Subscribe to the Teradata Blog

Get the latest industry news, technology trends, and data science insights each week.

I consent that Teradata Corporation, as provider of this website, may occasionally send me Teradata Marketing Communications emails with information regarding products, data analytics, and event and webinar invitations. I understand that I may unsubscribe at any time by following the unsubscribe link at the bottom of any email I receive.

Your privacy is important. Your personal information will be collected, stored, and processed in accordance with the Teradata Global Privacy Policy.

Ensuring Actionable Answers from Analytic Models

Ensuring Actionable Answers from Analytic Models
Teradata recently introduced Teradata Vantage, the platform for pervasive data intelligence. 100% of your relevant data, all the time, deployed anywhere – providing answers to what matters most to your business.

As a former Teradata customer, independent consultant and now Teradata employee, this is very exciting as the emphasis is on providing the answers required to achieve business outcomes. As we all know technology alone is never enough, rather success requires a balance of technological and human capabilities. The same thinking should be extended to analytics, whether it is a rules engine, machine learning, deep learning or artificial intelligence.   

The third-party report Analytics in Action says: “Success depends upon the cultural, not technological, evolution for your team. Judge progress based on the ability to learn specific business insights that continually improve ways of serving customers. For example, the conversations should not be about how to get data out of Hadoop, but how to better serve customers at store #123 who are vegetarian.

How can we (the humans) help ensure answers are practical and can be applied through taking business actions?

We first need to take the spotlight off the analytic models and put it directly on answering the business questions required to solve the business problem (or opportunity). Understanding the business problem as well as the key business questions provides a focus for the analytic models as well as helps to identify the data required to provide a practical, actionable answer.

With that focus, we can come up with the right analytical model, that will deliver an answer we were actually searching for.

But this is not enough: we also need to understand when it is better to provide guidance to the model to produce a more practical answer and when it is better to let the model have more freedom to seek the answer. In simple terms: you need to monitor the answers you get and be willing and ready to adjust the model or feed in additional information, if needed.

I would like to illustrate this with an example for each instance:

Focus on answering the business questions

We were asked to develop a model that helps gather insights into how the size of a case package should be configured. While the results delivered made sense, our customer did not get the insights they were looking for. Therefore, we determined we needed to feed the model additional business rules to produce a result that was more tailored and actionable by the specific customer.

Provide guidance to the model

We built a markdown optimization model. This model allowed the customer to understand how to maximise margin and sell-through percentage by optimizing markdown price. Initially the model had a minimum sell-through constraint fixed at 95% of inventory – to ensure that by the end of the season at least 95% of the product inventory would be sold. This meant that prices were continually lowered until that target was reached. But pricing could go to an extreme low of just $0.99 per item.

In this instance, the team recommended changing the sell-through constraint so that it was not fixed a percentage. This resulted in reaching 87% of stock clearance on average. At first glance this may seem as a worse result, however it was proven to be a better answer, since the model was able to balance the sale price to inventory clearance levels, so the overall profit margin of sales was increased, not decreased as might be expected.
When to provide a model more guidance and when to back off will vary depending on the specific question the analytical model is answering. It also can vary when the same question (e.g. what is the optimal markdown price for a specific product?) is asked by different stores. This is because of the variability between stores with regards to key factors such as starting stock position and sales volume.

It all comes down to where organizations put their priorities. Put the focus on technology and there may be business questions that cannot be answered as the required data, or analytic capability may not be available. Put the focus on the staff and then there may not be enough hours in the day to answer the priority questions. If the focus is on balancing technological and human capabilities to provide practical, actionable answers for business then they will be able raise their expectations as to what to expect from analytic solutions and rise above to solve the previously unsolvable business problems.
Portrait of Monica Woolmer

Monica Woolmer

As a Senior Business Consultant, Monica’s role is to help organizations answer key business questions through analytics. That is, to utilize her diverse experience across multiple industries to understand client's business and to identify opportunities to leverage analytics to achieve high-impact business outcomes.

With over 35 years of experience, Monica has been leading analytic solutions implementations for 23 years.  Prior to joining Teradata, Monica was the Managing Partner of Formation Data Pty Ltd, a Specialty data management, data warehousing and analytics consultancy in Australia. View all posts by Monica Woolmer

Turn your complex data and analytics into answers with Teradata Vantage.

Contact us