Rafael Fanchini, CCO
(To download the pdf version, click here)
It's no wonder that many companies are struggling with how, when, and where to adopt the newer technologies of artificial intelligence, consisting of advanced analytics and machine learning, in particular. Let's define each in order to understand what their advantages are. Advanced, "augmented" analytics comprise a combination of mathematics, statistics, and computation to produce machine-learning algorithms that can help any organization process and consider all relevant data in order to optimize decision-making. Machine learning algorithms provide computers with the ability to identify hidden and often very complex performance-driver relationships within historical data. In turn, these analytic insights can be used to make predictions by considering current performance-driver values.
This technology is a key element of artificial intelligence (AI), and its use outperforms human intelligence very significantly on complex use-cases with large datasets, complex data forms, or very high-volume scenario comparisons. Machine learning is the key capability, for example, that has allowed companies like Google, Facebook, Amazon, and Uber to transform the way companies do business.
As the technology evolves, new forms of applications can be developed. Augmented analytics is the latest, most powerful, and user-friendly form of data analytics now available to businesses of all sizes. It provides optimal recommendations to decision-makers at the time a recurring decision is made.
Today, most organizations, large and small, still rely on the intuition of skilled and experienced individuals to make complicated decisions—most often with far too little analytic support to guide them. Yet, researchers at Google have proven that an analytically enabled decision process will outperform what they call the “ad hoc” process that it replaces by at least 10%. In reality, the analytic performance difference is often an advantage of nearly 20% in terms of higher revenue, increased profits, and better quality, etc.
Since machine learning and augmented analytics are complex disciplines to adopt, it is not surprising that the early adopters have been primarily technology-focused companies with deep pockets and distinguished science teams. In general, investments made by these companies have paid off in the long run, but only after efforts have been made to translate insights into actions. Early adopters have often been frustrated to find that valuable data-science insights do not automatically or even easily translate into meaningful business results.
There are a couple of reasons for this. The underlying data science is complex and hard to execute. Very often, data-science practitioners are specialists who do not have a complete understanding of how organizations behave. The typical solution has been for domain experts to team up with data scientists in order to overcome this problem. This approach has only partially resolved the problem, though, since from their side, the business-domain experts are not always well informed about the full set of recurring business decisions that can be automated and augmented with analytics.
Think of it as plumbing. Even if you live nearby an adequate water supply, you do not automatically have fresh water delivered to your faucet. You need plumbers to provide pipes to transport water from the source to your glass. The so-called “last mile” of analytics is the same. Delivering great insights to decision-makers at the time they need them is essential for the promise of analytics to be realized.
Seen from this perspective, it is not surprising that the early adopters have not been entirely satisfied with the impact that they have received from analytics. In fact, organizations of every size face the same challenge. They wonder, “How can I use analytics to optimize the important decisions that I make every day?"
We believe that advanced analytic microapps are an essential and significant tool for covering that last mile, bringing predictions and prescribed actions to the end-user in real time. Microapps are at the opposite end of the spectrum from the large, pre-packaged software platforms traditionally subscribed to by companies; these require large investments in time and resources to integrate, deploy, and maintain. Even more important, introducing new software into an existing workflow inevitably results in a lag time for adoption by end-users.
In contrast, microapps are focused on a single, important, recurring decision. The scope is clear. The implementation is far easier, and the output of the microapp can be used directly by the decision-maker. The insight-to-results gap is eliminated using microapps. Each microapp is designed to deliver optimal, best-practice, decision recommendations. Since the scope of each decision is limited, it is far easier to design the microapp so that it runs efficiently in real time. Likewise, adoption is easier, because the microapp fits into the formal or informal workflow of the decision-maker, and its intent and alignment with current processes make it easy to understand.
Microapps are designed to be “bite-sized” and focused—an easy-to-use tool for individual decisions. For example, we have developed a 30-minute demand-prediction microapp for quick-service restaurant (QSR) chains. This app is used to optimize when to start cooking or preparing food. This is an important factor for an individual QSR location that needs to harvest all of the demand that comes in the door. If customers see a long or slow-moving line, potential orders are lost. Likewise, food that is made too early may “time out” before it can be served. The results from this simple microapp are impressive. QSR customers report that their revenue increases by 2% to 6% when they use the 30-minute demand-prediction microapp to know when to start the cooking process.
Analytics2Go has built what we call The Power Plant, which seamlessly ties together all of the technology tools required to provide end-to-end SMART AI using microapps as one component. The Power Plant is built to optimize the effectiveness of microapps by continually processing real-time data feeds, automating algorithm selection, and monitoring algorithm-performance criteria. We source and maintain fully validated, external data for the businesses we support. And Analytics2Go maintains an automated analytics operating service for each client. Our client businesses can very easily expand their analytic capabilities by utilizing any or all of the Analytics2Go microapps driven by our Power Plant. So you can build your capabilities one business decision at a time.
Analytics2Go offers about a dozen microapps for the QSR industry. Our external data model includes all the data we recommend be used to operate any or all of the QSR-focused microapps. We also specify internal data models required from our clients (to be streamed to the Power Plant) to optimize the predictions and recommendations our microapps provide.