Walk into any RPA vendor’s resources page and you’ll find a wall of eye-popping numbers: hundreds of percent ROI, millions of dollars saved, payback periods measured in months. Those numbers show up again in sales decks, analyst webinars, and — if we’re honest — in plenty of blog posts that repeat them as settled fact. We pulled the actual case studies behind three of the most commonly cited figures from UiPath, Automation Anywhere, and Microsoft Power Automate, and looked at what’s really being claimed, who calculated it, and how it was verified (or wasn’t).
The short version: every number below is a vendor-sourced or vendor-commissioned claim, not an independently audited result. That doesn’t make them worthless — self-reported customer outcomes are still a useful signal — but they should be read as marketing evidence, not proof of what your organization will achieve.
At a Glance: Vendor ROI Claims vs. Independent Verification
| Vendor | Named Case Study | Claimed ROI/Result | Independently Verified? |
|---|---|---|---|
| UiPath | Royal Mail Group (UK postal/logistics) | £55M in value delivered to date; “over 300% above expectations” vs. an original £13.7M target; 662,000 employee hours returned | No. Published by UiPath on uipath.com as a case study; quotes come from Royal Mail staff, but no external auditor or methodology is cited. |
| Automation Anywhere | Cerner (healthcare IT) | 628% ROI; 6-month payback; $130,007 average annual benefit; ~60% developer productivity gain | Partially. Figures originate from a Nucleus Research “ROI Case Study” one-pager (an analyst firm, not Automation Anywhere itself) — but the page discloses no calculation methodology, and the underlying study appears to date back roughly a decade, recycled in current marketing. |
| Microsoft Power Automate | Composite organization built from interviews at 6 companies (incl. Evri, CoreLogic Asia Pacific, Brisa) | 248% ROI, $39.85M NPV, under 6-month payback (3-year, risk-adjusted) | Partially. Conducted by Forrester Consulting, a third party — but the study was commissioned and reviewed by Microsoft, models a hypothetical “composite” company rather than reporting any single real customer’s P&L, and assumes Power Apps is deployed alongside Power Automate. |
Short Verdict
None of the three big RPA ROI claims in heaviest circulation — UiPath’s Royal Mail numbers, Automation Anywhere’s Cerner case study, or Microsoft’s Forrester-commissioned Power Automate study — have been independently audited in the way a financial statement is audited. The Forrester study is the most methodologically transparent of the three because Forrester publishes its assumptions and composite-company model, and because Forrester states outright that it “maintains editorial control” even though Microsoft paid for and reviewed the study. The UiPath and Automation Anywhere figures are closer to traditional customer-reference marketing: real customer names, real quotes, but no visible methodology and no third party checking the math.
What’s Actually Being Claimed, Case by Case
UiPath + Royal Mail Group
UiPath’s own case study page on uipath.com, published under UiPath’s byline, credits Royal Mail Group with £55 million in value “delivered with RPA to date,” against an original internal target of £13.7 million — described as “over 300% above expectations.” The £55 million figure is a blend of four very different categories: £4.8 million in employee time saved, £22.4 million in “additional revenue,” £17.5 million in cost reduction/avoidance, and £10.3 million in “capacity creation.” Mixing time savings, cost avoidance, and new revenue into one headline number is a common pattern in RPA marketing, and it’s worth unpacking before treating “£55 million saved” as a literal cash figure — much of it is estimated value, not verified savings on a balance sheet. The quotes in the piece come from named Royal Mail staff (a program lead, a business analyst, and others), which lends it more credibility than an anonymized case study, but there’s no mention of an external audit of the numbers.
Automation Anywhere + Cerner
The “628% ROI” figure attached to Cerner (the healthcare IT company, now part of Oracle Health) traces back to a Nucleus Research ROI Case Study, a short one-page report style Nucleus has produced for many enterprise software vendors over the years, marked in this instance as a past “ROI Award Winner.” It states a 6-month payback, a 1:11.1 cost-benefit ratio, and an average annual benefit of $130,007, attributed to Ben Mitchell, a systems engineer at Cerner Ambulatory. Nucleus Research is an independent analyst firm rather than Automation Anywhere itself, which is a meaningful step up from a vendor writing about its own customer — but the publicly visible excerpt discloses no calculation methodology, and these vendor-commissioned ROI one-pagers are typically paid for by the software company being evaluated. The dated, awards-badge styling of the page also suggests this is an older study still being recirculated in current sales and marketing material rather than a fresh 2026 result.
Microsoft Power Automate (Forrester Total Economic Impact Study)
Microsoft’s most-cited Power Automate ROI figure — 248% ROI with $39.85 million in net present value over three years — comes from a Forrester Consulting “Total Economic Impact” (TEI) study commissioned by Microsoft in 2024. Forrester interviewed representatives at six organizations (Microsoft supplied the customer names and introductions, but Forrester conducted the interviews itself) and built a hypothetical “composite organization” — modeled at 30,000 employees and $10 billion in revenue — to represent the blended financial impact. Forrester is explicit that it “maintains editorial control” of TEI studies even when a vendor pays for them, and the published report states plainly that the benefits assume customers are also running Power Apps alongside Power Automate, and that Forrester “advises readers to use their own estimates within the framework provided” rather than take the composite numbers as a guarantee. That’s a materially more transparent disclosure than either the Royal Mail or Cerner case studies offer — but it’s still a commissioned study, not an independent audit, and the 248% figure describes a modeled composite, not any single real company’s actual results.
Why These Numbers Deserve Skepticism — Without Being Dismissed
- Self-selection bias. Vendors choose which customers to feature. A case study is, almost by definition, a success story a company agreed to let the vendor publish — failed or disappointing deployments rarely become case studies.
- Blended metrics. “Value delivered” often mixes hard cost savings with soft estimates of “capacity created” or “additional revenue,” which inflates headline totals in ways that are easy to miss on a skim.
- Undisclosed methodology. None of the three studies above publish the underlying calculation in enough detail for an outsider to reproduce the number.
- Age and reuse. ROI figures — especially from analyst-firm one-pagers like the Nucleus Research reports — can circulate in current sales material for a decade or more after the original study was conducted.
- They’re still useful as a prior. A case study that names a real company and a named employee is harder to fully fabricate than an anonymous number, and it does tell you the deployment happened and that the customer was willing to publicly associate their name with a positive outcome.
How to Build Your Own ROI Case Instead of Borrowing a Vendor’s
- Run a time-boxed pilot on your own processes and measure before/after cycle time, error rate, and FTE hours directly — don’t extrapolate from someone else’s numbers.
- Separate hard savings (headcount, overtime, error-correction cost) from soft/estimated value (capacity “created,” revenue “enabled”) in your own business case, and present them separately to stakeholders.
- Ask the vendor directly for the methodology behind any case study or analyst report they cite, including whether the study was vendor-commissioned.
- Treat any vendor-supplied ROI calculator (several of which the three vendors above offer on their own sites) as a sales tool, not a financial forecast.
Bottom Line
UiPath’s Royal Mail numbers, Automation Anywhere’s Cerner figures, and Microsoft’s Forrester-commissioned Power Automate study are all real, named, traceable claims — not fabrications — but all three are vendor-sourced or vendor-commissioned, none has been independently audited, and at least one (Cerner/Nucleus) appears to be a much older study still being used in current marketing. Use them as a sanity check that RPA can produce real value in the right circumstances, not as a substitute for measuring your own pilot.
FAQ
Are vendor RPA ROI case studies fake?
No — the case studies we reviewed name real customers and real employees. But “real” doesn’t mean “independently verified.” They’re vendor-published or vendor-commissioned marketing content, selected because the outcome was favorable.
What’s the difference between a vendor case study and a Forrester TEI study?
A vendor case study is written and published entirely by the software company about its own customer. A Forrester TEI study is commissioned and paid for by the vendor, but conducted and written by Forrester, which states it retains editorial control and discloses its modeling assumptions — a meaningfully more transparent (though still not independent) process.
Why do RPA ROI percentages look so much higher than typical software ROI?
RPA case studies often count the fully-loaded cost of human labor hours displaced against a relatively small software and implementation cost, which can produce triple-digit ROI percentages even when the absolute dollar savings are modest for a large enterprise.
Can I ask a vendor for the raw methodology behind a cited ROI number?
Yes, and you should. If a vendor or its account team can’t explain how a headline ROI figure was calculated, treat the number as directional marketing rather than something to build a budget justification on.
