax arrived on.
The saving here isn't only hours. It's the ability to run intake as one function with one backlog metric, which is what makes staffing decisions possible at all.
What a stale referral costs a behavioral health group
The revenue side is bigger than the labor side and softer, so present it as a range with visible assumptions rather than a confident number.
Access is the pressure point. More than a third of the US population lives in a mental health professional shortage area, per workforce data compiled by NIHCM. KFF survey work found that among women who tried to get mental health care, 16% couldn't get an appointment at all, a third of them because providers weren't accepting new patients — and among those who did get one, 24% waited one to two months and 13% waited longer.
In that environment, a referral sitting unopened for two days is functionally a practice that isn't accepting the patient. They call the next name on the list.
The way to model this without inventing a number is to invert it: rather than forecasting a conversion lift, ask how many additional converted referrals per month it would take to cover the subscription by itself. For most behavioral health groups that number is small — often a handful of new patients — and a partner who won't accept a conversion estimate will usually accept that the threshold is low.
Two other revenue lines worth naming. Authorization letters that reach the coordinator promptly get acted on inside the window; ones discovered late become sessions delivered without authorization. And misfiles, which manual indexing produces at a rate nobody tracks, cost staff time to find and correct and occasionally more than that in a behavioral health chart.
What buyers underestimate
Four costs consistently get left out, and leaving them out is how a defensible decision turns into a disappointing quarter.
Implementation time is yours, not just the vendor's. Someone at your group defines the document taxonomy, sets confidence thresholds, maps routing rules to owners across locations, and validates output during the ramp. Budget real hours from a real person — usually your operations director or lead intake coordinator.
The exception queue needs a permanent owner. This is an operating cost, not a launch cost. It's much smaller than the work it replaced, but it isn't zero, and an unowned exception queue quietly becomes a second inbox.
The accuracy ramp runs weeks. Early straight-through rates sit lower while thresholds are conservative and the system is being validated against your document mix. Model the first quarter below steady state.
Vendor-reported outcomes are marketing until proven otherwise. Published numbers in this category come from vendors, describe their best deployments, and are rarely independently audited. Treat them as an upper bound. Ask instead for a reference customer with a comparable document mix and location count, and call them.
What drives the number up or down
Two groups with identical volume can land in very different places. The difference is usually one of these.
- Document variety. Seven distinct types with five owners produces more savings than one type in a folder, because manual sorting was doing more work.
- Current handling time. A group with a well-drilled coordinator at 90 seconds a document has less to recover than one averaging five minutes on an unstructured process. The worse your current process, the better your ROI — which is uncomfortable but true.
- Number of locations. More sites means more forwarding steps that vanish under centralized routing.
- Scan quality of your top senders. Counties, courts, and school districts often transmit degraded scans. If they're a large share of your volume, your straight-through rate will sit below the category average and your model should say so.
- EHR write-back depth. A system that files documents but can't create tasks leaves half the manual work in place. Confirm what "integration" means concretely before modeling savings on it.
Turnover belongs in the model too, if not as a guaranteed saving then as a risk the current process carries. Behavioral health settings run high administrative attrition, and repetitive document sorting is exactly the work people leave over. Replacing an intake coordinator costs recruiting, onboarding, and a productivity dip while someone learns your taxonomy. Meanwhile 84% of medical groups reported year-to-date operating costs up over the prior year, averaging about 11%, with labor the leading driver — so the alternative of hiring your way out gets more expensive each year.
How to build a defensible model in two weeks
You can produce a number your partners will believe without spending anything.
- Count for ten business days, per location. Every inbound document tagged by type. A tally sheet works. This gives you volume, mix, and the cross-site distribution.
- Time the handling for three days. Whoever works each queue logs start and stop. Compute an average and note the spread.
- Get the loaded rate. Base pay × 1.25–1.4 for the staff actually doing the work.
- Compute current monthly cost. Volume × time × rate × 21 days, summed across sites.
- Measure time-to-first-contact on new referrals. From fax arrival to the first outreach attempt. This is your revenue-side baseline and the metric your clinicians will notice moving.
- Shadow-mode before you model savings. Run live traffic through the candidate system with staff reviewing decisions, and measure the actual straight-through rate by document type. Replace the vendor's number with yours, then haircut the first quarter for the ramp.
Honey Health's Fax Triage agent is one option to run this math against. What decides it isn't the pitch — it's the shadow period, where you find out what share of your actual documents reach the right chart untouched, broken out by type and by location.
Frequently Asked Questions
What straight-through processing rate should we expect?
For a mixed behavioral health queue, 60 to 80% of documents filing without human review is a reasonable steady-state expectation — higher on structured payer correspondence, lower on agency and court documents and multi-page packets. Ask any vendor quoting above 90% to break that out by document type on your own samples.
How long until it pays for itself?
Groups with high volume and slow current handling often see the labor math clear the subscription within the first few months of steady state. Lower-volume groups may never get there on labor alone, which is why measuring your own baseline matters more than any published payback claim.
Should we count headcount reduction as savings?
Most groups redirect rather than cut. The recovered hours move toward work that was already backed up — authorization follow-up, waitlist outreach, referral callbacks. Present that as capacity rather than payroll reduction, because a partner who was promised a smaller payroll and got more capacity will feel misled.
Does consolidating fax numbers across locations make sense?
Not necessarily, and you don't have to. Triage can consume documents from multiple existing lines and route by rule after classification, so each site keeps its published number while the work becomes one queue. Consolidating numbers is a separate decision with its own referral-source disruption.
How do we account for implementation cost?
Ask the vendor for the full picture: subscription, one-time implementation, EHR integration if priced separately, and their estimate of internal hours during configuration and ramp. Add your own staff time at the loaded rate. Vendors quote the first item readily and the last one almost never.
Is this worth it for a two-location group?
Sometimes. Two locations is enough for queue fragmentation to cost you something, but the labor math still has to clear on volume. Under roughly 75 documents a day per site with queues that clear reliably, a good coordinator and your existing fax service usually win. The case strengthens with document variety more than with raw page count.
