When does each automation investment pay back its build cost, and how do cost savings compound over time?
№ 206 · Automation ROAS curve
Definition
MetricCumulative savings from an automation divided by cumulative build and run cost, tracked from launch
Unitratio of cumulative savings to cumulative cost (x), plus the payback month
Cumulative cost-saved from automation divided by build cost, over months. Like CAC payback but for internal builds. Build by quantifying labor saved per automation. Identifies which automations were worth it. Example: High-volume automation 1.0x at M4, hits 4.4x at M24; low-volume 1.0x at M18.
- ROAS
- Return on ad spend. Revenue attributed to a campaign divided by the spend on that campaign.
- CAC
- Customer acquisition cost. Fully loaded sales and marketing spend in a period divided by new customers won in that period.
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| payback beyond 18 months, or a curve that never crosses | payback at 9 to 12 months, three year return 1.5x to 2.5x | payback inside 6 months, three year return above 2.5x |
Enterprise process automation and robotic process automation, drawn from vendor-commissioned Forrester Total Economic Impact studies (2022 to 2024) built on composite organisations, plus practitioner survey data on expected against achieved payback. Composites skew large: the Microsoft study models a 30,000 employee, 10 billion USD revenue organisation. · Almost every published figure here is vendor-commissioned, which biases the results upward in a predictable way: the TEI method builds a composite from interviewed customers who chose to stay on the platform, so it is closer to a best-realistic case than a median. Reported three year returns range from 97% (UiPath composite) to 262% (Automation Anywhere composite) on the same broad method, which is itself a warning about precision. The single most useful figure is the gap between expectation and outcome: organisations that piloted RPA expected a 9 month payback while those that implemented and scaled it reported 12 months. Build the internal case at 12 months even where the vendor case says 6. Note also that this is a different family from marketing ROAS despite the shared shape: the denominator is a one-off build cost plus a recurring run cost, not recurring media spend, so the curve should be read for the crossover and the slope after it, not for a steady-state ratio.
Category split omitted: Published automation returns are segmented by vendor platform rather than by process type, and the underlying composites are not built on a comparable basis, so no honest category split is available.
When it looks bad
The curve flattens before it crosses breakeven, because run and maintenance cost grows at the same rate as the savings and the automation never amortises its build.
Build cost 84,000, monthly saving starts at 9,000 then decays to 4,500 by month 8 as exception volume accumulates. The curve crosses at month 19 instead of the modelled month 7.
What to do about it
- Model the run cost from the start, not just the build. Licence, monitoring and exception handling are what flatten the curve, and vendor TEI composites include them explicitly while internal business cases routinely do not (Forrester TEI, 2024).
- Size the first automation on volume rather than on visibility. Forrester's Power Automate study located the largest per-employee saving, around 200 hours a year, in high-volume repetitive work such as data entry, invoicing and document gathering, not in the processes that get the most executive attention (Forrester TEI for Microsoft, 2024).
- Budget payback at 12 months even where the vendor case says 6. Organisations that piloted RPA expected 9 month payback while those that actually scaled it reported 12, and the gap is the honest planning number (Deloitte global RPA survey, via AIMultiple).
- Track exception rate on the automated path weekly and treat a rising rate as a stop condition. A growing exception rate converts a labour saving into a labour transfer, which is the most common single reason these curves decay after month six.
Sources
- Forrester Consulting, commissioned by Microsoft Benefits of 55.93 million USD over three years against costs of 16.08 million USD, net present value 39.85 million USD, ROI 248%, payback under six months. The composite trains 1,800 automation builders. tei.forrester.com ↗
- Microsoft Employees in high-volume repetitive tasks such as data entry, invoicing and document gathering saved 200 hours per year. One pharmaceutical organisation saved 11,000 hours running 72 RPA automations for document processing. Consolidation onto a single platform produced an estimated 9.5 million USD in legacy system savings. microsoft.com ↗
- Forrester Consulting, commissioned by Automation Anywhere Payback period under 12 months and 262% return on investment within three years, with the majority of value coming from redeploying employees to higher-value work rather than from headcount reduction. automationanywhere.com ↗
- Forrester Consulting, commissioned by UiPath Benefits of 12.08 million USD over three years against costs of 6.14 million USD including licence fees and planning and implementation, giving an ROI of 97%. Materially lower than the other composites on comparable methodology. roboticsai.co.uk ↗
- Forrester Consulting, commissioned by SS&C Blue Prism Payback calculated at under six months, and the study notes that organisations receiving only productivity benefits from an automation programme are leaving the majority of the value unrealised. blueprism.com ↗
- Deloitte, reported by AIMultiple Organisations that piloted RPA expected a 9 month payback on average, while those that implemented and scaled it achieved 12 months. On average an RPA bot costs one third of an offshore employee and one fifth of an onshore employee, and respondents expected roughly 20% of full-time-equivalent capacity to be automatable. research.aimultiple.com ↗
- IDC, commissioned by UiPath Three quarters of respondents said RPA deployment reduced the need for labour but only 13% said it resulted in actual job cuts, which means most modelled savings must be realised through redeployment rather than payroll reduction. Flagged as 2021 data in a 2026 reference. uipath.com ↗
- Softobiz Enterprise business process automation payback ranges from under 6 months for well-scoped platform deployments to 18 months for complex multi-process implementations, with peer-reviewed cross-industry returns spanning 30% to 300%. Explicitly flags that all vendor-sponsored figures must be read against their methodology. softobiz.com ↗