Build a Strategic Analytics Muscle, Not a Dashboard Team
I have watched companies model promotional depth to two decimal places and then decide which market to enter based on a conversation in a hallway. Both decisions were made by serious people. Only one of them got any rigor applied to it, and it was the one worth a fraction as much.
I have written before that data teams are not service teams, and that what a business remembers is the experience of using data rather than the feed itself. Both of those were about how the function operates. This is about what the function is for, which turns out to be a harder question and the one that decides whether any of the rest matters.
The pattern I keep running into is not a tooling problem. Organizations have built real analytical discipline around a particular class of decisions, and that class is defined by convenience rather than importance. The decisions that get analyzed are the ones that repeat, that are reversible, and that generate clean feedback. Send time. Promotional depth. Channel mix. Creative variants. Pricing inside an established band. The value there is real, and every one of those decisions optimizes inside a strategy somebody else already set.
The decisions that determine where a company ends up look nothing like that. Which business are we in. Which customer do we build around. Do we own this capability or rent it. What do we stop doing. Where do we place a bet that will not pay off for three years. They happen once, they cannot be cheaply undone, and there is no historical analogue to model against. So they get made on instinct, by people who would never approve a promotion without a lift analysis. We have built rigor exactly where rigor is cheapest and left the expensive decisions to intuition. Closing that gap is what a strategic analytics muscle is, and almost nobody has one.
Reporting is not analytics
The harder version of this needs saying out loud, because it is what makes the rest possible. Building dashboards and reports is not analytics. It is measurement. It is necessary, it is often excellent work, and it is bookkeeping for the business.
Reporting tells you what happened. Analysis tells you what to do about it. A dashboard contains no argument. It presents numbers and leaves interpretation to whoever is looking, which means the interpretation varies by viewer and usually confirms what that viewer already believed. An analysis has a conclusion in it, and a named person accountable for that conclusion being right.
There is also a structural difference that shows up over time. A report is an artifact that has to be maintained forever. An analysis is an act that ends. Every dashboard in your environment is a question somebody had on a particular Tuesday, frozen in place and preserved long after the question stopped being interesting, and the accumulated weight of those frozen questions is what consumes the capacity of most analytics teams. I have seen functions where the large majority of effort goes to sustaining reporting nobody has made a decision from in two years, and all of it gets counted as analytics in every capacity conversation. It is not. It is infrastructure with a person attached. Until an organization is honest about that, it will keep concluding it needs more analysts when what it needs is fewer reports.
This is also why the standard maturity ladder is a poor diagnostic. Descriptive, diagnostic, predictive, prescriptive is a useful teaching device that measures technique rather than influence. A company can have production models running against a well governed lakehouse and still have no strategic analytics at all, because every one of those models is answering a question somebody else already decided was worth asking.
One caution on foundations, since I am often the person arguing hardest for them. Organizations get the plumbing working, the complaints stop, the dashboards finally reconcile, and the relief of that is so genuine it gets mistaken for arrival. A trustworthy data foundation is necessary and nowhere close to sufficient. It buys the function the right to be taken seriously, and most stop there because stopping there feels like winning.
Rigor for decisions that happen once
Strategic analytics is the practice of bringing discipline to decisions that only happen once. It does not look like a model, because there is no training data. It looks like structuring the problem, naming the assumptions the decision rests on, identifying which of those assumptions is both most uncertain and most load bearing, and finding evidence that moves the estimate on that one thing.
It is closer to how a good investor underwrites a deal than to how a data science team builds a classifier. The output is not a prediction. The output is a decision made with its reasoning exposed, so that when the world turns out differently you can tell which assumption failed and learn something durable rather than concluding that forecasting is hard.
Most executives already do a version of this in their heads. What they rarely have is anyone whose job is to make it explicit, write the assumptions down before the decision rather than after, and go find evidence on the one that matters most. That role is missing in most companies, and it is not the same role as the one producing the weekly numbers.
Experimentation is how you buy information
Experimentation is where the muscle gets built, and it is the piece most organizations underweight badly. Reporting looks backward at data that happened to you. Experimentation creates data that would not otherwise exist. It is the only mechanism available for establishing that one thing caused another rather than merely accompanied it, and a function that never runs experiments is permanently limited to explaining a past it had no hand in shaping.
The obvious objection is that you cannot experiment on a decision that happens once, and as stated that is true. You cannot A/B test whether to enter a market. But you can almost always decompose the decision into the assumptions holding it up, and a surprising number of those are testable cheaply and fast. The bet on a new segment rests on a belief about willingness to pay you can probe in a week. The bet on a new product rests on a demand assumption a landing page and a small spend will interrogate honestly. Find the assumption that is both most uncertain and most load bearing, then buy information on that one thing before committing everything. That is what an experimentation practice is for, and it is a much larger idea than testing subject lines.
The organizations that do this well have made experimentation a habit rather than a project. They run many small tests, accept that most return nothing interesting, and treat the portfolio rather than the individual result as the unit of value. The ones that do it badly run a single large test, get an ambiguous answer, decide experimentation does not work in their context, and go back to deciding by seniority.
Why this is now urgent
If the value proposition of an analytics function is turning questions into answers, that work is commoditizing quickly. Writing a query, assembling a chart, summarizing a table, explaining what moved. All of it is becoming ambient, available to everyone in the building without a ticket and without a wait. A function whose identity rests on being the place that produces answers will find that identity worth less every quarter, and the people running it will experience this as a slow loss of relevance they cannot quite explain.
What does not commoditize is knowing which question is worth asking and having the standing to put it on the agenda when nobody requested it. That requires context about the business that does not live in the data, judgment about what leadership is avoiding, and enough credibility to be believed when the answer is inconvenient. A strategic analytics muscle is not the advanced version of the analytics function. Increasingly it is the only version with a durable reason to exist.
Four questions
If you want to know where your own function sits, do not look at the tooling and do not look at the model inventory. Ask these instead.
What decision changed last quarter because of analytics. Not what insight was delivered, not what dashboard shipped, not how many requests were closed. A decision that would have gone one way and went another. If the honest answer is that the work informed decisions without changing any of them, you have a confirmation function.
How many analyses in the last year concluded the company should not do something. A function that only ever validates is either not being asked hard questions or not answering them honestly. Willingness to say no is the clearest signal that analysis sits upstream of the decision rather than downstream of it.
Is analytics in the room before strategy is set or only after. Analysis that arrives after a decision can only grade it. That is a real service and it is not the same thing as shaping direction, and the two get confused constantly because both produce documents that look similar.
How many experiments ran last quarter, and how many tested an assumption underneath a real decision rather than a variation on something already running. A function that generates no new evidence depends on evidence arriving by accident, and evidence that arrives by accident tends to arrive too late to matter.
The part that is yours
The rest of this is addressed to the person the analytics leader reports to, because almost none of it is inside the analytics leader's control.
The barrier here is not raw talent. There are capable people inside most analytics teams already and technology stopped being the constraint some time ago. The barrier is appetite for risk, and that appetite is set in the C-suite rather than in the data function. Reporting is institutionally safe and nobody gets criticized for an accurate dashboard. Having a thesis means being wrong in public occasionally and telling executives things they did not ask to hear about decisions they have already committed to. Your team has read that tradeoff correctly and adjusted, which is why you are getting reports.
Appetite alone does not produce capability. The muscle has to be built on purpose, because the skills involved are different in kind from the ones the function was originally hired for. Framing an ambiguous problem before anyone has agreed on the question. Decomposing a decision into the assumptions holding it up. Designing an experiment worth the cost of running it. Sitting with irreducible uncertainty and still making a recommendation. Defending a conclusion an executive does not want to hear. None of that is an advanced form of SQL, and a reporting team does not grow into it through encouragement. It happens when someone hires for that profile deliberately, develops the people already there who show the instinct, protects capacity so the work has somewhere to live, and is honest about fit when the fit is not there. That is a multi-year commitment rather than a reorganization, and it competes for budget against things with faster and more legible returns, which is exactly why it needs an executive sponsor rather than a champion.
Three things are in your hands and nobody else's. Bring the function into decisions before they are made, because when analysis is invited determines whether it can matter at all. Absorb the first few uncomfortable answers without consequence, because that is where the credibility of the invitation gets established. And notice what you do when an analysis concludes something you like is a bad idea, because everyone else is noticing.
None of this reaches you as a platform request or a headcount line, and that is why it usually does not reach you at all. It is a decision about whether your company wants its largest and least reversible choices examined before they are made, followed by the patience to hire and develop the people who can do that examining well. A dashboard team is what you get by default. A strategic analytics muscle is what you get on purpose.