Embedded BI puts analysis software inside ERP and other real-time transaction systems. But you still must ask the right questions and make the right decisions.
The really strategic, bet-the-company decisions by CEOs get most of the attention. But what matters even more are the ones frontline employees make every day--the operations, sales, marketing, customer service, and other decisions that determine if a grand plan flies or flops.
That's why companies need smarter enterprise applications--and why apps such as ERP and CRM that are already critical to companies will increasingly be used to deliver so-called embedded business intelligence and analytics.
Your company almost certainly has at least one BI system, and probably several. But is decision support delivered within the transactional interfaces that plant managers use to make production decisions? How about in the software that customer service reps use to guide interactions with each customer? "Executives can't just stand up at the front of a room and say, 'This is the new corporate strategy.' You have to make sure your organization executes on that strategy in every transaction," says decision management expert James Taylor, co-author of the influential book Smart (Enough) Systems (Prentice Hall, 2007).
Some of the midtier enterprise application players, including Microsoft, Lawson, and Epicor, have advanced the notion of embedding insight with their latest product releases. Market leaders SAP and Oracle are poised to raise the bar. With its move to acquire Sybase and in announcements made last week, SAP made it clear it intends to deliver real-time BI and advanced predictive analytics directly within its applications. It's even promising to eliminate the need for separate BI and information management infrastructure, a development that would dramatically lower IT costs.
Oracle, too, will make embedded BI a big part of its Fusion Applications, due later this year. And it's sure to tie in its high-powered Exadata appliances, which can process both transactional and analytic workloads. It's a different route to the same destination: real-time insight within the application.
Don't count out dedicated BI and analytics vendors. IBM, SAS, MicroStrategy, and others can make a strong argument that most companies have diverse applications, so companies need analysis tools that are agnostic to the data source and application.
There are two key tests for companies: Can they get better, timelier, and more usable intelligence from their apps and vendors? And can they drive operational decisions that are tied to their unique strategies? They can, though only if they carefully align those insights with their business goals and not depend on cookie-cutter dashboards and analytics.
Trouble In Boston
At Primo Water, a bottled water supplier that sells through retailers such as Albertsons, Kmart, Kroger, and Lowes Foods, decision insight is delivered through custom applications, including a Route View application built on top of Microsoft Dynamics GP ERP and Dynamics CRM. It combines order information from the ERP system with store-level sell-through data from customers. The combination gives Primo daily updates on orders and inventories.
As a result, Primo now constantly adjusts shipments in response to demand, and integrations with CRM let distributor partners and customers see when Primo plans to make its next shipment. Primo customers now often go long periods without running out of inventory. Even when unusual events lead to stockouts, the company can react faster--like when a major water main broke this month near Boston, forcing 2 million residents to boil or buy their water.
"Every retailer in the area ran out of water within hours," says Mick Gunter, Primo's senior VP of operations. "We were able to adjust our distribution routing on the fly because we could see that sell-through." Primo also adjusted its forecasts, which in turn triggered orders for more bottles, caps, and minerals the factory needed to ramp up production.
Primo, like many companies, is building intelligence into its business applications rather than asking users to "go do BI" in separate interfaces. You can use conventional BI systems to embed decision support within applications. But the promise of embedded BI is that by tapping both historical data and live transactional information, this new breed of up-to-the-moment insight will be delivered in the context of business processes, and it will be predictive and more actionable. Frontline workers can actually do something to seize opportunity or avoid bad outcomes.
Too good to be true? You're wise to be skeptical. This isn't the first time embedded BI has been touted as the way to get insight in front of more employees, since conventional BI has been too complicated and expensive to spread broadly. Also, the embedded BI capabilities that companies are delivering--with scores of pre-built reports, metrics, and dashboards packaged with these applications--won't serve your every need. You have to make sure the insights promote the right decisions.
Think of a call center application with built-in key performance indicators, including the inevitable measure of call duration. It's easy to measure that number and to create incentives for the call center manager to keep call times to a minimum. But what if your strategy is to be the premium-quality, high-margin provider? Should your service reps be rushing profitable customers off the phone, ones your marketing team spent a fortune to land?
Experienced users of embedded BI know better than to settle for simplistic, tactical insight. Gunter at Primo Water thinks of this as building "little data tunnels" into the application, and he emphasizes the importance of letting the business units call the shots and experiment with the available data. "We've done a lot by putting OLAP cubes together and encouraging our people to play with the data," he says. "That has helped them find interesting stats that we then make part of the next [custom] dashboard or report."
Hands On The Data
One complaint about embedded BI is that it's difficult to use data from outside the application. It's fine to be able to analyze ERP data inside the app, but the most powerful insights often come from blending data from multiple apps--including, increasingly, software-as-a-service apps. That's becoming more feasible.
Since 2007, ACTS Retirement-Life Communities has used the BI suite from Lawson software to draw insights from its Lawson S3 ERP system and combine it with separate occupancy and culinary/meal planning systems. ACTS has 23 retirement communities in eight states. The BI suite delivers drillable dashboards, alerts, and reports for finance, staffing, maintenance, housekeeping, and environmental managers--all data from inside the ERP app. But a separate, non-Lawson system tracks tenancy and vacancies in the various communities.
"That's what drives the bottom line," says Rob Yancey, director of business systems for ACTS, "and with the current housing market and the economy being in tough shape, our occupancy levels have taken a hit."
Over the last year, ACTS has used Lawson BI and data-integration capabilities to develop a one-day-latency dashboard of occupancy-related stats that's used by top corporate managers and the executive directors of each community. And in a wrinkle that may sound familiar, the company has resolved a "single version of the truth" discrepancy--in this case, on resale stats, which track new residents moving into an ACTS community.
"Executives look at resales on a daily basis, but marketing would have a different number than finance," Yancey says. The fix involved using Lawson's BI suite to add a "pending-customer" column to resale analyses and integrating data from yet another software system, a homegrown sales lead- and pipeline-tracking application.
Advantage In Analytics
Don't write off conventional BI suites, which can be used to embed insight into third-party interfaces using Web services or conventional APIs. In fact, standalone BI may be the only option for embedding if the application is entirely homegrown or if the analysis desired is sophisticated. SAS and IBM SPSS, for example, have customers who are embedding advanced predictive analytics into their CRM applications.
YouSee, a Danish digital TV and broadband provider, integrates customer-churn, cross-sell, and up-sell predictions generated from its SAS data mining and predictive analytics software with its Salesforce.com CRM data using Web services. Salesforce offers built-in BI query, analysis, and reporting features, but nothing approaching the SAS capabilities, says Anders Sorensen, YouSee's CRM executive.
YouSee developed customer-specific scores for more than 1 million subscribers based on deep analysis of transaction histories. The scores are then delivered through simple messages at the bottom of service reps' screens, such as, "This customer is likely to buy broadband." A nearby button calls up related sales scripts. "Our approach has been to make the analyses sophisticated on the inside but make it simple for the agents in the call center to put them to use," Sorensen says.
In a similar deployment, Sogecable, a digital TV and pay-per-view service provider in Spain, developed customer-churn and up-sell scores using data mining and predictive modeling tools from IBM SPSS. The scores for more than 2 million subscribers are embedded in the company's homegrown CRM system through color-coded alarms, so contact-center agents can see if a customer is a churn risk or cross-sell prospect.
"Conversion ratios for the cross-selling suggestions are very good, and sometimes they're even higher than similar outbound call campaigns," says Omar Rois Merino, a customer analysis manager at Sogecable. The main goal, he says, is to improve customer satisfaction with a customized experience. But here's a case where taking action on insight pays immediate dividends.
The advantage enterprise apps vendors are now promising that stand-alone BI vendors may find hard to match, except through custom projects, is the blend of historical data with real-time analysis of transaction information. App vendors say they'll also offer the ability to trigger events and take action within business processes. The biggest application vendors, SAP and Oracle, are so far only promising this capability. But if they deliver, it'll unite worlds that have long been separate.
SAP And Oracle Cometh
SAP this month announced plans to acquire Sybase, including its IQ column-store analytic database. By combining IQ with its own in-database technology, SAP says it will let customers explore real-time transactional information as well as historical data. SAP execs even go so far as to predict the elimination of the classic BI and information management steps of extract, transform, and load for data warehousing.
"Layers and layers of complexity exist simply in aggregating and organizing data until it can reach the decision maker," says Vishal Sikka, SAP's CTO. "All those layers can go away when you connect the decision maker directly to the transaction information sitting in in-memory systems."
Another jewel in the Sybase portfolio that may advance this trend is Aleri, the complex event processing vendor Sybase acquired in February. CEP is used to spot patterns in fast-moving data. Pioneered on financial trading floors, the technology is being adopted for telecom network monitoring, smart power grid management, Web clickstream analysis, and supply chains. CEP could also be used to quickly pass data from a transactional environment into an analytic environment. IBM, Oracle, and Microsoft also have CEP technology, with IBM furthest along in putting it into practice.
Oracle, too, is promising next-generation embedded BI capabilities as part of its Project Fusion application suite rewrite, which it has talked about for years but now says will arrive by year's end. What'll it have? When users visit their home pages within any Fusion application, they'll see a work list highlighting key things they need to do and know, based on their roles. Built-in Hyperion and Oracle BI technologies will let users drill down on the data behind exception conditions and do what-if analyses to determine the best course of action to address problems. If it delivers, it will blend the best of BI into the transactional application.
"We all bought these big systems with the thinking that we'd have these beautiful reports, but nobody ever looked at them," says Ray Wang, an Altimeter Group analyst. "Oracle got it right by starting with the end in mind, which is, 'What information do you need so that a person in a certain role can make a decision?'"
Some smaller vendors already are delivering. Epicor, for one, offers embedded BI in Epicor 9.05, released in April. By adding its Enterprise Performance Management suite to the ERP application platform, companies can mix real-time and historical analyses with rules and event triggers within the platform's business process management, or BPM, engine.
So, say a manufacturer's strategy dictates that every plant maintain a certain profit margin. This requirement could be monitored with KPIs set to specific thresholds, but without a direct link into the transactional environment, alerts wouldn't go off until unprofitable orders are a matter of history. With BI integrated with ERP, the process engine can detect problematic orders, fire off alerts to an approver, and channel the order to an exception queue. Yes, simple rules that don't allow products discounted more than X% can already be used to trigger exception handling. But embedded BI could allow automatic, up-to-the-minute analysis of current plant conditions--from capacities to the impact an order has on volume purchasing--that could affect borderline transactions.
Independent BPM software vendors will argue they've been blending BI, real-time monitoring, and process execution for years. That's true, but with the advent of next-generation service-oriented enterprise applications, vendors including Epicor, SAP, and Oracle now offer their own BPM tools, some of which are pre-integrated with application functionality, transactional data, and historical data.
Higher Standards, Same Goals
So get set to hear a lot more about BI embedded into your enterprise applications. But keep in mind that the basic advice on what tools to use probably still applies: If most of the data relevant to your queries and analyses is generated within one set of apps, then by all means consider the BI tools and capabilities available from that app vendor.
But if you have multiple, disparate applications and no dominant platform, or your company is adept at BI and has sophisticated requirements and expectations, you're likely to appreciate the flexibility and consistency of a standard BI toolset. If application vendors succeed in delivering real-time transactional insight with reduced information management infrastructure, it would be a game changer.
Whichever route you choose, the good news is the expectations are rising--to produce practical guidance within applications to help people make better decisions. But it's still up to business unit leaders, business analysts, and supporting IT executives, architects, and developers to ensure that the dials and dashboards are tuned to the big-picture business objectives.
Doug Henschen is editor in chief of IntelligentEnterprise.com. Write to him at email@example.com.
Part of an ongoing series assessing the future of the Internet. For more, check out internetevolution.com. Contributors to its ThinkerNet blog include David Silversmith, VP of IT at FirstBook.org, and cybercrime activist and analyst Jart Armin. Internet Evolution features industry-specific content in its IT Clan and Security Clan sections.
InformationWeek Tech Digest, Nov. 10, 2014Just 30% of respondents to our new survey say their companies are very or extremely effective at identifying critical data and analyzing it to make decisions, down from 42% in 2013. What gives?