A Practitioner's Guide to Self- Service BI and Analytics

Transcription

1 A Practitioner's Guide to Self- Service BI and Analytics What you need to know before you try or buy Publication Date: 12 Dec 2014 Product code: IT Surya Mukherjee

2 Summary Catalyst It is happy hour in the self-serve data visualization market for the business intelligence (BI) novice. Over the last two to three years, vendor interest has continued to peak in this segment, driven by the realization that there is a strong unmet demand for easy-to-use analytics solutions and considerable untapped budget sitting within lines of businesses. Ovum finds self-serve data visualization to be one of the fastest growing segments of the BI market, with all BI vendors falling over each other to claim their own share of the rapidly expanding data visualization pie. This report provides a comprehensive guide for BI practitioners to understand the market, navigate some of the available choices, and understand how to choose among vendors. This report tells practitioners of BI what to expect from self-service solutions. In upcoming reports, Ovum intends to touch upon self-service from the perspective of the ad hoc user, as well as the applicability of data visualization and self-serve to Big Data. Ovum view Data is enjoying a meteoric rise in profile among business users: analyzing it for outcome-improving insights consistently ranks as a top priority among enterprises globally. The technology required to do this has traditionally been focused on delivering capability to a small segment of the potential user base power users effectively capping its value. Intuitive accessibility has risen up the agenda and is all about breaking down the barriers to adopting and using data technologies. Visualization, which represents the natural way for humans to consume data, is emerging as the primary method to enable this transformation. Visualization is a loosely worded term; the front end of most self-service BI solutions is visual, but beneath the covers it encompasses federated querying across a wide variety of data sources. It also involves faster analytic architecture such as in-memory and advanced techniques such as flexible columnar compression to store data. The technology is easy to use because it effectively abstracts incredibly complex computations from the user. For most enterprises today, self-service and visual BI solutions cannot be a "yes we do" or "no we don't" discussion. Organizations that ignore the prevalence and virally expanding popularity of free-totry desktop analytic tools are doing so at their own peril; shadow IT usually thrives in the most locked down of IT environments. Instead, it is best to understand the pros and cons of these solutions and work self-service into the overall analytics strategy. Doing so not only helps prevent compliance mishaps but also helps organizations discover newer insights that could positively influence business metrics. Ovum believes the key to unlocking self-service BI apart from providing a visual approach to analytics is to, first, help end users access a range of clean and consistent data sources and, second, integrate them painlessly. This, of course, has profound implications for what, how, when, and where data is sourced. Self-service BI implementations that focus heavily on solving the complexity of data access and integration while incorporating governance and compliance are therefore more likely to gain sustainable enterprise adoption Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 2

3 Key messages For self-service BI to succeed, enterprises must be willing to change their data culture to one where sharing and collaborating within secure boundaries is encouraged and not penalized. Self-service is one of the primary enablers of agility being able to dissect data with minimal intervention from IT to empower line-of-business users. Self-service is not a magic bullet for tackling all enterprise data and analytics issues, and the ideal approach should be to supplement existing tools and databases with data visualization. Data management and governance become extremely important in a self-service paradigm, and given the nascent nature of this market, enterprises should be ready to deploy additional tools along with a self-serve solution to ensure compliance, at least for now. A code-free, intuitive interface with contextual analytics is the most important aspect of any self-serve BI tool. Collaboration in self-service tools should be in-process and in-context, which makes storyboarding an essential part of the overall experience. While visualization will be important, intuitive search will ultimately reach the most underserved of users. Recommendations Recommendations for enterprises Change IT culture to one of inclusion Self-service BI can be very fruitful for organizations across industries as it opens up a new world of data exploration, often leading to richer insights on existing and new data. However, like all new models, it needs to be supported by a change in culture as much as it needs a change in technology. For self-service BI to work effectively, IT departments must be open to creating an environment amenable to exploration. A great self-service environment will likely fail if it does not allow open and transparent access to rich data and content. On many occasions, such data lies inside the enterprise but is poorly used because people do not know about it or access is prevented by overly restrictive policies. Ovum advises IT departments to act as self-service enablers for business users by providing access to interesting curated data sets (from enterprise applications and the enterprise data warehouse) in server-side deployments of self-service tools. Where possible, assigning a quality score to such data sets could further help gain user trust. The logical next step is for vendors to provide features for users to rate or share their assessment or other useful insights about data sets themselves similar to offerings in the consumer world, such as Yelp. Tableau already has a feature that lets users do this. In making organizational data sets available quickly, IT not only ensures that data is properly governed and protected but also disincentivizes users from acquiring such data in unsecure flat file dumps, which most ad hoc BI users do. This is the same logic that the media industry uses to reduce piracy: Shortening the theater release to DVD/online release window of a movie actually reduces 2014 Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 3

4 piracy, as most people do not see the point of downloading a shoddy, barely legal copy of a movie when the original version is available in a short while for a small price. Separate test and production where possible A great way for IT to enable data sharing without losing control in self-service environments is to strongly segregate production and test environments (or server and desktop deployments). Data exploration is best done in a test environment, and analytical apps that prove useful can then be put into production. Big Data is an exception here, as data exploration might not be possible in a test environment with a large data set. However, for traditional BI with internal, known data sets, a data sandbox or something like that where business users can test-drive data sets works well. When IT is in control of a self-serve BI environment, it can use the test environment to offer a higher degree of analytic freedom with a higher degree of security and authentication. When apps come to production, IT can then look to limit the application capabilities to their intended usage. The culture of change works both ways. It is critical to remember that without IT involvement, selfservice cannot transcend becoming a siloed desktop analysis tool. It is also equally important for IT to realize that shadow IT systems exist and flourish in even the most locked-down of organizations. Instead of trying to respond to genuine analytic requirements with more padlocks on data, IT should look to use self-service as a great way to create analytic sandboxes for users. This does not mean a free-for-all enterprise data mart; instead it is about a culture of making a majority of organizational data available to the right (authenticated) users for experimentation in a controlled environment. Recognize the limitations of self-service BI The need for self-service BI has certainly been validated by the entry of all major BI vendors into this space. However, organizations should look at self-service BI in the context of overall enterprise information requirements and consider if self-service BI is indeed the answer to all of their information requirements. For example, data mashups/blending is an area where self-service capabilities may struggle to match complex enterprise requirements. Best practices on how to implement and use data and application mashups are still nascent in areas such as location intelligence. It is therefore important that organizations understand when not to use self-service tools for a particular job. For example, data mashups are highly applicable when there is a long backlog of end-user requests for new or modified reports, dashboards, and analytic applications that understaffed IT development staff struggle to meet, or if there is a mature federated data environment with limited data warehouses or data marts. Equally, there are certain applications and scenarios where it is not well suited, such as missioncritical operational BI applications that need to run smoothly 24x7, complex analytics queries that require large table scans or time-series processing, or when there is a need to minimize the impact of ad hoc queries against federated, transactional source systems. Similarly, reporting for compliance adherence and pixel-perfect reporting is likely not suited to self-service. Organizations should therefore consider self-service BI to be complementary to, rather than a complete replacement of, traditional enterprise BI tools for now. Create a strong governance environment for data and users Self-service BI can only be sustained with a strong governance program. Providing governance controls for the data sources being used to create data models, reports, and analyses is critical Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 4

5 Ensuring a clean, consistent, and up-to-date set of data sources is one of the most important responsibilities that enterprise IT departments must undertake before rolling out self-service BI capabilities to end users. This is not possible if business does not want to own or get involved with data quality to some extent. Equally, business users also need to be closely governed and monitored in a self-service BI environment. Organizations must ensure they are not defining "rogue" metrics and KPIs. Self-service BI always carries the risk of "overambitious" users submitting costly, long-running queries, which bog down query (and network) performance, at the expense of more important queries. They might also be running variations of BI reports and analysis that may already exist in the enterprise, effectively producing redundant BI content. Ovum suggests that organizations set up a governance committee that comprises a representative mix of IT, data management professionals, and business users to guard against "report clutter" by keeping regular checks and tabs on the number of reports being produced, particularly those that are redundant or have not been accessed and used over a long period of time. Apply a sliding scale of data quality There is no one-size-fits-all paradigm when talking about data quality, which opens up opportunities for both long-established BI vendors and pure-play self-service vendors to engage in this market. Ovum advises enterprises to apply a sliding scale of data quality when dealing with self-service initiatives. The sliding scale dictates that different measures of data quality (and therefore different types of data quality capabilities) are applied to different kinds of data. In other words, a sourceagnostic data quality initiative is not advised as the primary approach in self-service environments. Data quality need not be entirely visible, as non-expert users are unlikely to understand complex notions of data quality. Whatever happens under the covers, a simple, clear estimation of quality should be provided to the user. Figure 12 summarizes the sliding scale of data management in self-service environments Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 5

6 Figure 12: The sliding scale of data management decides treatment based on data type Source: Ovum Manage self-uploaded, semi-structured data from flat files natively Flat files are frequently acquired from organizational data in Microsoft Excel, CSV, or log file style dumps. They can also be acquired from static data dumps that originate from online sources such as Google AdWords; from cloud applications such as Eloqua and Salesforce; or from enterprise applications such as human capital management, financial management, or sales planning. While the structure of this data might not be known beforehand, there is an inherent structure to this kind of data in most cases, which makes it easy to model. However, since the data is inadequately formatted or not aligned to an analytics-friendly format, it is important that the self-service tool possess basic data quality measures to help the user transform and analyze the data. IBM Watson Analytics (or IBM Cognos Insight) and SAP Lumira provide a set of handy graphical tools to selfservice users for "wrangling" such data. For example, IBM Watson Analytics applies data-cleansing algorithms (under the covers) as soon as data is uploaded by users to identify common issues such as duplication, mismatched headers/keys, and how sparse the data is. However, the job of data quality does not stop at identification alone. More advanced features in the solution include tabular/spreadsheet views (which are a good way for humans to format structured data) which highlight semantic mismatch and allow the user to change the data so that it fits a particular schema/pattern. For example, a country name column in a spreadsheet could have both "USA" and "United States of America," which would create problems when aggregating these entries. The solution recognizes "USA" to be a loose synonym/acronym for a given country name and helps the user change all nonconforming entries to one standard at the data level with one operation Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 6

7 Figure 13: Watson Analytics automatically profiles uploaded data and points out interesting aspects Source: Watson Analytics, Ovum Depending on organizational maturity and requirements, Ovum believes that self-service solutions should allow organizations to build a rough ontology most relevant to themselves. Similarly, these solutions should help visualize data types, so that closely related entities that may need to be merged and standardized may be seen visually. At the end of a data load cycle, the tool should also be able to assign a data quality score to data sets, like the one shown in the Watson Analytics illustration below Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 7

8 Figure 14: IBM Watson data quality score Source: IBM, Ovum Task the enterprise data warehouse with sanitizing its own structured data As far as possible, cleansing and managing structured data should happen inside the EDW and not in the self-service BI solution. Expecting self-service users to cleanse EDW data will likely be counterproductive; Ovum believes that the onus of data quality lies with the EDW first. However, it is important to ensure that basic transformation steps are built into the self-service tool so that the end user is able to select and filter a cut of the data according to their analysis requirements. In some cases, this can be achieved inside the data connector itself. Most self-service vendors also allow IT to publish curated data sets on the server side of their platform for the use of self-service BI users. These data sets carry a data quality score which is regularly refreshed, which helps assure users that the data they are using are trustworthy. A key requirement for such data environments to work is that data are current and refresh cycles are frequent. Since users of self-serve BI platforms typically use live visualizations in presentations/storyboards, infrequent or unpredictable data refreshes might make analysis stale by the time new data are uploaded. Apply some native quality for structured data from enterprise applications Ovum believes that the onus of data quality in this case, similar to data acquired from an EDW, lies with the enterprise application. However, enterprise applications tend to be inherently transactional and not analytical; it is therefore important to ensure that somewhat advanced data transformation capabilities are built into the self-service tool so that the end user is able to modify the data according to their requirements. Such capabilities could involve data virtualization, where the source data is understood and mapped to the server side of self-serve BI tools before conducting any analysis. Organizations that invest in mapping source data well to their target platforms will clearly be at an advantage here, as a source-aware self-service BI tool (such as Tableau) could then run a federated query across mapped sources, without ever needing to pull the entire data set into its active memory Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 8

9 Data virtualization in this case helps aggregate and transform data at run time, which eliminates the need for an intermediate staging layer and data mart for each mapped source. Keep the source responsible for data hygiene for unstructured data Unstructured data is the final frontier for self-service BI and data visualization solutions. Unstructured data tends to be the most difficult to analyze, as it may or may not possess a known structure. Popular practices to deal with unstructured data involve running pattern-matching algorithms or visualizing the data to identify visual patterns without patterns or a semblance of pattern, it is not possible to understand the data at all to perform analysis. Given the inherent challenges in dealing with unstructured data, Ovum does not currently anticipate self-service BI vendors to be very active in this space. A more viable near-term route is to tie up with intermediaries that can refine and assign some structure to unstructured data before it lands in the self-service BI tool. This is where Ovum believes data preparation tools for Big Data exploratory analytics will pick up, and conventional selfservice BI tools, with their own ETL, will leave off. For example, unstructured text data responds well to quick search, which is one of the first steps in assigning some structure to it. Tableau ties up with MarkLogic, a NoSQL database, to support full text searches as well as complex searches on unstructured data that is stored in MarkLogic. For Tableau users, these searches appear to be in Tableau's flavor of SQL (VizQL), but the engine translates them to a format that MarkLogic understands. The heavy lifting, which involves actually indexing the data, forming metadata, and searching through full text, is done by MarkLogic. Indexing of raw data tends to be extremely resource intensive and Ovum believes that such high-overhead activities are best left to specialized analytic databases. Similarly, enriching or further filtering through searched data is again done by MarkLogic, which provides a semantic layer to provide customized data that enriches the data discovery process. SAP Lumira uses SAP HANA at the back end to achieve the same result. The ultimate goal of self-service solutions in this case should be to get the data set to a point where there is sufficient structure and/or lack of sparsity so that it can be visualized without exerting massive overhead on memory. While it is not necessary to fit all the data in-memory with most self-service solutions, there is a gradual/graceful performance degrade as the system starts processing large data sets which have active tables outside of memory. Unstructured data can also arrive in self-service solutions as flat files (or a native extract.tde for Tableau or.qvf for Qlik). For such cases, the treatment should be similar to self-uploaded, semistructured data. However, enterprises need to keep in mind that running visual operations on large, semi-structured data sets could have a very high overhead it is therefore advisable to either do this in server mode with ample analytic infrastructure, or to only work on representative samples of such data. Appendix Methodology The research is based on Ovum data and ongoing consultations with Ovum clients, discussions with industry vendors, and extensive scanning of online technical references Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 9

10 Further reading 2015 Trends to Watch: Business Intelligence and Enterprise Performance Management, IT (October 2014) Ovum Decision Matrix: Selecting a Business Intelligence Solution, , IT (July 2014) Author Surya Mukherjee, Senior Analyst, IT Information Management [email protected] Ovum Consulting We hope that this analysis will help you make informed and imaginative business decisions. If you have further requirements, Ovum s consulting team may be able to help you. For more information about Ovum s consulting capabilities, please contact us directly at [email protected]. Copyright notice and disclaimer The contents of this product are protected by international copyright laws, database rights and other intellectual property rights. The owner of these rights is Informa Telecoms and Media Limited, our affiliates or other third party licensors. All product and company names and logos contained within or appearing on this product are the trademarks, service marks or trading names of their respective owners, including Informa Telecoms and Media Limited. This product may not be copied, reproduced, distributed or transmitted in any form or by any means without the prior permission of Informa Telecoms and Media Limited. Whilst reasonable efforts have been made to ensure that the information and content of this product was correct as at the date of first publication, neither Informa Telecoms and Media Limited nor any person engaged or employed by Informa Telecoms and Media Limited accepts any liability for any errors, omissions or other inaccuracies. Readers should independently verify any facts and figures as no liability can be accepted in this regard readers assume full responsibility and risk accordingly for their use of such information and content. Any views and/or opinions expressed in this product by individual authors or contributors are their personal views and/or opinions and do not necessarily reflect the views and/or opinions of Informa Telecoms and Media Limited Ovum. All rights reserved. Unauthorized reproduction prohibited. Page 10

11 CONTACT US INTERNATIONAL OFFICES Beijing Dubai Hong Kong Hyderabad Johannesburg London Melbourne New York San Francisco Sao Paulo Tokyo