Going MAD: A Framework For Delivering Pervasive BI Solutions

Editor's Notes

• #2: I talk to a lot of BI professionals and most cite self-service BI as a benefit of investing in data warehouses and business intelligence tools. Their argument is logically flawless: BI empowers user to get the data they need directly from a data warehouse and create their own reports without having to wait forever for IT to build them for them. Self-service BI eliminates IT as an intermediary and frustrating bottleneck. Self service is a good thing, but we know in life that sometimes too much of a good thing can make you sick! Sometimes I wonder if self-service BI isn’t actually self-serving BI. Are we eager to embrace self-service BI because it truly meets the needs of our user community or because it offloads us from the unwanted work of creating custom reports? The answer to that question is the focus of this session.
• #4: To start this analysis, let’s take a short walk back through time. Before there were data warehouses, IT was in prison when it came to delivering information to business users for decision making purposes. IT created all the reports, usually by hand coding them in a 3GL or 4GL language. If an executive or manager wanted a slightly different view of the data, they would submit a request to IT and then wait, perhaps weeks or months before getting the custom report, which often didn’t provide the view the user was looking for or was no longer needed. In the eyes of the business, IT was slow, incompetent, and a very expensive cost center. No wonder many companies sought to outsource all of IT to third party providers. From the IT perspective, IT was in jail. But in some respects, users were in jail too, since they were held hostage by IT to get information to make decisions.
• #5: Then along came the data warehouse whose purpose was to create a repository of data designed exclusively for query and reporting purposes. The data warehouse was supposed to suck up all corporate data like a black hole and make it available directly to end-users. The data warehouse in essence was a playground in which users could ask any question they wanted or create any report they could imagine without bogging down the performance of operational systems or submitting requests to IT for custom reports.
• #6: And this was great news for IT. The data warehouse could liberate them once and for all from reporting prison and finally get on the good side of business users by giving them what they had sought for years: direct, unfettered access to data for query and reporting. Shortly thereafter, IT realized that data warehousing was also a full employment act and a great career path. So, by now in the mid-1990s, IT professionals were jumping for joy about data warehousing.
• #7: But things didn’t pan out as planned. While some users took the bait and ran with it, most still continued to request reports from IT. Running SQL directly against a DW proved very challenging for most people. In fact, impossible! And IT in some ways didn’t want to lose complete control of the reporting environment. So instead of reducing the report backlog, it got BIGGER!
• #9: JSo what did IT do next? Well, we decided if giving users the data wasn’t enough, we needed to give them better tools, self-service BI tools to be exact. These tools could enable them to access, analyze, and report on the data without IT intervention. And thus a new mantra was born. Self-service BI. It’s now the holy grail of BI professionals. Let’s empower the users to help themselves. ust as Marie Antoinette was completely oblivious to the needs of the French citizenry, so too are many BI professionals completely oblivious to the needs of their users when they recommend a tool that is impossible for them to use.
• #10: Ok, after one false start, now we’ve figured it out, and we are really happy now!! We’ve fixed this report backlog problem!
• #12: Let’s examine what happens when we give users unfettered access to data. First of all, the 20 percent of your users who had been bugging you for direct access to data embrace data warehousing and BI tools with gusto. These are typically business analysts or power users who are hungry for data and not intimidated by technology. They end up creating tens of thousands of reports and in doing so, waste a tremendous amount of their time creating and formatting reports for themselves and their colleagues. Some of them use the new BI tools to expedite the way they’ve always done things. That is they issue a humongous runaway query to download all the data they might ever need to stick in an Access database, predictive modeling tool, or spreadsheet where they perform their real analysis. These runaway queries bog down the performance of the whole system discouraging other users who suffer poor query response times. Most importantly, these users – although they are sourcing data from the same place – often create their own metrics, account structures, and definitions which they embed in their reports. Consequently, there is no consistency among these reports which confuses other users and executives seeking the final word on corporate performance in specific areas. Worse yet, the other 80% of your users – the casual users who look at reports and information periodically to do their jobs – get completely lost in the shuffle. They can’t find the report, the tools are too complex for them to remember how to use from one sitting to the next and the response times due to runaway queries are poor. Often, the reports they do find don’t exactly map to their processes but they don’ t know how to custom the view so they either stop using the tool or ask a colleague or IT to develop the view for them. And guess what, we are back to where we were before data warehousing ever appeared on the scene. So how did we get into this pickle? How did we spend millions of dollars on data warehousing and BI tools and infrastructure only to create shelfware and users unhappy about IT backlogs and lack of responsiveness? One common thread that I’ve noticed is that power users – those 15 to 20% of most user populations – wield way too much influence over the selection of BI tools. Once you let the chicken (i.e. power users) out of the henhouse, it’s hard to get them back in.
• #13: Survey respondents cited numerous impediments to usage, none of which stood out from the rest. Rather, BI teams looking to accelerate usage of BI tools face a plethora of challenges. The top two challenges, if I may paraphrase the answer choices here, are change management and data quality. The hardest thing about getting users to adopt a new tool is to get them to stop using their current tool. As one respondent wrote, “The only way to move users within our company to new BI tools is to eliminate their ability to rely on the crutches they currently use.” Lack of executive backing is also a challenge. If executives aren’t promoting and using the tools, all their subordinates won’t feel compelled to either. But the reverse is also true: if the executive uses the tools, by default, everyone else in the organization must as well. Ironically, data quality is as much a change management issue as a technical one, because users often reject new reports and BI tools that present data in a different format with different metrics or results even though the data is accurate. In an ironic twist, one BI manager said they knowingly published erroneous data that users had become used to seeing to ensure the adoption of a new tool. Obviously, we humans don’t deal with change, even if it’s for the better! My notion about tools being too complex registered in third place, while poor performance finished in a close fourth. Performance expectations among casual users are tough to satisfy. Says one BI director, “People won’t use a tool that doesn’t perform. Users expect to run a query and get results back within seconds. I call this the Google effect.”
• #14: The good news is that BI teams have quite a few levers they can pull to increase usage. Chief among these is integrating the BI environment with Microsoft Office. Since many users need to work in Excel to develop complex business plans and models and use PowerPoint to display results to managers and teams, it only makes sense to bring Microsoft Office more closely into the BI tool orbit. Many BI vendors have done just that in recent years, making these tools full-fledged clients to their BI applications. Second on the list, performance dashboards provide a highly interactive and intuitive interface that resonates with casual users and requires little or no training to use. When designed properly, a performance dashboard can replace a multitude of “legacy” reporting systems. We’ll talk more about how to design performance dashboards when we discuss the MAD framework. Many users also said that the best way to increase BI usage is to embed it into existing processes. And many respondents believe that delivering users a high degree of report interactivity and self-service will increase usage.
• #15: This sounds good and few would argue with it. But I see a couple of problems here: 1) Most users can’t create reports 2) users don’t really know what they want 3) and when they run into difficulty, they are still going to ask IT for help.
• #18: As you can see here, each leverage point is complex in its own right, consisting of multiple components and subcomponents, each of which can affect the usability of the system or the effectiveness of a project overall. Typically, BI teams need to address all the components within a leverage point to ensure a successful outcome. This is the equivalent of juggling multiple balls at once without dropping any. I can’t possibly review every element on this chart to show you how to deliver a more usable BI system, which will help create positive reinforcing cycle, that will lead to a successful and pervasive BI solution. But suffice to say, the four key components of usability, which are arrayed at the center of this diagram, are: good design, great user support, a solid architecture, and attention to change management. It is interesting to note that two of these four elements have little to do with technology and a lot to do with communications and people skills. Excellent technical skills are only ticket that get you in the BI game, and to succeed you need to master the soft skills of sales, marketing, education, and support. I drew much of this concept map based on responses to an open ended question on the survey that asked, “What strategies do you use to drive adoption of BI tools and increase usage?” So, most BI managers already understand the importance of the “soft stuff.” For the rest of this Webinar, I’d like to examine a single element in this chart. If you follow the path from the word Usability in the center to Design just above it, and then to Framework, you’ll see the word MAD. I suspect some of you might want to know what this is all about or if I’ve gone mad myself! By the way, the report goes into more detail on each of the elements displayed here.
• #21: So what is tailored delivery? Very simply, tailored delivery involves having a central group generate a standard set of interactive reports for specific domains of the business. These reports usually take the form of a dashboard, scorecard, or parameterized report that enables users to change a predefined view by selecting filters from a pick list or other graphical mechanism. A performance dashboard is the best. As a result, these reports often give users the impression that they performing ad hoc queries against all the data in the warehouse when, in reality, their access is circumscribed by a predefined set of metrics, attributes, and filters. The BI team needs to monitor usage and examine which elements are being used and which aren’t and perform periodic selective pruning to keep the reports relevant and fresh.
• #22: Actually, it’s more like 60% to 80% of the questions for casual users. A well designed interactive report can replace dozens if not hundreds of existing reports. Each report typically focuses on a specific domain and contains about 20 dimensions and 12 metrics. These standard reports are broad without being overwhelming because the present a simple, intuitive interface that makes it easy for users to apply filters to navigation and analyze the predefined data set. Once users navigate to a particularly useful view of information, they can subscribe to that view so it’s available with fresh data the next time they open their reporting tool.
• #23: This is the ideal framework to implement tailored delivery. It is the conceptual architecture for a performance dashboard. The pyramid represents layers of data. Graphical data is usually delivered via a dashboard/scorecard display or portal. Summarized data is delivered via an OLAP tool. Detailed data is delivered via a DW or OLTP systems. Each layer has an accompanying activity (Monitor, Analyze, Drill thru – hence MAD). Each level is used by all users, but each major group may start in a different place. Managers start at the graphical layer, analysts at the summarized layer, workers at the detail layer. KPIs at the top should be cascaded all the way to the bottom. The layers also represent the evolution of BI. In the 1980s, BI was just detailed data. In the 1990s, we added the analysis layer, and in the 2000s, we added the monitoring layer. Instead of delivering these separately, you should integrate all into a seamless, layered information delivery system called a performance dashboard. (See my book: Performance Dashboards: Measuring, Monitoring, and Managing Your Business – available online.)
• #24: Cisco has deployed this in marketing, HR, and now in finance. Beautiful!
• #25: Unfortunately, most companies turn the pyramid upside down! This is insane, not MAD! Systems that look like this overwhelm users with reports that contain all metrics and all data. And each layer is supported by a different tool so that users have to switch between tools to move from one layer to the next, and usually the data is not synchronized.
• #26: There are several adjacent applications that are gradually being integrated into the MAD sandbox. Planning today creates the goals and targets that comprise the conditional formatting of KPIs in the graphical layer. It also supports the what-if modeling to help organizations understand the ramifications of proposed actions. Wouldn’t it be great if the models were baked into the dashboard? And the plans were automatically integrated with the dashboard? Advanced analytics includes the predictive models that are being built into dashboards. Decide involves collaboration, such as annotating charts and tables and embedding workflow, which is common today in scorecards but not dashboards. Act involves closing the loop between decisions and updates, sending updates implemented via application APIs, stored procedures, or SQL. Many ERP vendors are creating composite applications that close the loop between analytical and operational applications.
• #29: Information Technology Department Understands process and requirements! Populates subject areas in the DW Develops the dashboard framework Develops the initial dashboard Trains power users to expand the dashboard Evaluates adherence to guidelines Select Power users – “SuperUsers” Expand dashboard with new views/reports Submit designs to IT for review Submit requests for new dashboards