[ 01 ] RPA ROI Calculator
RPA ROI calculator: what do the bots actually return?
Model one robotic process automation case with licence and orchestrator fees where they belong, an automation rate set by what the bot completes unattended, and payback measured against the real cost of building and maintaining it.
The short answer
RPA ROI is calculated with the same formula as any workflow automation, but three inputs decide whether the case is honest: bot licences and orchestrator fees go in monthly operating cost, bot development and testing go in implementation cost, and the automation rate must reflect only the share of cases the bot finishes without a human. Attended bots and high exception rates are where most RPA business cases quietly fail.
What is RPA ROI?
RPA ROI is the return a software robot produces against the full cost of owning it. Gross value is the handling time the bot gives back plus the rework it prevents, priced at a loaded hourly rate and scaled by the share of cases it completes end to end. Total cost is bot development plus the recurring stack it runs on: per-bot licences, orchestrator or control-room fees, the virtual machines it drives, and the engineering time that keeps it working when the underlying screens change.
- 3-year RPA ROI
(3-year gross value − 3-year total cost) ÷ 3-year total cost × 100- Payback (months)
bot build cost ÷ (monthly gross value − monthly licence, orchestrator and maintenance cost)- Effective automation rate
cases completed unattended ÷ total cases
Straight-through rate, not deployment count, is what moves an RPA business case. A bot that touches every case but finishes 40 percent of them earns the conservative scenario, not the aggressive one. Full methodology and a worked example →
[ 02 ] Calculator
Model one bot.
Put licences and orchestrator fees in monthly operating cost, bot development in implementation cost, and set the automation rate to your measured straight-through rate.
Start with an RPA example
Illustrative inputs only. Replace every value with observed operating data before making a decision.
[ → ] Projection
Unattended invoice bot business case.
Steady-state annual net value
$12,975
gross workflow value minus twelve months of operating cost
Hours / year
525
Payback
11.1 mo
3-year ROI
95%
Business-case breakdown
- Annual cost of the current workflow $30,625
- Gross annual value returned $18,375
- Annual operating cost $5,400
- First-year net value $975
- Three-year net value $26,925
Scenario comparison
| Scenario | Payback | 3-year ROI | 3-year net |
|---|---|---|---|
| Conservative 40% | 21.0 mo | 30% | $8,550 |
| Expected 60% | 11.1 mo | 95% | $26,925 |
| Aggressive 75% | 8.2 mo | 144% | $40,706 |
[ 03 ] Cost modelling
Where RPA costs actually sit.
The formula is not the hard part. Putting each cost in the right field is, because RPA carries recurring fees that other automation does not.
Bot licences → monthly operating cost
RPA platforms charge per bot, usually with separate rates for attended and unattended runtimes. That fee continues for as long as the bot exists, so it belongs in monthly operating cost, not in the build. Counting it as a one-time number is the single most common way an RPA business case is overstated.
Orchestrator and infrastructure → monthly operating cost
The orchestrator or control room that schedules, queues, and logs bot runs is licensed separately on most platforms, and each unattended bot needs a machine to drive. Add the orchestrator tier, the virtual machines, and any monitoring the operation depends on into the same monthly figure.
Development and testing → implementation cost
Process discovery, building the bot, and testing it against real cases are one-time and belong in implementation cost. So does the work to document the process, because an undocumented bot becomes unmaintainable the first time the person who built it moves on.
Maintenance → monthly operating cost
Screen-scraping and UI-driven bots break when the applications underneath them change. Budget recurring engineering time for that, in the monthly figure. If you have no estimate yet, modelling it as zero will produce a payback number you should not act on.
Attended vs unattended → automation rate
An unattended bot that runs a whole process on a schedule can justify the expected or aggressive scenario. An attended bot that assists a person through part of a task cannot: the person is still in the loop for every case, so the recovered time is a fraction of the handling time, and the conservative scenario is the honest starting point.
Exceptions → automation rate and rework
Every RPA queue has cases the bot cannot finish. Those return to a human, and the triage costs time the manual process did not spend. Set the automation rate to the straight-through rate you measured, and keep rework minutes at a realistic level rather than dropping them to zero.
What this calculator does not model.
Stated plainly, so the number you leave with is one you can defend in a budget conversation.
Enterprise licence negotiations
Platform pricing is negotiated, tiered, and often bundled. Enter the per-month figure your own quote implies rather than a list price, and re-run the projection if the tier changes.
Revenue, risk and capacity effects
Faster cycle times, fewer compliance breaches, and freed capacity are real, but we do not price them here. Including them makes almost any bot look justified, which defeats the purpose of running the numbers.
Whether RPA is the right tool
If the systems involved have APIs, an integration is usually cheaper to run and far less fragile than a bot driving a screen. The calculator will happily model an RPA case that an API would beat — comparing the two is a design decision, not an arithmetic one. See data entry automation for the integration-first approach.
Common questions.
How do you calculate RPA ROI?
Price the handling time the bot gives back, scale it by the share of cases it completes unattended, then subtract what it costs to build and run. Three-year RPA ROI is three-year gross value minus three-year total cost, divided by three-year total cost, times one hundred. Bot licences, orchestrator fees, virtual machines and maintenance all belong in the recurring cost, and bot development belongs in the one-time cost.
Where do bot licences and orchestrator fees go in the model?
Both go in monthly operating cost, because both recur for as long as the bot exists. Putting either in implementation cost shortens the payback period artificially and is the most common error in RPA business cases.
Should attended and unattended bots use the same automation rate?
No. An unattended bot running a whole process on a schedule can justify the expected or aggressive scenario. An attended bot keeps a person in the loop on every case, so it recovers only part of the handling time and should start on the conservative scenario.
How do exceptions change the number?
Exceptions come back to a human and cost triage time the manual process never spent. Set the automation rate to your measured straight-through rate rather than the platform capability, and keep rework minutes realistic instead of zero.
What is a good RPA ROI?
There is no universal threshold; it depends on the payback window your business accepts. A common bar is payback inside twelve months and positive three-year ROI on conservative assumptions. If a bot only clears the bar on the aggressive scenario, treat it as unproven rather than approved.
Is RPA always the right choice for these workflows?
No. Where the systems involved expose APIs, an integration is normally cheaper to operate and much less fragile than a bot driving a user interface, because it does not break when a screen changes. Model both before committing.
[ 04 ] Next step
Take your number into a scoping call.
Bring the projection and the process behind it. We map where the bot would sit, say whether an integration would beat it, and tell you plainly if the math does not hold.