The make-vs-buy decision is a design decision.
Build gives you control and customisation. Buy gives you speed and lower initial cost.
The honest comparison uses total cost of ownership: build cost plus integration, maintenance, and the cost of being stuck with your own roadmap.
What are you building right now that you could buy or share?
Cost-benefit analysis (CBA) tells you whether total benefits exceed total costs. It doesn't tell you who bears the cost and who gets the benefit.
A service can have a positive cost-benefit ratio and still be regressive, charging the people who can least afford it to fund benefits that go elsewhere.
Distributional thinking adds the question CBA can't answer.
A public good is non-excludable (you can't stop people using it) and non-rivalrous (one person using it doesn't reduce it for others).
Street lighting. Disease surveillance. Navigation charts.
The market under-provides these. That's the original reason government exists: to provide what markets won't.
"We think we contributed, but we're not certain."
That's not a weak evaluation. That's an honest one.
The rigorous version of claiming impact has three checks: would it have happened anyway? Who did you measure? What else changed at the same time?
Vanity metrics feel good. They're easy to increase, hard to interpret, and rarely connected to whether the service is working.
The test for any metric: what decision would be different if this number went up?
If the answer is nothing, it might be a vanity metric.
The cobra effect: a policy designed to solve a problem that makes it worse by changing the incentives.
Every organisation has one.
Before you set a target, ask: how might someone reasonably game this? The answer often redesigns the metric before it causes harm.
Incentives shape behaviour more reliably than values or instructions do.
If the system rewards activity over outcomes, you'll get activity. If it rewards compliance over judgement, you'll get compliance.
The uncomfortable question: what does your system actually reward?
Most risk management is actually worst-case management.
We imagine the bad scenario and design to prevent it. Useful. But used alone, it makes organisations over-cautious and slow.
Expected value thinking asks a different question: across all plausible outcomes, what do we get on average? That's a different frame, and it often leads to a different decision.
What decision are you making based on worst case alone?
Little's Law in plain English: the more work you start, the slower everything gets.
Not because the team got worse. Because the system got fuller.
Reducing work in progress is often more effective than adding capacity. Most teams resist this, starting feels like progress. Finishing is what actually matters.
What's your team's WIP limit?
A moat in public services isn't about profit. It's about why your service can do something no one else can replicate at the same cost, quality, or reach.
Network effects. Switching costs. Data accumulated over time. Regulatory position. Most services have one of these. Few capitalise on it deliberately.
What's your service's sustainable advantage?
Simon Whatley