Insights
Practical writing on measuring what programs, policies, investments and AI tools actually change.
How to measure the ROI of AI: a plain guide
Usage counts and staff surveys do not tell you whether AI paid off. A baseline, a fair comparison and full costs do. Here is how to set that up, step by step.
Why is Canada's inflation high? Gasoline, not prices broadly
Canada's inflation was 3.1% in September 2026, but 2.1% once food and energy are removed. Gasoline, up 27.8%, explains the gap. What that means for rates and budgets.
Why are US utility bills rising faster than pay?
Since January 2019 US electricity prices rose 47%, piped gas 47% and water, sewer and trash 39%. Pay rose 37% and all prices 34%. Here is why and what it means.
Canada's new business rate fell to 12.3% in 2023
In 2023, 12.3 of every 100 Canadian employer businesses were new and 11.9 closed. The 0.4 point gap is the smallest since 2016, apart from 2020. What it means.
Why do mobile phone prices differ so much across the EU?
From 2019 to 2025 mobile service got 37% cheaper in France and 51% dearer in Poland. One single market, very different bills. What the gap says and what to do.
Are companies hiring fewer people because of AI?
A New York Fed survey of service firms finds AI is changing hiring more than it is cutting jobs. 61% use AI, 15% hired fewer people because of it and 4% laid people off.
Can AI forecast earnings better than analysts?
A preprint tested 13 AI models on 45,964 company filings. Their earnings forecasts land near the middle of analyst forecasts and do not consistently beat individual analysts.
AI agents in UK finance: wide use, little autonomy
The Bank of England and FCA found 80% of responding UK financial firms run AI agents, but only 2% give them broad autonomy. The agents recommend and people decide.
Do executives overstate AI productivity gains?
About 750 US finance executives say AI raised labour productivity 1.8% in 2025. Calculated from the same firms' own revenue and staffing figures, the gain is 0.6%.
AI-fabricated content in court: 2,231 cases logged
A public database lists 2,231 court decisions that involve AI-made-up content as of 29 October 2026. Self-represented litigants account for 1,288 and lawyers for 880.
Did Canadian pay keep up with groceries and rent?
Canadian pay rose 33% since January 2019 and all prices rose 27%, so pay won on average. But groceries, rent and car insurance rose faster than pay, and those are bills nobody can skip.
Canada vs US phone and child care prices since 2019
Since January 2019, cell phone service prices fell 50% in Canada and rose 2% in the US. Child care fell 30% in Canada and rose 36% in the US. The same services took opposite paths.
US car insurance is up 49% since 2019, pay up 37%
US car insurance prices are up 49% since January 2019, against pay up 37%. Car insurance outran pay by the widest margin of any household bill we checked. Shop the policy every year.
US car repair prices are up 61% since 2019
US car repair and maintenance prices are up 61% since January 2019. That is close to twice the rise in all prices and well above pay. Get a second quote every time.
UK vet prices are up 67% since 2019
UK vet and pet care prices are up 67% since January 2019, nearly twice the 36% rise in all prices and above the 43% rise in pay. A new competition order will force vets to publish price lists.
Difference-in-differences, explained with one real example
When you can't run an experiment, compare how the group you changed moved against a group you didn't change. Here is how difference-in-differences works, using the study that made it famous, and when it goes wrong.
Can AI agents do senior expert work? What the evidence says
Studies now have experts grade AI work products against their own. They also show where AI breaks: open-ended judgment, choices of method, and the gap between how productive people feel and how productive they are.
What North Dakota wheat research must deliver to break even
At the new 20-mill checkoff, a year of North Dakota wheat research pays for itself if it adds about 0.06 bushels per acre. That is far too small to see in state yield data, which is why showing the return takes a different kind of evidence.
How to audit an AI system that produces analysis
If an AI system's analysis reaches a board, a regulator or a funder, someone will eventually ask whether it can be trusted. An audit answers that with evidence: baselines, tests of each part, traced numbers, and a score leadership can act on.
What can labor-saving orchard equipment cost and still pay?
In a Washington Gala orchard, each 10% of hand labor a machine displaces saves about $530 per acre a year. That supports about $2,500 per acre of equipment. Whether a machine pays then comes down to how many acres it covers.
Agentic AI for research and reporting: where it helps, and where it quietly fails
AI agents can now search, read, calculate and draft. The risk is not obvious nonsense; it is a polished report that leaves out the one comparison that mattered.
What is an estimand, and why does your program evaluation need one?
Most disappointing evaluations fail before any data is analysed, because nobody wrote down exactly what was being measured. The estimand fixes that.
How should a producer organization judge the return on its research?
Commodity commissions invest levy money in research whose payoff can take a decade or more to arrive. That long horizon shapes how research should be chosen, evaluated and governed.
Testing a policy before it exists: what agent-based models add
When a policy is still a proposal, there is nothing to evaluate. Agent-based simulation lets decision-makers see how individual choices and market prices might respond first.