Power BI, Embedded App Development, and the Problem of BI Disuse
Microsoft's new Power BI Embedded service gets at a significant issue for data management professionals: anemic BI uptake, what we could call BI disuse.
- By Steve Swoyer
- August 17, 2016
Whether you're building Windows, OS X, iOS, or Android apps -- apps for any platform, really -- Microsoft has something for you: an embedded Power BI service for Azure.
Last month, Redmond announced Power BI Embedded, a version of its Power BI service for Azure that exposes RESTful APIs. Developers can call Power BI Embedded APIs from their apps, effectively embedding business intelligence (BI) and analytics functionality -- reporting, charts, visualizations, and so on -- into non-BI or analytics apps. That's the idea, anyway.
The Value of Embedded BI
"Power BI Embedded ... gives ISVs the ability to create, deploy, and manage data-rich applications with ease," Nick Caldwell, Microsoft's Power BI general manager, wrote on the software giant's Power BI blog. "ISVs can bring compelling and interactive data visualizations to their customers. This empowers users to explore and engage with data visualizations which are built into their applications."
If anything, Caldwell is underselling the case. Thanks to the hype surrounding microservices, embedded app dev is a suddenly sexy technology subject. However, Microsoft's new Power BI Embedded service gets at an issue that has special salience for data management professionals: anemic BI uptake, or, to put it differently, BI disuse.
To find out what's at stake, let's briefly put Power BI Embedded aside to explore the fraught history of embedded BI, long touted (and never quite realized) as a means of exposing BI to the masses.
The failure of BI is in a sense a problem of insularity: BI is too often exposed and supported via dedicated front-end tools from vendors such as IBM, Information Builders, Microsoft, MicroStrategy, Oracle, or SAP. BI vendors have talked a good game about embedding BI functionality in operational apps -- call center apps, inventory management apps, and so on. In practice, however, they've failed at executing on this strategy.
The upshot is that when line-of-business consumers need BI information, they usually have to switch from the operational app in which they more or less live to a separate BI front-end tool with a very different user experience.
Not surprisingly, this has the effect of dampening enthusiasm for, and usage of, BI.
Most Enterprises See Only Moderate BI Impact
In her last "Successful BI Survey" prior to taking a job with Gartner, veteran industry watcher Cindi Howson summed up the state of enterprise BI this way: "The portion of customers saying BI has delivered significant impact declined ... to 28 percent and is the lowest since the survey began. [Although] there are some very successful and impactful BI deployments, the majority are stuck in the middle, with only slight to moderate success and business impact."
Part of the problem is that embedding BI is hard. It requires a commitment to cooperation and collaboration on the part of different groups in IT (IT management, software development, and data management), each of which has different priorities. It also requires ongoing cooperation and collaboration between IT and the line of business.
Traditionally, then, it's been very difficult to realize embedded BI in practice. In that same "Successful BI Survey," for example, Howson found that less than half (49 percent) of respondents had successfully executed on an embedded BI strategy.
This was in spite of her recommendation that "integrating BI with an operational application (embedded BI) or providing BI ... via a mobile app are two ways to improve BI adoption beyond power users."
That isn't all. Remember the BI tools vendors? Traditionally, they had little to no incentive to make their products more embeddable -- especially to the extent that embedded BI is effectively hidden BI.
Roadblocks for Embedded BI Products
The ideal embedded BI product is logo-less and bereft of distinctive features of any kind. In fact, the ideal embedded BI product is a service -- for example, reporting, charting, or visualization -- that can be embedded in and which appears indistinguishable from a front-line operational app.
Some BI vendors (such as Information Builders Inc.) embraced this logo-less, embedded model; most didn't -- at least not on the terms preferred by enterprise app developers.
There are both good and bad reasons why a market for such services failed to materialize, according to Mark Madsen, a research analyst with information management consultancy Third Nature.
"Historically, your data warehouse was in one place with its BI tools and stuff. Where are you trying to embed BI? Inside an operational app, which lives on a completely different system. So how do you shoehorn your MicroStrategy BI tool inside of SAP? In most cases, you don't," Madsen explains, describing this scenario as a "technical mismatch" that's long plagued BI efforts.
"If you can get past all of those things [technological and logistical issues] so that you can actually shove ... context-specific reports or visualization into [operational] apps, you have another kind of mismatch: the tools that were available often were not designed to be embedded, so you had logos popping up in places where you don't want them to appear. Even if you didn't have logos, you couldn't 'skin' [the report or BI attribute] to look like the app [in which you were embedding it]."
Power BI Embedded Offers Solutions
One indication that the embedded BI market is changing is the availability of technology such as Power BI Embedded, which Microsoft plans to offer globally across all Azure regions.
In the context of Azure, Power BI is a managed service; security and authentication (including row-level security), API versioning, data backup, and archiving are all handled by the Azure platform-as-a-service. In the same way, Azure supplies other technologies or services (SQL OLTP database, SQL data warehouse).
From a developer's perspective, the Power BI Embedded service abstracts what's going on in the background -- SQL queries to Microsoft's Azure SQL Data Warehouse, calls to other Azure services such as Azure Event Hub, Azure Stream Analytics, Azure Data Factory, Azure Catalog, Azure Data Lake, Azure HDInsight, etc.
If Power BI Embedded takes care of many technological issues -- and it does -- it also addresses important logistical issues, according to Microsoft.
Consider the problem of billing: what works in an on-premises environment (per-named user, per-processor, etc.) doesn't work in the cloud. Microsoft and other vendors are experimenting with different pricing strategies and mechanisms. Redmond seems to have devised a customer-friendly pricing scheme for Power BI Embedded.
"Starting September 1st, developers will no longer be billed per visual render but instead per report session. A report session is registered each time a user loads a report within the iframe, and it is good for 60 minutes. This means that no matter how many pages or visuals a report has or how many times a user clicks around that report, you will only be billed for one report session unless the entire report is reloaded," writes Josh Caplan, senior program manager with Microsoft's Power BI team.
New Paradigm Embraces Embedded BI
The emerging app-dev paradigm -- in which embedding is expected, if not mandated -- takes care of the rest, suggests Madsen.
There's also this: Microsoft's Power BI Embedded service, like similar cloud services, is in a process of continuous evolution: Redmond has promised to support interactivity between apps and embedded reports (via a new client-side API) as well as to support connectivity to on-premises data sources as the service evolves.
"Now pretty much all apps ... are being built with Web interfaces; they're making service calls to do functions, so now you can actually call the server and if the server can spit out a chunk of HTML or an iframe, then you can actually build it in," Madsen points out.
"I think the time is approaching where it will be a lot easier to embed this stuff. Certain of the problems don't go away, but there's a lot more potential."