- Guide
- 3 min read
How to measure what analytics and AI work is worth
Most data teams still cannot show what their work is worth. A value ladder, a way to set a baseline, and a one-page sheet to agree before anything gets built.
Gartner’s 2025 survey of chief data and analytics officers found 30% naming the measurement of business impact as their top challenge, and only 22% of organisations had defined, tracked and communicated business impact metrics for most of their data and analytics use cases.1 IBM’s study of 1,700 data leaders later that year found 29% with clear measures for the value of data-driven outcomes.2
The fix is unglamorous: agree the measure before the work starts. The rest of this guide is about doing that properly, whether the project is a dashboard, a data product or an AI system.
Why published studies disagree
Four widely quoted studies from 2025 and 2026 give very different pictures of AI returns, because each asks a different question.
| Study | Question asked | Result |
|---|---|---|
| McKinsey, 20263 | Does AI contribute to EBIT? | 37% some, about 6% at 5% or more |
| IBM CEO study, 20254 | Did initiatives deliver the expected ROI? | 25% yes, 16% scaled |
| BCG, 20255 | Is value captured at scale? | 5% future-built, 60% laggards |
| MIT NANDA, 20256 | Is there measurable P&L impact? | 95% none, as reported |
Your own business case needs the same discipline these studies show at their best: write the question down first, in words a finance partner would accept.
A value ladder
Value shows up in stages, and every project should report against each rung and know which one its sponsor cares about.
- Use. People open the report or call the system. It shows demand and nothing else.
- Process. Time per task, error rate, cases handled per person.
- Business result. Revenue per customer, cost to serve, conversion, churn, stock-outs.
- Financial effect. The impact on profit, with the cost of running the thing included.
A project that stalls at the third rung usually lacks a baseline or has ignored its running costs. Dashboards are particularly prone to stopping at the first rung, since view counts are easy to collect and say little.
Set a baseline and a comparison
Measure the current process for at least a month, recording volume, time and error rate. Then choose a comparison: run the new approach for half the cases or some of the teams while the rest continue as before. Fix the measurement window in advance, long enough to cover a full cycle such as a month-end close or a sales quarter.
Count every cost: build, running cost, review time, data preparation and licences. One respondent in five in McKinsey’s 2026 survey said their organisation was limiting AI use because of operating costs, and costs counted from day one are much harder to be surprised by.3 Last, write down the result at which you will expand, adjust or stop.
Where the value tends to sit
BCG estimates that 70% of AI’s potential value sits in core functions such as sales and marketing, manufacturing, supply chain and pricing.5 Its future-built group reported about twice the revenue growth and 40% more cost reduction than laggards. Those companies differ in many ways, so the figure is a reason to look at core functions first and proves nothing about any single project.5
The one-page value sheet
Agree these eight lines with the sponsor before any build starts, and revisit them at every review.
- Use caseOne sentence naming the decision or workflow it changes.
- OwnerThe business leader who answers for the result.
- MeasureOne primary measure, and the rung of the ladder it sits on.
- BaselineCurrent value, period and how it was measured.
- ComparisonControl group, or a before-and-after design.
- WindowStart and end dates.
- CostsBuild, monthly running cost and review time.
- Stop ruleThe result at which you expand, adjust or end it.
Sources
All sources were published in 2025 or 2026. The MIT NANDA report is cited through trade press coverage, which is named in the entry.
- Gartner Survey Finds One-Third of CDAOs Cite Measuring Data, Analytics and AI Impact as Top ChallengeGartner, 20 February 2025
- IBM Study: Chief Data Officers Redefine Strategies as AI Ambitions Outpace ReadinessIBM, 13 November 2025
- The state of AI in 2026: On the road to ROIMcKinsey & Company, 25 August 2026
- IBM Study: CEOs Double Down on AI While Navigating Enterprise HurdlesIBM, 6 May 2025
- AI Leaders Outpace Laggards with Double the Revenue Growth and 40% More Cost SavingsBoston Consulting Group, 30 September 2025
- MIT Report Finds Most AI Business Investments Fail, Reveals "GenAI Divide" (coverage of MIT NANDA, The GenAI Divide: State of AI in Business 2025)Virtualization Review, 19 August 2025
Keep reading
More from Resources
Ninety-five per cent of what? Reading the most quoted AI statistic
The claim that 95% of generative AI projects show no return is everywhere. The study behind it measures something narrower.
AnalysisCulture is the barrier everyone names
Why culture tops every list of data and AI obstacles, and the three places the 2025 and 2026 surveys say it shows up.
ExplainerDecision intelligence for BI teams
What decision intelligence adds to BI and machine learning, and how to model one decision before buying anything.
BIAAS 2027, Amsterdam
Hear it first-hand from the people doing the work
Two days of keynotes, panels and 1-on-1 meetings with senior data, analytics and AI leaders, on 23-24 March 2027.