CFOs are Taking over the Analytics Department

I was speaking to a CFO recently. They exclaimed:

This reflection hit me like a ton of bricks, because as a CFO I too took over the analytics department at my last company. It was originally built under engineering, but we got to a point where it would benefit the masses to have all our data in one place, across finance, product, and sales data.

We'd established that my org was the best location for it to sit in the company (IT was overly systems focused, Product wasn't financially focused enough).

And boy did I have a ton of imposter syndrome.

Yes, I enthusiastically put my hand up to say "Gimme that!" … but deep down inside I was thinking "How do I drive this car?"

In fact, I thought SQL was referring to the second God Father or Jaws movies.

But… I could def tell a story with data if you gave me enough rope. I could position the numbers within our company narrative.

The question became: how do both me and my team move up the analytics escalator?

The following is a guide for finance leaders who are either building out or taking over the analytics org at their company.

It's written from the perspective of someone who's an Excel guru, but does not have a degree in machine learning.

Here’s what we’ll cover today:

Analytics is not:

Speaking of that… dashboards are…basic.

Analytics:

Let’s board the analytics escalator.

First stop: Descriptive Analytics

Descriptive analytics lays the foundation by organizing historical data into clear, accessible formats like dashboards. This step is about tracking trends and spotting anomalies. But it doesn't explain why they occur.

For example, a dashboard might show that revenue is up 10% this quarter, with strong growth in the US and a decline in Europe. While this is interesting, it's only the first step.

Start small here. Pick one key metric that aligns with your team's goals: revenue, churn, or gross margin. Use tools like Tableau, Looker, or Power BI to build dashboards that highlight trends and comparisons over time.

But remember, this is not a book report or a school art project. Keep dashboards simple and intuitive. If an exec can't understand them in five seconds, they're too complicated.

Congrats! We’ve now taken the first step to “describe” something that is going on, complete with trends showing how we got here, and isolating one output metric so we can go deeper and diagnose.

Get your stethoscope, because that’s our second step.

Diagnostic analytics is all about going deeper to uncover the drivers behind the data. This step uses techniques like correlation analysis and segmentation to explain changes in performance.

We're going a level deeper than the simple dashboards we created before, trying to get upstream to see what levers were pulled in the organization to get to an output.

At this stage, the best advice is to focus on a single anomaly from your descriptive analysis, such as declining performance in a specific region or product line.

You’ll want to use segmentation to break down the data by key dimensions, like geography, customer type, or sales channel. Look for relationships between metrics (e.g., churn rates and product usage) to uncover actionable insights.

For instance, if sales in Europe are lagging, diagnostic analysis might reveal a connection to turnover among tenured sales reps. Or, if churn is decreasing, it could show that proactive outreach by customer success is driving retention.

Jesper explained how diagnostic analytics transforms raw data into meaningful insights:

We want to tell people “why”.

Another powerful approach is to lean on cluster analyses. You can see that sales people are traveling more, but are the ones who travel the most also selling the most?

Or maybe you create cluster charts depicting sales org tenure and performance, segmented by region. You see that you've lost a number of tenured sales professionals in Germany, and you've had to replace them with less experienced ones. As a result, they're not selling as much. Now you have a correlation (and a story to tell).

In terms of tooling and presentation: You can move past a simple Excel dashboard now to a data visualization tool like Tableau, Looker, or Power BI.

With this data, you're getting closer to predicting what will happen next.

Predictive analytics uses historical patterns and external factors to forecast future outcomes. This stage helps finance teams move from reactive to proactive planning.

Contrary to popular belief, predictive analytics is not super complex (or at least, doesn’t have to be). It's actually pretty straightforward. You need an algorithm, and that algorithm needs to be one you've proven can tell history. You've tested it on past data, so you feel confident it can also predict the future (within certain bounds).

Start by forecasting key metrics like revenue, churn, or expenses using simple tools like Excel's regression analysis or Anaplan. Then incorporate external data (market trends, customer sentiment, economic indicators) to enhance your predictions.

In my opinion, back testing is the most important step to get right before “passing go”… If it accurately predicts history, it's more likely to be reliable for the future. Not enough teams spend sufficient time backtesting.

Jesper described how predictive analytics sharpened his ability to challenge assumptions:

OK, now we're in the game of "What Should We Do?"

Prescriptive analytics is the most advanced stage. It doesn't just forecast your outcomes; it provides recommendations for improving them.

The goal is to develop clear "if-then" decision frameworks. For example: "If usage drops below X, then launch an engagement campaign." Or: "If customer health score falls below Y, then trigger a CSM intervention."

This requires the soft skills to partner with stakeholders across different departments (e.g., sales, customer success, product) to identify actionable recommendations based on your predictions. You're no longer just reporting numbers; you're influencing decisions.

This is where you start to invest in advanced tools like ThoughtSpot or DataRobot to scale prescriptive recommendations across teams.

Jesper called this stage the most impactful:

If you really want to advance your analytics agenda, you need access to an AI-enabled analytics tool that has the ability to sift through thousands of algorithms and determine which one best predicts your future outcomes. These tools can process a metric ton of variables (both internal data sitting in your data lake and external data from market sources) and identify the elements that are highly correlated to your performance.

Imagine a tool that runs through two million different data points and surfaces the three variables that matter most to your business. That's the power of prescriptive analytics at scale.

The reality for most finance teams is they’ve invested in the first layer of the stack. They have ERPs, planning and budgeting tools, and maybe data visualization tools. But the advanced analytics package? That's where the gap exists. And that's the reason why so many companies struggle to generate true insight.

And what they call "insight" often isn't actionable because it doesn't answer the "why," the "what might happen," or the "how to make it happen." True prescriptive analytics closes that loop.

Here's the thing about building an analytics function: you can't mature one dimension at a time. You need to evolve your people, process, and systems in parallel, making each more sophisticated as you climb the analytics escalator.

Too many CFOs make the mistake of thinking "I'll hire a data scientist, and they'll figure it out" or "I'll buy ThoughtSpot, and we'll be prescriptive overnight." That's like buying a Porsche 911 when you haven't learned to drive stick.

Let me show you how to sequence this properly.

People: At this stage, you don't need data scientists. You need analysts who understand the business and can build dashboards. Think: strong Excel skills, business acumen, and the ability to translate data into stories. One or two analysts can get you started.

Process: Focus on establishing a single source of truth. Define your key metrics as a company. What's revenue? How do we calculate churn? What’s the definition of a customer? Get everyone aligned on the vocabulary before you start building.

Systems: Your ERP and a data visualization tool (Tableau, Looker, Power BI) are sufficient. You're connecting systems, pulling data into one place, and creating basic dashboards. Keep it (stupid) simple.

People: Now you need to add technical depth. Bring in someone who can manage your data infrastructure. Brace yourself: this is where data engineers enter the picture. You'll also want to distinguish between:

The semantic layer is your delineation between the front door and the back door. The front of the house is where you make data real to people, talk about business outcomes, and link directly to what they're working on day-to-day. The back is where the sausage gets made: more technical than most people care (or need) to understand.

Process: This is where you establish governance. Who owns what data? How do we handle changes to our systems (like new Salesforce fields)? Create a communication protocol between your business systems team and your data team so nothing falls through the cracks. Lock down input fields on your sources of truth.

You're also starting to build repeatable diagnostic workflows. When a metric moves, what's your standard process for investigating why?

Systems: You're investing in your data warehouse and semantic layer. Tools like Snowflake or Databricks for storage, DBT for transformation. The goal is to have clean, structured data that your analysts can query without constantly bothering the engineers.

People: Here's where sequencing matters most. I've seen companies bring in data scientists too early, and it never works. You need the foundation first.

One CFO I know brought in a data scientist about a year into their analytics journey. They quickly realized they were trying to run before they could walk. The team wasn't ready for that skill set. So they pivoted: they changed that person's role to focus on being a senior analyst for their go-to-market teams. This person was extremely well-versed in data and could handle the day-to-day analyst work while also tackling more complex projects on the side, like building predictive churn models.

Don't hire for Level 4 skills when you're still building Level 2 capabilities. Grow your team's sophistication in parallel with your systems and processes. You will eventually get there, but you’ll set a new hire up for failure and disappointment if you overpromise what they will be working on day to day.

Process: You're now building model validation frameworks. Every forecast needs backtesting. Every prediction needs a confidence interval. Create a rhythm where your analytics team regularly validates models against actual outcomes and refines their approaches.

You're also establishing feedback loops with business teams. Sales might be optimistic in their forecasts, and your job is to pressure-test those assumptions with data and create a dialogue about closing the gap.

Systems: You're adding forecasting and modeling tools to your stack. Maybe it’s Anaplan for planning, perhaps Python or R for more sophisticated statistical analysis. You might start incorporating external data sources to enhance your predictions.

People: Your team now spans the full spectrum:

The key at this level is that your analytics team isn't just producing numbers; they're business consultants who help the business make better decisions. In fact, the soft skills of business partnering now mean more than ever.

Process: You're operating with "if-then" decision frameworks. When customer health scores drop below a threshold, the system automatically triggers interventions. When leading indicators shift, you're already modeling scenarios and recommending actions.

Your process also includes continuous improvement. You're tracking which recommendations were followed, what the outcomes were, and feeding that back into your models.

Systems: This is where AI-enabled analytics tools come in. You will want platforms that can sift through thousands of algorithms, process millions of data points from your data lake and external sources, and surface the three variables that actually matter to your performance.

But remember: you only get value from these advanced systems if you have the people who can interpret the results and the processes to act on them. I fully realize I, as the CFO, was not that person, at least not at the start.

Here's what progression actually looks like in practice:

Year 1: Hire 1-2 analysts. Build basic dashboards. Establish data governance. Get your data warehouse set up.

Year 2: Add a data engineer. Formalize your semantic layer. Create diagnostic workflows. Start simple forecasting.

Year 3: Bring in BI engineering capacity. Invest in modeling tools. Begin building predictive models. Validate, validate, validate.

Year 4+: Add advanced analytics capabilities. Implement AI-enabled tools. Scale prescriptive recommendations. But only if the foundation is solid.

Not to sound like a broken record, but the mistake most CFOs make is jumping straight to Year 4 investments without building Years 2 and 3. You end up with expensive tools that no one can use and data scientists who spend their time building dashboards because the infrastructure isn't there (and I’ve found they really hate that).

Your systems should be slightly ahead of your people's current skills (to encourage growth), but your people's skills should be slightly ahead of your processes (so they're not constrained). Keep all three moving forward together, and you'll climb the analytics escalator without falling off.

Remember that CFO from the beginning who felt like no one taught them what to do with their new analytics org? I’m still here! That imposter syndrome is real. And it's exactly why this escalator framework matters.

You don't need to know SQL or how to code data pipelines. You don't need a machine learning degree. What you need is a clear roadmap that starts where you are and moves you systematically toward insight, foresight, and better decisions.

Start with one dashboard. Find one anomaly. Build one predictive model. And keep your people, processes, and systems moving forward in lockstep. Before you know it, you won't just be the CFO who inherited the analytics org; you'll be the CFO who transformed how your company makes decisions (think the pros call it “data driven decision making”).

Don’t forget - that’s the whole point of getting on the escalator in the first place.

Hoping you don’t get your shoe laces stuck in the escalator,

CJ