The Business Relevance of Analytics

The Business Relevance of Analytics

by Carlos Andre Reis PinheiroData Scientist, Teradata

Analytical modeling plays a key role in distinguishing companies within any marketplace by driving opportunities for competitive advancement. This competitive advantage enables companies to be more innovative, helping them stand ahead of the others in their market. Analytics helps identify when and what products to launch, informs which services will maintain customer loyalty, and optimizes product and service price, quality, and turnover. But perhaps even more important than being innovative is remaining innovation—and where analytics becomes crucial. Innovation means solving problems with simple solutions rather than complex ones, in an appropriate amount of time, and using a feasible replication process.
In order to be innovative, companies need to create a suitable environment to identify problems, threats, and opportunities; to create quick solutions that address the issues; and to deploy these solutions into the production chain.
If the solution is too complex to be readily understood, too expensive to be deployed into a production process, or if it takes too long to be developed, then this solution certainly isn’t innovative. It is just a solution, not feasible, not touchable, and not applicable to the company.

Be simple. Be fast. Be innovative.

Although easy to say, it isn’t as easy to do. Companies around the world are trying to do exactly this, although many aren’t succeeding. What’s the secret? Unfortunately, there is no secret. It’s like they say in the cartoon movie Kung Fu Panda. The secret of the amazing noodles was … there is no secret at all. It was all about the passion, love, and care put into making the noodles. The secret is in each one of us, who work with passion, love, respect, and who put all effort and energy into doing our best. The secret is to look at the past, to observe the present, and to try to foresee the future. So the secret is all about us, is it? Is it just a matter of looking at the past, observing the present, and foreseeing the future? Then why doesn’t this approach always work? It doesn’t always work because of the heuristic factor.
Even though we may proceed on a well-trodden path and follow a typical pattern, there are so many unpredictable variables in our world, each with so many attributes to be considered that even a small change in the overall environment might alter everything.

You can do everything exactly the right way, but perhaps not in the right time. You could do it in exactly the right time and the right way, but perhaps not considering all the variables involved. You can even do everything in the proper way, in the right time, but by not taking into account some small external factor (which is unknowingly completely relevant to the entire model) that may be unmapped, unpredictable, and untracked. This one factor may be something you would never consider being important to your model—it could be a natural event, a political fact, a social change, or an economic disruption. The same approach, by taking the same steps, performing exactly the same routines, would thrive perfectly if done a bit earlier, or even a bit later, but not right now. This is the imponderable! This is something you cannot predict. So what do you do? How can you manage this heuristic world? I have no other tip—but to keep trying to model the scenario at hand.

Organizations need to use the best tools they have available in order to adapt themselves to the scenarios, to both current and future business environments and to both current and pending changes. Analytics will help you to understand all these changes, all these business scenarios, for all corporate environments. Paradoxically, even when an analytical model fails, the analysis of that failure will help you understand why it happened, why your prediction didn’t materialize as expected, and, hopefully, what you would need to do to different next time. Even the failures help you reroute your strategy and drive your company toward the proper trail.
Being innovative is not the destination; it is the journey. In order to be innovative, companies need to steady themselves into this analytical path by monitoring modeled outcomes and improving models over time. Increasing the usage of these models and by converting this whole analytic environment into an operational process that exists across the enterprise drives access to innovation. The organization’s strategy should totally direct the analytical environment, and in turn, the analytical environment should totally support the company’s strategy.
This relationship between organizational strategy and the analytical environment can be envisioned in three distinct steps. These steps may be performed at different stages of analytics maturity—or may even occur simultaneously. Irrespective of timing, each of these steps resonates with different stages of applied analytic procedures, each of which are aimed at addressing specific business issues, and with particular goals.

(1) Stage One Analytics. This first layer of analytics provides long-term informational insight, helping organizations analyze trends and forecast business scenarios. Data warehouse, data marts, multidimensional applications (OLAP—On Line Analytical Processing), and interactive visual analysis usually support this stage one purpose.
These inputs support analyses geared toward identifying trends, historical event patterns, and business scenarios.
his analysis is concerned with presenting information about past sales by region, branches, products, and, of course, changes that have occurred over time. You can easily replace sales by any other business issue such as churn, subscription, claims, investments, accounts, and so on. Also, you can replace the dimensions region, branch, and product by any business factor you may be interest in analyzing. However, you can’t replace the dimension time, which should be always a consideration in this exploratory analytical approach. Often there is a production environment available that readily provides this kind of analytical information in a pre-defined environment, usually through a web portal.

(2) Stage Two Analytics. A second layer of analytics maps out the internal and external environments that impact the question at hand. This can include market considerations, the customers’ behavior and the competitor’s actions, as well as details about the products and services that the organization offers. Questions that are explored in this stage include: How profitable are my products/services? How well have they been adopted by the target audience?
How well do they suit the customer’s need? Statistical analyses support these tasks, with correlations, topic identification, and association statistics methods. Usually in these cases there is an analytical environment available to perform such queries and analyses. However, further distinguishing it from the first stage, there is typically no production environment that provides real-time answers, nor a predefined web portal or any other interface for rapid response to such questions. This stage of analysis is performed on demand, when business departments request deep information about a particular business issue.

(3) Stage Three Analytics. Finally, the third layer of analytics is driven by to the company’s strategy. Model development is directed by core business issues such as cross sell, up sell, churn, fraud, and risk, and models are also deployed and used once the results are derived. Data mining models that use artificial intelligence or statistics commonly support these types of endeavors, deploying supervised and unsupervised models to classify and predict some particular event and to recognize groups of similar behavior within the customer base for subscribed business change.
For example, let’s consider what analyses are required when a company decides to launch a new product.
Before establishing the proper packaging or price, the company may decide to run a deep study about the marketplace, the competitors, and the consumers. This study should take into consideration current customer needs: Are customers willing to adopt the product? What price might they be willing to pay for it? Do competitors have similar products in the market? If so, how much do they charge? Does the company have pre-existing products that compete with this new one?
All these questions might be addressed by using the second layer of analytics. This task is completely on demand, and it would be required to support the product launch.

A more in-depth analysis regarding how customers consume similar products, taking into account historical information about sales, might lead to a predictive model that establishes the likelihood of customers to purchase this new product. This predictive model would support sales campaigns, by defining target customers who have higher likelihoods of purchase, for example. This type of procedure would be associated with the third layer of analytics.
And finally, once that product has been launched, the company could monitor the sales success over the time. Business analysts might have a clear view about how well sales for the product occur in relation to different customer segments, different types of promotions, how profitable the product is in different branches, regions, sales channel, and so forth. This type of task is associated with the first layer of analytics, by delivering readily available insights through a web portal, published reports, interactive visualizations, and other ad hoc queries about that product across different business dimensions.

The entire analytical environment, including applications that support the three stages of analytics, should all relate to the organization’s strategy, cover all business issues, and be aligned with the company’s priorities. For a new company just starting out, a key objective might be to acquire as many customers as possible. In this scenario a customer acquisition dashboard should be deployed as part of the first stage of analytics, in order to monitor the changing size of the customer base. A market analysis that describes customers’ needs should be performed in the second stage, to understand which products and services must be launched. And a predictive model that targets acquisition strategies to the most appropriate prospects should be developed in the third stage of analytics.

On the other hand, if this company is well established and there are several other players emerging in their market, similar applications should be put in place but would be focused on monitoring different activities and events. In this situation the organization would want to monitor, understand, and predict churn (i.e., the rate at which customers leave the company), as well as predictive models that target cross and up-sell marketing activities. Nevertheless, the same three layers of analytics are used to adequately cover all the relevant business issues and organizational priorities.

Related Articles of Carlos Andre Reis Pinheiro:

Evolving Analytics

The Monty Hall Problem

Heuristics and Randomness in Analytics

Business Applications Based on Human Mobility Analysis and Travel Prediction

You may also like...