The CIO: Where the Money Goes

It is 8:30 on a Monday and Daniel, Northwind’s Chief Information Officer, is the last person to speak in a meeting already forty minutes old. Sales has blamed the product. Product has blamed the March price rise. Marketing has asked, twice, for a bigger discount budget. Everyone is confident, and no two of them are confident about the same thing.

Then the CEO turns to him. “You keep telling us we should be smarter with our data. You have $600k. Show me.” Daniel now owns the most consequential desk in this course, and the reason has nothing to do with technology. He will not choose a single tool this week. He will choose which capabilities exist for the rest of the project — and what everybody after him can and cannot do is about to be settled in a spreadsheet with seven rows in it.

What lands on Daniel’s desk

The board hands over less than it thinks. Revenue booked $11.3m last year and is tracking to $9.6m, a fall of about 15%. Churn — the share of customers who cancel in a year — has doubled from 9% to 18%, which on 400 accounts is about 36 extra departures, and at an average contract of $24k a year that is some $864k of recurring revenue which will not repeat. The ratio the board watches most closely, lifetime value against acquisition cost, has slid from about 7:1 to 4:1.

What does not land on his desk is any explanation. Nobody can name the accounts that left, or say whether they went over price, or the export feature that broke in March, or onboarding that quietly stopped working. Four theories, zero shared facts. The board has also fixed two constraints before Daniel says a word: $600k, and 26 weeks. Both are still binding in module 12 when someone asks whether it worked.

What a CIO actually does

People assume the CIO’s day is spent picking systems. In a well-run company it is closer to the opposite: the CIO translates a business problem — “revenue is falling” — into priorities before anyone writes a line of code, then defends those priorities against six executives who each arrived with a solution already in mind.

The core skill is telling a symptom from a root cause. Cancellations are the symptom; they are what you can see. The root cause is whatever produces them, and at Northwind that is genuinely unknown. This sounds academic until you notice what it does to spending: every proposal on the table — discounts, more salespeople, a product rebuild — is a solution to a cause somebody has assumed. Buying one means betting $600k on a hunch and finding out in month nine.

That last line is the one Daniel repeats all week: Northwind has not got worse at winning customers, it has got worse at keeping them — and you cannot discount your way out of that, because a discount lowers the $24k too.

The software on Daniel’s desk

None of Daniel’s software is clever, and that is the point. ServiceNow — an IT service management tool, essentially a shared queue with rules — matters because it puts every department’s request in one list. When the asks sit side by side, they compete on merit rather than on who caught the CIO at the coffee machine. The $1,100k of requests against $600k of budget is visible as a fact rather than a feeling.

The Power BI portfolio view supplies the context that makes a “no” defensible: what is already running, what it already costs, and which teams are already committed. Then the actual allocation happens in Excel, because it always does, and travels to the board as a PowerPoint slide, because funding decisions are made by people reading one page. If you take one thing from this desk, take that: the most consequential decision in the whole course is made in a spreadsheet and communicated in a slide.

The decisions Daniel makes this week

Deciding what the project is actually for

Before a penny moves, Daniel writes one sentence: the objective is not “reduce churn”, it is “find out which customers are leaving, why, and early enough to act — then act”. The target he attaches is churn back toward 9% within a year.

This looks like semantics and is not. “Reduce churn” is an outcome anyone can claim credit for; a good quarter would satisfy it. Daniel’s version names a mechanism, and a mechanism can be funded, tested and disproved. It also carries an assumption he writes down, because an unstated assumption is what kills a project quietly: we believe customers about to leave behave differently, visibly, in time to be saved. If that is false, the project should stop. Ben tests it in module 6, finding that leavers’ logins slide for about ten weeks before they cancel — and it is only because Daniel wrote the assumption down that anyone recognises the answer when it arrives.

Splitting $600k across $1.1m of asks

This is where opportunity cost stops being a textbook phrase. Every dollar Daniel puts in one pot is a dollar that is not in another, and because he is $500k short, funding anything properly means telling somebody no to their face. The temptation — enormous, because these are colleagues he sees at lunch — is to give everyone roughly 55% of what they asked for. Nobody storms out. Nobody is singled out.

And nothing works. Half a data platform is not half as useful as a data platform; it is pipelines that break on Tuesdays and a customer table two teams do not trust. An even split funds seven half-capabilities and buys none of them, which is why the sharpest thing to know about this desk is that an even split is not a compromise, it is a refusal to decide — and the refusal is paid for by whoever inherits the half-built thing, with no budget left to fix it.

So Daniel applies a harder rule: fund the smallest number of things to the level where they work, say no clearly to the rest, and write down what has to be true for the outcome to happen. Here is where he lands.

Data infrastructure, $175k, and AI and analytics, $150k, funded close to their asks because they are the only pots that answer the question why. Everything else on the list responds to an answer nobody has yet. Deliberately, this is the largest commitment in the split: $325k combined.

Product engineering, $100k. Not the $250k asked for — that was a roadmap wrapped around a bug fix — but enough to pay down the March release properly, including the export feature Corvex Manufacturing has raised 23 tickets about.

Customer success, $75k, half the ask: enough to add retention capacity, not enough to hire the way they wanted before anyone knows which accounts are worth calling.

Employee training, $50k, and governance, $50k. Small amounts, and Daniel protects them first rather than last. They are cheap insurance — and cheap insurance is always what gets trimmed at the end of a budget meeting, by people who will not be there when the claim is made.

Marketing incentives, $0.

Telling marketing no

The discount budget is the hardest conversation, because the case for it is genuinely good: it produces revenue this quarter, and this quarter is on fire. Daniel refuses it anyway, and his reasoning is worth borrowing.

A blanket discount cannot tell who it is talking to. It retains the customers who were leaving over price — the Fenwick Retails of the base, demanding 30% off — and hands the same money to everyone who was staying anyway, which is most of them. Subsidising the majority to save a minority is arithmetic that only works if you never do the arithmetic. Worse, it treats the symptom so well that it hides the cause: renewals tick up, and nobody learns why March broke.

What Daniel says to marketing is not “never” but “not until we know who” — a different sentence, and the one that keeps a colleague on side while still refusing them $200k.

Where this goes wrong

In real companies the failure is almost never a wild bet on the wrong technology. It is the polite version: the budget is spread evenly, and the CIO leaves the room having offended no one and committed to nothing. Six months later the data platform is unreliable, the model built on it is not trusted, the support team was never trained to read the output, and the governance review has no capacity, so it takes three weeks in the worst possible month. Each is a small underfunding. Together they are a failed project, and no single decision looks like the cause.

The second failure is quieter: choosing the technology first. Buy a well-reviewed retention platform in week two and every question after that gets answered in terms of what that platform happens to do. Northwind has chosen nothing yet — not indecision, but the correct order.

What Daniel hands on

The deliverable is a strategic mandate: the problem stated plainly, the target outcome (churn back toward 9%), the $600k split, the measures of success, and the written assumption that leavers can be spotted early enough to save. It goes to Omar, the Finance Business Partner, with one question attached — before we build anything, is this worth the money?

Watch what travels with it, because this is the rope, not the relay. The $325k of data and AI money is now the hard ceiling on the spending envelope Omar can approve in module 3; he can approve less, never more. The $50k of training money decides how many accounts Jade’s operations team can work through each week in module 11 — a ranked list of at-risk customers is worth nothing if nobody has been taught to read it. The $50k of governance money decides whether Elena’s risk review in module 9 is a one-week formality or a three-week scramble. And the $100k of product money decides whether, when Corvex Manufacturing’s broken export surfaces again in module 11, there is a real fix to escalate to — or only an apology.

Daniel has just done the most consequential thing in this entire project, and no technology was involved.

Spot the decision

Read each situation, decide what a good CIO does and why, then tap a card to flip it.