After our conversation, Ajay posted an intriguing blog post on PivotLink's blog entitled "Is Reporting Overrated?" that has already got a great comments thread running including James Taylor (his excellent blog here) and Jerome Pineau (his excellent blog here):
Late last week I had lunch with Nenshad Bardoliwalla, an ex-VP from SAP’s Business Intelligence group. He has written a great book on Corporate Performance Management . He said (and I paraphrase)” that users don’t really know what to look for in a report and that information is useless without context. So even if you gave all the required reports to a customer they wouldn’t know all the right things to look for.” His point was that the BI industry needs much more than reports. Customers need guidance on what to look for.If you've read this blog or Driven to Perform: Risk-Aware Performance Management From Strategy Through Execution, you'll know that reporting is one small part of a very rich set of capabilities needed to manage a business effectively: goal setting, risk management, compliance management, initiative management, planning, budgeting, and forecasting, predictive analytics, data mining, simulation and other types of modeling, and optimization.
But as to reporting specifically, for as long as I've been in this industry, and from what I can tell, as long as this industry has been around, we have not been able to get more than 20% of the users in an organization to use query, reporting, analysis, etc. tools despite continuous attempts. We've made the reporting tools significantly easier to use, with attractive options available from SAP BusinessObjects, Oracle, and IBM Cognos having been available for years. We now have a new generation of SaaS BI players like PivotLink as well as Birst, and Good Data that also do a credible job of providing the functionality that the on-premise vendors do, but with the significant TCO advantages that SaaS can provide, with a much lower time to implement, compelling UIs, and nowhere near the manageability headaches of their on-premise counterparts. Will these newer BI tools be able to break the barrier of increasing the adoption of BI technologies in the enterprise?
While I have good friends at every single vendor above and wish all of them nothing but the utmost success, unfortunately, I don't believe so. I think even vendors with very low costs and very compelling options like the aforementioned SaaS BI vendors will still hit the 20% adoption wall because although their reporting capabilities are excellent, the metaphor of the report is not what end user's want. Reports, the kind that have a grid with a lot of numbers, a number of dimensions, a few metrics, etc. require too much work to create and too much work to interpret. If I'm a sales manager in the middle of a task and I have to look at the sales pipeline report, I have to expend a lot of cognitive effort to figure out if the right filters have been applied, what this number means versus another, etc. It's a very specific metaphor whose rightful place is on the analysts' desk, not on the desktop of every user in the enterprise. Those end users definitely need the information that is contained within those reports, but delivering it in the report format is unlikely to work, no matter how easy it is to create. But, I readily acknowledge that it is too early to tell. I am an empiricist and always welcome contrarian opinions and more importantly data that refutes my hypotheses.
For those users in the enterprise who DO need reports, such as analysts, I think the SaaS BI offerings like PivotLink's and Good Data's are very compelling from a value perspective compared to their on-premise brethren, and all have their sweet spots in terms of differentiation. But that post is for another day!
Posts
Hey Nenshad,
ReplyDeleteInsightful post as usual. I agree that too often, reports are merely a static materialization of a given slice of information. As such, their usefulness is clearly limited as you point out. To me, one holy grail is about providing "point-of-need" information you can act on it in real time (tactical) - static reports don't support that easily. You need to have a timelined view and you need to have a mechanism to bring in information from disparate sources (much like instruments in a cockpit) with the ability to access other people and their own silos as needed. So in that sense being able to mash up diverse streams of information intelligently while sharing information with peers at all levels is key. It's not clear that one single application can deliver this efficiently at the moment BUT what is clear (to me) is that an analytic app need to "talk" to other systems seamlessly and facilitate collaboration among users.
This allows its reports (and whatever else the app happens to deliver) to take on a more multidimensional "aura". Reasons for the low adoption numbers you mention are numerous but IMHO a lot of them pertain to ease of use (as you describe) - Too many of these products are rocket science, users have to "wedge" and tweak them into what they really need -- because the developers know better of course :) -- and they don't play nice with the rest of the ecosystem. In a word, they're a pain in the ass to use!
In all those areas I believe SaaS can make a difference (but not simply by virtue of being cloud-based) and of course you can imagine which one, in my opinion, is already leading the charge in that endeavor :) - So yes reporting is a small part - Because it's the way BI is done & delivered we need to change.
Easier said than done for sure - stick around ;)
Nenshad, very valid points. What David Abdo, our CEO and I have been advocating has been BI tools for front-line employees, that actually help them do their job better and can impact the company's bottom line within the operational time-frame where it would do the most good.
ReplyDeleteThat involves, of course, monitoring, analysis and then action. Action, we think, has always been the missing link from BI tools. Obviously the store manager or the warehouse manager has to know what the heck he's looking at, but once he has access to something that is easy to understand and relevant to his or her role within the company, then we need something that would enable him or her to make the right decision, and even to kick-off a project for corrective action.
This, then, goes to your point of who is BI really for, analysts or operatons & sales folks? We think it's for the latter, we think that it should be put to practical use.
Just my .02 cents.
Fernando A. Labastida
KPIOnline
@Fernando - I would argue that BI needs to be "democratized" throughout the entire business. Actionable answers (and not just reports) are important for tactical and strategic decision-making as well. Analysts tend to be more strategic, your ops guys (and gals) are tactical and your sales folks are probably both. So in my opinion, there is no reason to restrict BI to one group or another - BI is really for everyone and as such should be delivered as a commodity and not something requiring special jargon or a PhD. This is the "right" way to make sure BI is put to practical use I believe.
ReplyDeleteJust my 2 cents.
@jerome I actually think we're on the same page - maybe I was extreme in advocating BI should be for the ops folks instead of analyists. Of course it should be democratized.What's exciting is to see how BI is coming down from it's ivory tower!
ReplyDeleteFernando A. Labastida
Nenshad; your post hits the current state of BI on the head.
ReplyDeleteMy past experience in dealing with BI adoption rate at clients; I would encounter the "E" word (excel); the fact that it required little IT intervention; and it suited what the operation folks needed to perform. But as we know this creates a silo approach and doesn't scale well when trying to integrate information from multiple applications. In many of these organizations these ops people also had access to BI solutions that they felt were to cumbersome to get the same information from.
One of the most vital pieces I always felt that was missing when asked about their BI app; and the reports that were generated; was what was the original question being asked here? Reports as you mention are static; numbers; results; but don't add the context of the question into the picture. In order for BI to right the ship; is to start thinking in terms of questions and answers; pretty much the same most of use deal with solutions to a problem. Its a matter of the context the answer is presented to the requester; and who has control over the original question and of course the information provided to answer the question.
Is SaaS BI going to be the answer to this contextual problem; I feel today the answer is "no"; but I also perceive its heading in the right direction. The provision to integrate on-line systems and external data sources, along with the ability to ask the question and modify it with ease will allow it to extend the adoption rate that dogs its on-premise sibling.
Its also the matter of finding the right segment along with problem to solve in order to gain user trust and adoption. But that conversation is for another post.
Hey Ryan, I think you're right on there. As a matter of fact it's also how I decided to look at various SaaS BI players in a past post on my blog - At the time I wanted to be able to ask questions in English actually, or as close to it as possible. I found GoodData's what/how/filter UX very compelling for this reason (as in, look ma: no SQL!) - I still think one of the most brilliant piece of UI ever designed is the Outlook rules generator (I copied it in numerous software apps myself). On the context angle, you're right, there's still plenty of room for progress there but IMO SaaS is better positioned to bring the pieces together (no pun intended) as opposed to monolithic on-premise apps. On that front we as vendors need to do a LOT of listening still - Now I'm up to a nickel I think :)
ReplyDeleteThis comment has been removed by the author.
ReplyDeleteNenshad
ReplyDeleteA very well described point of view. If I were to benchmark the BI industry's products against the simplicity of some of Google's products, Salesforce.com products and the iPhone platform then a lot of what you say makes a lot of sense.
Additionally a user experience that creates such a massive adoption is not a result of just UI alone but some serious rethinking of how things are architected under the covers. The mobile application space was fragmented with hundreds of vendors and a very complicated and proprietary stack that prevented adoption. Every company approached it more or less the same way - using synchronization. Apple rethought all of that and started from a clean slate.
BI needs a similar approach. We need to look at technologies and where they will be 3-5 years from now and seriously challenge ourselves. Does a BI deployment need dimensions, star schema, facts, hierarchies? If we were to start fresh how would things look like? What role can flash memory, Amazon EC2 like offerings, RESTful web services play? We at PivotLink (http://blog.pivotlink.com) are constantly thinking about these.
Check out this video by Sir James Dyson, who rethought the design of a vaccum cleaner and now a fan. A breakthrough design that is completely different from what has existed for hundreds of years - a fan with no blades.
http://www.fastcompany.tv/video/dyson-reinvents-fan
@Traveller - Your analogy of PivotLink's contribution to the world of BI by comparing it to a vacuum cleaner and a new improved fan hits a personal nerve.
ReplyDeleteIn my household, I am responsible for very few things domestic (bless my wife) besides vacuuming and I can promise you the day hasn't arrived yet when I'll plop 5 bills down for a sucking motor with a ball no matter how cutesy it looks :) I've done quite well for years with the cheaper Sears models. These Dysons are nothing more than a classic mousetrap with a pretty dress and significant marketing behind them - let's keep our senses there. I doubt you meant to compare PL's approach to this :)
But going back to BI, your question "Does a BI deployment need dimensions, star schema, facts, hierarchies?" leaves me a bit perplexed, especially as you refer to a "deployment" in this context. How do you guys at PivotLink envision BI and its deployment without these fundamentals? How about some concrete examples? It would surely be educational. Being iconoclastic is noble (IMHO) but where's the beef? :)