Analytics as Simple as Asking a Question
Note: This series of articles features technologies demonstrated at Summit Sneaks Night during Adobe Summit, the Digital Marketing conference. Sneaks demonstrations are a preview of the ideas and technologies percolating inside Adobe Labs, and should not be viewed as a representation of product plans or features.
http://blogs.adobe.com/conversations/files/2015/03/sneak-datatone.jpg
It’s a profound irony of digital marketing — the more data you have, the harder it is to make use of it. What’s more, as the speed of marketing accelerates you need to find the right data quickly to make informed decisions and reach customers in real time. Your business data loses value when it’s locked away, or routed through a single, overwhelmed gatekeeper in your organization.
That’s why Adobe is constantly thinking about how to make data analytics and access to marketing intelligence easy-to-use and accessible to everybody. Wouldn’t it be great if everybody in the marketing organization had instant access to an analytics expert? What if finding the right data point was as simple as asking a question out loud? It turns out the answers to those questions aren’t pipe dreams — they’re closer than you think.
As part of the Sneaks session at Adobe Summit, the Digital Marketing conference, Adobe Research Scientist Mira Dontcheva demonstrated a technology prototype called DataTone with the potential to revolutionize the way you access your business data. “We’re trying to make a more natural interface for working with the Marketing Cloud, whether it be through a browser, or through the phone,” she explains.
The DataTone project is a cross-lab initiative between the Creative Technologies Lab and the Imagination Lab in Adobe Research. It’s led by computer scientists Mira Dontcheva and Walter Chang in collaboration with Carnegie Mellon University, the University of Michigan and the University of Illinois, Urbana Champaign.
DataTone uses intelligent, personal assistant technology specifically developed around the needs and language of digital marketers. With it, marketers can access reports, look at the status of a campaign and navigate the Adobe Marketing Cloud — all via text queries in the browser or through a natural language interface on the mobile phone similar to Siri or Google Now. In other words, accessing your marketing information can be as simple as asking a question.
http://blogs.adobe.com/conversations/files/2015/03/datatone1.png
http://blogs.adobe.com/conversations/files/2015/03/datatone1.png
During the demonstration Mira navigated the Marketing Cloud using natural speech questions to access marketing data like revenue by marketing channel over time, and even complicated queries like show me how much revenue we got from the gold membership customers who subscribed to our newsletter and have purchased in the last 30 days.
“DataTone is able to automatically query this type of data and then visualize it appropriately,” Mira says. “For example, if I ask for revenue by channel, I might see a bar graph; but if I ask for that same data over time, it will generate a series of line charts that are more appropriate for that.”
http://blogs.adobe.com/conversations/files/2015/03/datatone3.png
The near instant results provided by DataTone make it seem simple, but in reality it’s a complex technology to master. “One of the hardest things about making a technology like DataTone work is that natural language is ambiguous,” she reveals. “It’s a problem that we, as computer scientists, have been trying to solve for many decades. But because we’re narrowing the scope of the problem to just digital marketing and to the language that marketers use, we’ve simplified the problem. It’s allowed us to make fast progress here.”
Although still in the concept stage, DataTone represents Adobe’s continued commitment to bringing more natural interaction to the Marketing Cloud, so that critical business data is easily accessible to all levels of the organization, especially for marketers on the go.
“We’re continuing to build out our prototypes,” Mira adds. “For us, what’s next is collecting a lot of data about the kinds of questions people ask. These sorts of systems learn over time, so we need real data to make these products work well, and incorporate that back into our algorithms. That’s our next step.”
See how _DataTone _works here:
Check out other sneaks previewed on the Summit stage: