Du descriptif au prescriptif : les différentes étapes du modèle de maturité analytique Alteryx

Strategy   |   Rachel Kast   |   Aug 30, 2024 TIME TO READ: 9 MINS
TIME TO READ: 9 MINS

As an analytic consultant at Alteryx, I work daily with customers who are eager to gain an edge with their data. However, when customers approach us, many want to jump straight into the biggest, buzziest trend of the day, whether that’s machine learning (ML), large language models (LLMs), or genAI.

And it makes sense why. It seems like everyone is getting into these technologies — and many are reaping huge rewards.

Unfortunately, you can’t just go from zero to one hundred, and many companies have limited analytics maturity. Jumping straight into advanced use cases isn’t always wise, practical, or even possible.

Analytics maturity exists along a continuum; different competencies are needed at each step. In this blog, I break down the different stages of analytics maturity and explain how the products within the Alteryx platform align with each stage.

What is analytics maturity, and what are the different stages? 

In his watershed book, The Five Stages of Analytics Maturity, Tom Davenport bucketed analytics maturity into five stages: Analytic Beginner, Localized Analytics, Analytic Aspirations, Analytic Company, and Analytic Competitors.

While this is a helpful framework, I prefer to look at analytics maturity in a more practical and actionable way, so I stratify the analytics maturity continuum by capability.

  1. Descriptive: What happened?
  2. Diagnostic: Why did it happen?
  3. Predictive: What’s going to happen?
  4. Prescriptive: What action should we take?

Each stage builds upon the previous and helps you gain incremental insights from your data. However, here’s a secret underlying the entire analytics maturity continuum: Analytics (and AI, for that matter) won’t help you unless you have high-quality data. In addition, I would also add an “unofficial” stage zero to this continuum: automation.

While automation isn’t an analytic capability per se, every piece of the analytics lifecycle can benefit from some degree of automation. Tasks that consume your analysts’ and data scientists’ time can be finished in seconds, freeing up valuable time for more strategic work. With that said, let’s take a closer look at each stage of the analytics maturity continuum and how Alteryx can help.

The Alteryx Analytics Maturity Continuum 

Stage zero: Automation 

What it looks like: Automation can enhance repeatable processes, such as aggregating data, running analytics workflows, creating reports, or even running advanced machine learning or AI models. Its capabilities serve as a strong foundation throughout the analytics maturity continuum.

Benefits: Time-savings, cost-savings, and fewer manual errors.

Alteryx product: Designer Desktop, our flagship product, is a great analytics solution for automation. Designer offers self-service data preparation and no-code, low-code modeling, helping you automate your data preparation, model building, and virtually any analytics workflow imaginable.

DoorDash saves 25,000 hours with Alteryx automation

Like many organizations, the accounting team at DoorDash was dealing with a growing volume and complexity of data combined with intense pressure to speed up their analytics processes. With Alteryx, they were able to build hundreds of workflows, saving 25,000 hours annually and realizing millions in ROI.

Read Their Story

Stage one: Descriptive

What it looks like: I’ve had several retail customers with mountains of order data. They usually have a gut feeling about when their busy season is or the big drivers causing fluctuations in orders. They use these gut feelings to drive decision making, but they’ve never tested them in the data. In one case, when we looked at their data in Alteryx, we found that their theory about how the seasonality of their part-order system works wasn’t true at all — and they were floored.

Descriptive analytics allow you to spot patterns and trends in your data to answer “What happened?”

Benefits: Understand key trends and patterns in your data.

While the past doesn’t dictate the future, it certainly informs it..

Alteryx product: Alteryx Designer is the perfect solution for descriptive analytics. Our Data Investigation tools, Time Series tools, and Summarize Tool all help you quickly understand what happened in the past.

Stage two: Diagnostic

What it looks like: If descriptive answers “what happened?” then diagnostic analytics answers why it happened. This stage helps you understand the underlying drivers of important events, such as which factors most impacted your sales, whether that’s seasonality, new contracts, specific part numbers, etc.

In the early stages of Diagnostic, you may have only one or two employees looking at your data. As you mature through stage two, you’ll end up with multiple stakeholders across the org leveraging a single source of truth for their decision-making.

Benefits: Pinpoint key drivers behind important events to improve decision-making and set the stage for predictive modeling.

Alteryx product: Alteryx Auto Insights uses machine learning to automatically surface the important patterns and trends in your data and explain it in clear language —  and it does it all in minutes.

Newport ONE boosts revenue by 30% with Alteryx Auto Insights

Newport ONE, a full-service fundraising agency for nonprofits, wanted to maximize the ROI of its direct response campaigns for its clients. With Alteryx Auto insights, the team at Newport ONE could automatically surface insights to improve campaigns. “With Auto Insights, we’re improving ROI by targeting the right people,” said Collin Ward, chief innovation officer at Newport ONE. “It’s pointing to key segments that are overperforming, and we’re able to feed [those] into a model and improve campaign performance by 20-30%.”

Read Their Story

Stage three: Predictive

What it looks like: Once you understand your key drivers, you can use those as inputs for your machine-learning models. Predictive models will help you answer: “What will happen?”

For example, perhaps you’ve identified a KPI you want to forecast, like sales by the end of Q4. If you know your key drivers for sales are marketing campaigns and product features, then you can input those into a model and predict total sales.

At the early stages of Predictive, you’ll start building and experimenting with models while learning what makes a model truly effective. As you become more proficient in stage three, you’ll understand that your real-world data is going to differ from picture-perfect models. Finally, you’ll also be performing modeling monitoring to look for changes in accuracy over time (model drift).

Benefits: The benefits of predictive modeling depend on your use case. For example, you can predict customer churn, classify data (such as fraud/not fraud), and even make predictions based on multiple variables and test the impact of certain variables. In general, by understanding what will happen, you can take preemptive action to mitigate risks and maximize results.

Alteryx product: Alteryx Machine Learning (our AutoML solution) and Alteryx Designer both have robust predictive modeling tools. Alteryx Machine Learning has many automated features (like automated feature engineering and model construction), while Designer has a library of low-code, no-code R-based tools for more hands-on, customizable predictive analytics and model monitoring.

Carnival Cruise Line evaluates 10,000 vendors in under a minute

The Fraud Investigation team at Carnival wanted to see if there were any high-risk vendors across the company. Rather than creating a manual process, they decided they needed something repeatable. The team built two Alteryx workflows, one to automate the manual testing performed on vendors and another that uses k-means clustering to identify risky vendors. In one year, they were able to build their workflow and review 10,000 “non-PO” vendors, giving each vendor a risk ranking based upon 275,000 invoices that represented over $2B in spending.

Read Their Story

Stage four: Prescriptive

What it looks like: Finally, prescriptive analytics prescribes what you should do. For example, returning to the customer churn model, if you can predict which customers will churn, you can then prescribe the best actions to retain that customer.

Benefits: Make informed decisions about the best course of action to reduce risks and costs and gain a competitive advantage.

Alteryx product: Alteryx Machine Learning has simulation functionality built-in to help you easily pinpoint the best action to take. In addition, Machine Learning provides a weighted score for each feature, and if you adjust a feature on or off or up or down, that changes the weighting — which can change the prediction and help you understand the best action to take.

McLaren runs 30M race simulations in minutes

Just one race weekend in the Formula 1 calendar generates 1.5TB of data. With over 300 telemetry sensors on each race car, McLaren Racing has plenty of data. However, they needed a way to maximize its value. With Alteryx, they can run over 30M race simulations to test how every race will play out. “Alteryx allows the fast combination and correlation of those data sets so the teams can focus on what changes they can make that will improve performance iteratively across the course of the season,” said Edward Green, head of commercial technology at McLaren Racing.

Read Their Story

Accelerate your analytics maturity

Analytics maturity is a worthy investment, and although it takes time to mature, there are actions you can take to fast-track your success. For example, you can attend conferences, host in-office hackathons or analytics days, and even build an internal center of excellence/enablement (COE).

Increasing Analytics Maturity

However, there are even more effective ways to move the needle, like working with experts who have gone through this transformation hundreds of times. I can’t recommend our

The second surefire step I’d recommend is our Maveryx Academy, which provides Alteryx lessons and Alteryx Certifications. Not only are the learning resources incredibly valuable, but the Certifications can help you effectively upskill your workforce while giving your data workers an edge in their careers (and greater motivation to stay with your company). Finally, the Maveryx Community is a great resource. You can scour the forum to see how other data teams are solving their toughest problems and pose your own questions to a community of over 400,000 members.

Additional resources

Learn more about success bundles, or to purchase one, contact your Alteryx Account Executive.

Try Alteryx Designer free for 30 days.

 

Tags