RESEARCH & RESOURCES

Q&A: Embedded Analytics Draws Increased Interest

As the value of analytics becomes clearer, organizations are turning to embedded analytics -- tightly integrating analytics into a business process or application. That can help analytics become more pervasive, but it must be done right.

As the value of analytics becomes clearer and as analytics become easier to use, organizations are turning to embedded analytics -- tightly integrating analytics into a business process or application. That can help the technology become more pervasive, but it must be done right.

In this interview, we talk with Mark Gamble, senior director of technical marketing for OpenText Analytics, which acquired Gamble’s previous employer, Actuate, in January. Gamble discusses what the term means today, making the business case internally, and where companies can stumble in embracing embedded analytics.

TDWI: What do we mean by the term embedded analytics?

Mark Gamble: We often see embedded analytics as part of a Web or mobile application that. ...produces and shows summarized data to its users. A common example is showing bank customers their historic spending patterns on a banking website. The goal is to seamlessly insert analytics into the experience in context, in a way that is transparent to users of the application. We’re finding more and more that companies are looking to embed these capabilities into their existing apps rather than code it themselves from scratch or rip out their applications and replace them with all new efforts.

However, there is more to it than just that, and [TDWI Research Director] Fern Halper did a great job of underlining that [in a recent TDWI webinar with OpenText]. To expand on the somewhat simplistic notion I described, embedded analytics isn’t just pictures on a Web page. It also means including analytics within the business process behind the application. ... It should be a big-picture approach, not just slapping a chart on a Web page and calling it embedded analytics.

As interest grows around this whole idea, what types of embedded analytics are we seeing?

As customers, the most common types of embedded analytics that we all tend to see are charts or graphs that have to do with an account we have with a company, such as a bank. We also see analytics providing information in context with the application for the users -- and there are two flavors of this.

First, there’s the simplistic way -- just give me my information at a glance, with the important things called out to me as traditional alerts, then let me get on with my day.

A second flavor of embedded analytics is much more interactive. In that case, it might be a different use case than a customer. It might be an agent on behalf of a customer, with more flexibility to access data in an ad hoc fashion, but still through an in-context, tightly coupled experience, through their front-end application. It still removes the necessity of having to leave an application, go to another one, conduct your analysis out of context, then return to the previous experience, all while retaining the context.

A third flavor that I’ve seen -- and Fern emphasized this really well -- is folks who embed analytics into their processes. For example, they might create an analytic model that will help determine the success of a given campaign. It might mean identifying a target demographic -- or the traits and habits of this demographic -- then predicting future purchasing patterns and using this data to help drive campaigns to engage customers. Finally, you could measure the success of those campaigns through the same analytical process.

In this case, you aren’t necessarily embedding charts for your customers. Instead, you’re embedding analytics in your processes to help you understand how to engage with customers better. I think that final category is something that often gets lost -- too many people are still stuck on the idea that analytics needs to be a chart on a Web page. In reply to that, we like to say that embedding takes on many flavors -- it can happen within the process and on different devices. It’s one of the reasons that embedded analytics is something of a misunderstood animal these days.

What trends are you seeing regarding embedded analytics?

The second category in my previous response -- more interactive analytics -- is something of an emerging trend, especially from a customer-facing perspective. From a functional perspective, the trend is toward more interactive analytic content. Early adopters of the concept, who tended to embed static charts and graphs in their apps, faced a take-it-or-leave-it approach. Today’s users expect to be able to do more with what they see on screen. They want to be able to filter data, to change perspectives or context. This expectation, I think, is driving the trend of interactive embedded analytics. More and more companies and applications really strive to embed capabilities and content that will respond to user commands.

From a market perspective, another interesting trend I’m observing personally is a mad rush of old and new BI players to join the embedded analytics party. I’ve seen many of the larger BI players, who were more traditionally standalone, hastily adding integration APIs to allow for insertion of charts and graphs on Web pages. However, any of these vendors that assumes that is all that’s really required, I think, is going to fall short. True embedding requires more than just plugging in charts. It requires true integration throughout the app, from security to data to backend processes and more.

How are most companies embedding analytics today? What do you see with OpenText customers and potential customers?

Typically, what I’ve found is that the larger the application, the greater the need for seamless embedding. That becomes especially important when it comes to customer-facing apps. Most companies really start with the basics -- they start very simple. They insert charts and graphs with limited activity into their applications. However, I’ve also had companies who reserve increased functionality and offer it as a premium. In that case, the interactive features and ad hoc querying might be provided for a fee. Companies see it as a way to monetize the apps -- they are charging nominal fees, but for large companies, it can be quite lucrative. Several of our customers are following that model today.

What challenges do companies face in getting ready for embedded analytics?

It’s all about the data. I find this day in and day out when I’m involved in proofs of concept. The biggest challenge is to make sure that you have your house in order. Your data should be organized and understood before you begin any effort to try and depict it. Here’s a good example that: I was involved in a proof of concept recently in which we were given sample data. When we started building out charts and graphs, things just didn’t add up. We started seeing anomalies that shouldn’t have been there, or strange characterizations, or weird rollouts. When we put that same data into our big data analytics tool for data analysis, we instantly started to find data quality issues -- sometimes to the great surprise of our prospect, who thought they had everything organized.

I share that example to drive home the point that, in the rush to just slap charts on a Web page without considering the data infrastructure very carefully, you’re probably not going to have much success. Although there might be other challenges, I’d say first and foremost, understand your data.

How do you make the business case for embedded analytics?

It’s interesting -- just five or six months ago I would have said that step one is to get executive buy-in -- to convince your management that the effort of embedding analytics will pay dividends in customer satisfaction and loyalty and improved processes and so forth. However, recently, I’ve noticed a huge uptick in interest in embedded analytics. It’s been spurred by industry analysts, I think, along with a general acknowledgement that embedding is the best way to gain user adoption. I would propose that many organizations now understand the business value. The challenge then becomes to resist the urge to dive in too quickly.

The outlined, measured approach that Fern Halper presented in the TDWI webinar recently can help insure success. She laid out a great recipe -- consider business and process, as well as systems and integration and security and data. All of this needs to be part of an embedded strategy. Don’t rush and expect it to be successful.

With Actuate’s recent acquisition by OpenText Analytics, has your analytics strategy changed?

That’s a common question. First and foremost, we’re staying the course. With the Actuate acquisition, OpenText becomes a leader in embedded analytics. Our road map is continued growth of the embedded analytics technology within the market, with our sales force and sales support teams.

OpenText will also take advantage of Actuate by embedding analytics in the rest of its product stack. There are aggressive projects underway to introduce or enhance analytics within the OpenText Information Exchange Suite, and so forth. I think it’s going to be a two-pronged approach -- let’s tackle the BI and embedded analytics market directly, just as Actuate did. However, we’ll also absorb Actuate’s capabilities into the rest of the product stack, so that ultimately it becomes pervasive throughout the OpenText product line.

TDWI Membership

Get immediate access to training discounts, video library, BI Teams, Skills, Budget Report, and more

Individual, Student, & Team memberships available.