How to Get Started with Advanced Analytics
Advanced analytics can help retailers and their trading partners meet the
challenge of motivating consumers to engage in favorable shopping
behavior. By adopting business analytics, companies can develop targeted
marketing, tailored assortment and personalized shopping experiences to
drive growth and market share.
“It’s all about data. It’s about understanding where data comes from and what business processes are providing data,” says Colin Linsky, predictive analytics worldwide retail sector leader, IBM Business Analytics. “If there isn’t any data, there’s no ability to do analytics.”
In approaching analytics, executives need “to start thinking about what data exists throughout the organization and how does that relate to the business processes that we are trying to manage.” Linsky said three questions should be asked: How are we doing now? Why implement advanced analytics? And what should we be doing in the future?
He offered some suggestions to get started:
The business requirements for retail success have changed. Here are three major business areas that provide many opportunities to collect data, understand what is happening in the operations, while including analytics to make better decisions about how to perform operations in each area:
- Build smarter operations by updating systems to better handle customer demands, and improving management across channels of labor, assets and business processes
- Deliver a smarter shopping experience by enabling customers to shop however, whenever and wherever they want; and matching inventory and brand experience online, in stores and via mobile devices
- Develop smarter merchandising supply chains by gathering customer information continuously at every touch point, and managing and delivering assortments based on customer insights.
The integration of insights into decision making and processes by using advanced analytics begins with capturing data. There are many different sources of data in most companies: Internal like POS, and external like demographics. Rather than using just one or two, an advantage can be gained by mixing them up and making “a bit of a brew” out of all the different sources of available data.
After capturing and storing the data, the next step is to start building models. These models derive from different techniques and publicly available algorithms. They help companies understand what the relationship is between the different sources of data, and the particular outcomes that they are trying to model. This may involve trying to understand what are the best recommendations to make, or what could drive increased performance in a store. It may have something to do with the staff, the products, or the overall profitability. Each model is an asset that helps the understanding of patterns in the day, but also provides something to be used in deploying the analytics.
Deploying the analytics involves the familiar business processes in retail, such as planning, forecasting, and optimizing assortments. It’s never a single process. It never tends to be just the analytics providing the answer. There’s a lot of complexity in the retail environment, which means that models can give very strong indications of what is likely to happen next. Local knowledge — for example, knowing that something is going to be put on promotion — is very important in working with the output from the analytics.
So there is this process of iteration, where the forecasts will be made and the planners will start to bring their local knowledge and expertise into the process. It is a melding together of those hard-nosed analytics, plus the business experience of the professionals who work in these departments.
‘Capture, Predict and Act’
Doing analytics for analytics sake does no one any good. It’s a completely pointless exercise. Companies have got to find somewhere to deploy and act upon the analytics. The mantra becomes: “capture, predict and act.”
Usually companies will start to act in a business process. For example, on the planning side, feeding in the planning and budgeting process, from the top down or bottom up, and trying to reconcile the two.
It also could be on a website, using analytics to deploy the interaction between the customer and the catalog, or the different choices that they could make. Or the analytics could be deployed in customer service to try to get the best interaction with the customer, to try to make them happier with the service they are getting.
Reviewing the Analytics
Then there is a need to analyze if these interventions are working. There’s no point creating all of these analytics, deploying them into the real world, and then never noticing whether they are actually working and providing value.
At this point, companies need to understand how key performance indicators stack up against their plans, and close the feedback loop for continuous performance management. There’s a continual need to set up extra metrics to work out whether personal predictions are useful. Additionally, the more useful the predictions are, the more behavior changes. Therefore, we need to start considering that the analytics will cease to become as accurate. Then we can start to review the models, and get to understand whether the models should be refreshed or replaced by new ones. There is nothing static in the world, and that applies to analytics, as well.
“There are lots of opportunities to change the way that retailers do their business, and analytics definitely has a role to play in that space,” Linsky concluded.
The above article was abstracted from a presentation on “Getting Started with Advanced Analytics” by Colin Linsky, predictive analytics worldwide retail sector leader, IBM Business Analytics. He spoke during an IBM online conference on “Using Analytics to Understand Customers and Predict Buying Behavior.” The conference continues to be available online at http://events.unisfair.com/rt/ibm~retailanalytics.
More information on the topic and a contact form may be obtained at