|

Considerations in Data Analytics Problem Structuring

By Fatema El-Wakeel, CMA
October 1, 2020
0 comments

Map out the complete structure of a project for a clearer picture of the path toward success.

 

Assessing the outcomes of data analytics projects, why is it that some projects are more successful than others? It comes down to understanding what’s being solved. A fully defined problem statement must be completed in advance of any significant project investment. A problem statement should clearly identify (1) who is sponsoring or requesting the project, (2) what problem it’s addressing, and (3) what the desired or target outcomes are. An effective problem-structuring process will dramatically improve project planning and performance, lending itself to ongoing refinement that may take place as part of an agile process.

 

Working in data analytics for years, I have witnessed problem structuring get far less attention than it deserves, which almost always leads to costly project remediation too late in the process. From my experience, this normally happens for a few reasons.

 

  • Technical team: The team members working on the project are very technical, so they want to dive deep into the project and start working in their area of expertise—for example, data engineering, modeling, or automation.

 

  • Problem-structuring skills: The skill set required to bridge and challenge both technical and business experts may not be available on the project. Problem-structuring skills are unique, involving an understanding of analytics and a willingness to challenge the business for a clear outline of what’s needed in the project.

 

  • Business culture: The business doesn’t want to invest time in a workshop, as management doesn’t understand the value of problem structuring. This is usually an education opportunity to explain the importance of problem structuring to the business. It’s imperative that the problem-structuring process play out and that key technical and business professionals don’t bypass it with a flawed set of assumptions.

 

PREREQUISITES FOR A PROJECT

 

Over many years, I have facilitated problem-structuring sessions for data analytics projects in both the private and not-for-profit sectors. Three key elements are necessary before giving a data analytics project the green light:

 

  1. Understand the real need: There’s a difference between a need and a want. The business might want a data analytics project to solve a problem. That problem might not be the root cause; it could be just a symptom. Several questions should be asked to understand the need. Examples follow:

 

What questions are being asked of the data analytics team by the C-suite, external customers, and other stakeholders?

 

What are the questions that we want to ask? Maybe team members don’t even know if they have the data available to answer the questions, so they’re avoiding them. This can also cover industry best practices or benchmarking if needed. It can even lead to a competitive advantage by identifying a unique selling point for the business sometimes if the discussion is steered in the right direction.

 

What are the areas of improvements in an ideal world, even in areas other than data analytics? Do we need to look into processes, standard operating procedures, or others?

 

  1. Understand the users: Now that we identified the need, we need to confirm that we have the right stakeholders in the room. Asking the right questions in a problem-structuring workshop can help with that:

 

Who is the visualization/model designed for? Who will use this piece of work? Is it for operational decision making, or is it strategic? There can be different levels of users: Board members and senior leadership will focus on strategic (quantitative or qualitative) tools or dashboards, while operational staff may be looking for key business metrics at a business line or functional level that represent the health and performance of the business. A subset of these metrics may be strategic in nature (and part of the executive scorecard).

 

How often is it used? A visualization that’s used by the directors to make regional decisions will be different than the visualization used daily by operational teams. The underlying data might be the same, but the visualization will be completely different.

 

What decision do we want the user to make and under what circumstances? Is it a model where we need to consider regional factors because we’re looking into the overall company strategic direction, or is it a model for a specific market where we want to understand specific patterns and market data?

 

What actions do we want people to take with this insight? Ask your users to specify the insight they’re after and what actions will be impacted. Are we considering actions in different departments? In this case, are those individuals involved in the project?

 

Data availability is crucial. The team may want to build a model to understand customer retention and loyalty programs compared to those of competitors. This will be a massive challenge, though, if the data to build this model isn’t available.

 

It’s important to look into data interactivity and where the calculations are needed. Sometimes the calculations can be easily done in the visualization tool, but other times it’s better to be done in the source, which would mean data engineering is involved. While you don’t want to overengineer a project, simplifying it to suit your interactivity and visualization presents its own challenge. Having a visualization workshop facilitates a clear understanding of how the users want to see the data.

 

  1. Understand the problem complexity. It’s important that both the analytics and the business team understand the complexity of the project, especially the time and skills needed.

 

Well-structured: This is a straightforward problem for which the team has confidence in the data. For example, creating a simple bot to automate a monthly process.

 

Semi-structured: In this scenario, the team doesn’t have all the data and/or the problem is new to the team. For example, marketing campaigns across regions when there’s a new product.

 

Ill-structured: This is when there is very little information to solve a problem. Critical problem-solving techniques (CPS) are usually ideal for those kind of problems. For example, political or economic unprecedented times such as Arab Spring, COVID-19, new regulations, etc.

 

Problem structuring is a science, and its methodical application is important for project success. Before you set off working with a project team, take time to get the stakeholders in a room and design a problem-structuring workshop. Time well spent now means less time wasted down the line.

 

Fatema El-Wakeel, CMA, has more than 10 years of experience in strategy development, data analytics, and financial planning in both multinationals and public sectors across the EMEA region. She is a member of the IMA Global Board of Directors and IMA’s Technology Solutions and Practices Committee. You can follow Fatema on Twitter, @fatemaelwakeel, or visit her blog, analyticsdrivenstrategy.com.
0 No Comments

You may also like