May 22, 2013

What role desktop BI plays?

All modern Information Technologies which are capable of improving the enterprise competitiveness fall in the scope of BI, such as ERP, CRM, Reporting tools, Data Computing, Statistical Analysis, Data Mining, OLAP, and ETL, etc. They can be divided into 2 categories: Desktop BI and Solution BI.

 Desktop BI runs on the desktop environment, and almost does not need the support of the servers. Usually, such tools only provide the core BI function with a relatively low requirement on the technical environments, for example, Microsoft Excel, StataCorp Stata, Raqsoft ES series, IBM SPSS, and RStudio.

On the contrary, solution BI cannot work without the support of dedicated server. It’s usually the integrate solution or multi-module platform. Besides the core BI functions, there are also some external functions like the authority management, resources sharing, and collaboration among various jobs. Such software includes the SAS, Spotfire, Tableau, Qlikview, and SAP BI.

Although there are obvious differences, these two types have many advocates. In the following sections, we will discuss their features respectively from 4 aspects, including software structure, function feature, technical requirements, and target users.

Software Structure

Basic Structure: The Desktop BI is a typical desktop application that can be installed on an average PC. With it, you can access the external database via network and analyze the local files like RStudio if network is unavailable. The Solution BI is usually the B/S or C/S architecture with its own dedicated server. Owing to this, Solution BI requires a high performance server, a complete network environment, and a disaster recovery system like Spotfire. Therefore, everyone can use Desktop BI easily because it does not require much on its environment for use. On the contrary, the solution BI has high environment requirements and can only be used in the enterprise environment.

Installation: Solution BI software can be roughly divided into 2 types: server software and client software. The IT department usually will be responsible for the server end. Although it is very complex, the end users need not to care for it. Most clients are browsers capable to run with zero installation like QlikView. Even for the fat clients, the Java WebStart and the relevant technologies can be leveraged to reduce the complexity of installation and future upgrade. By comparison, Desktop BI requires users to download and install the software and sometimes various plug-ins like RStudio, which adds difficulty to the upgrade and update. It is far less convenient than Solution BI.

Performance: The differences of basic structure result in the great differences of their performance. Desktop BI is operated on the common PC, which means that the computation involving great data volume, high I/O and performance cannot be performed on the desktop. On the contrary, Solution BI usually completes the computation on the high performance server end, so that Solution BI has its advantages in the massive data processing and the operation efficiency. Sometimes, Desktop BI adopts cloud computing, server cluster, multi-core-based parallel computing, etc. as the supplementary means to enhance the performance, such as SPSS. This is certainly not the typical application of SPSS.

Maintenance: Their maintenance overheads are different. Desktop BI is zero maintenance cost and ready to use once installed nearly without any assistance of IT department. By comparison, Solution BI requires a data center, a network administer, and a server administer with a relatively high maintenance overhead. In addition, Solution BI generally requires a set of corresponding supporting mechanism. The administrators and users are also required to go through the professional training at high expense, for example, SAS.

Functions and Features

The functions of Desktop BI usually are specific and focused on a single-purpose, while Solution BI functions are diverse and complete.

BI-specific vs. miscellaneous functions: DeskTop BI only provides the core BI function, excluding the functions like access control, resource assignment, collaboration, and other non-BI functions – even if these functions are available in some software, for example Excel, users almost do not use them. However, the Spotfire, Tableau, and other Solution BI software provides a complete set of non-BI functions such as ACL, logon, logging, performance monitoring, mobile and share, and other non-BI functions. In addition, according to module, the user types can also be divided into system administer, database administer, reporting template designer, common business user, business manager, and other roles.

Single purpose vs. multiple purposes: This means the single usage of Desktop BI. For example, Excel and esCalc mainly focus on the spreadsheet function, while their reporting and ETL functions are relatively weak; RStudio and esProc are mainly for data analysis without any spreadsheet functions. Solution BI integrates a great number of BI modules. For instance, the SAP BI not only provides the ETL and Data Warehouse functions, but also the Reporting and OLAP functions.

In this case, we can find it is a typical example of 80-20 rule/Pareto Law that Desktop BI only offers 20% functions of Solution BI but is capable to achieve 80% BI goal. It is certainly not to deny the value of Solution BI. In fact, a secure, robust, and highly expandable system is a worthy investment.

Study curve:  Due to single-purpose and highly specific functions, Desktop BI is easy and fast for users to learn. Its study curve is smooth and gentle. The learning cost is very low, and users can fully grasp it after a few days of learning. Excel is even assigned as the learning contents for the primary school students. On the contrary, the study curve of Solution BI is quite high because Solution BI software consists of many modules, and provides many complex functions. The robustness and collaboration factors of Solution BI also add the cost and difficulty to learn.

Applicable scenarios: The complex access control and cooperation based on division of jobs determines that Solution BI is only fit for the job involving thousands of documents, rigid work, and routines. For example, for the jobs of reporting, Dashboard, ETL, OLAP, software of Spotfire, Tableau, Qlikview, and SAP BI all belong to this type. For this reason, it always takes quite a long time to complete a Solution BI job. For example, the OLAP analysis with SAP BI requires the multi-step and various jobs collaboration on model design, data acquisition, and task scheduling. Even the apparently simple job of report preparing requires the database administer to compose the stored procedures, the business expert to decide the business logics, the art designer to design the style, and the report designer to write the report scripts.

Instead, Desktop BI is a tool for users to work creatively. Desktop BI requires users to deliver the reliable results before the commercial opportunities slip away. For example, use R to analyze the reason for the sharp increase of website visits. Considering this, its main concern is not the access control, robustness, and cooperation but the agility and convenience.

Technical Requirements

The technical requirements of Desktop BI are lower than that of Solution BI.

Self-independence vs. Teamwork: Desktop BI usually allows users to complete the data computing and statistical analysis independently. The computational goal is highly related to the business, and thus the technical assistance becomes unnecessary. By comparison, Solution BI requires the strong technical assistance to complete the server deployment, report scripts and SQL statements, secondary development, and other IT-related work.

Work Cycle: Because of the lower technical requirements, the work cycle of Desktop software is relatively short. Solution BI software instead requires the cooperation of experts and coordination between various departments, and thus the duration of the cycle is relatively long.

The deviation of computational goal: The relatively low technical requirements can ensure that the business experts can fully grasp the Desktop BI all by themselves. As we know, the “B” of “BI” stands for “Business”. The whole computational goal of BI is to ultimately serve the purpose of business. Therefore, facing the business requirements, to get a result of the most practical and actual to business conditions, and the closest to the computational goal, it is compulsory to enable the business experts to use the tools by themselves, analyze from the business perspective, and turn their business expertise into algorithm.

By comparison, Solution BI instead requires the technical assistance from IT experts. A lot of core algorithms are even proposed by the technical experts, such as the modeling procedure of Qlikview. In the process of implementation, the IT experts usually play the role of constructing the underlying infrastructure on which the business experts will carry out their work. As we know, a solid foundation determines the lifecycle of the superstructure of a building. Therefore, the infrastructure laid by the IT experts will have great impact on the computation procedure of business experts. The ultimate computation results may easily deviate from the computational goal. Every business expert may ever hear these words: No way. You cannot drill like this because we did not consider this in designing; No. The desired data is hard to obtain through SQL and please consider to choose other data; No. we need stored procedure to achieve this, and we will have to discuss it with the DBA and IT manager; No. The cost is too high for IT department spending a few months on it. We need to change the computational goal.

Of course, the high technical requirement is not always a bad thing. For the BI work of relatively huge data volume and clear standard like reporting, the high technical requirement and cooperation based on division of work usually leads to the high quality of output. That is to say, the report is more refined, the algorithm is more efficient, and the model is more complete, just to name a few.

Target Users

According to the above analysis, we can conclude the most suitable user groups for these two types of BI software.

Data computing oriented vs. Data management and presentation oriented: Users of Desktop BI software all focus on the data computing, such as the Microsoft Excel, Raqsoft ES series, RStudio. These types of users usually have to perform the complex procession on data to get a certain valuable computational results. For example, find the product whose sales value has ever been rising for 3 consecutive months. Users of Solution BI are usually data management and presentation oriented. For example, use Dashboard to present the performances of major departments dynamically.

Business-oriented vs. Technology-oriented: Users of Desktop BI are mostly business-oriented. For example, the accountants, account manager for banking, business analyst, and stock analyst. Users of Solution BI are mostly technology-oriented, for example, the report designer, enterprise portal designer, and data warehouse administer. Of course, such division is not always clear, for example, the stock analyst may very likely the expert in both mathematics and finances.The report designers not only understand the reporting scripts, but also have in-depth knowledge in business.

Temporary and sudden demand vs. Daily routine requirement: Users of Desktop BI usually have to confront to the temporary and sudden requirements. Such requirements are often in short of the existing and ready model to use. Instead, it depends on the concrete analysis according to the specific factors at that point. For example: find the main reason for the sharp increase of complaints in last week. Or, of the clients accounting for the top 50% of the sales values, find the ones whose ranks of this year rise again. By comparison, the results of Solution BI are mostly typical and daily common ones. For example, the monthly balance sheet and cash flow report. Or, the 3 dimensions of client, network of dealers, and product, and the 2 metrics of sales value and sales volume for OLAP analysis.

No technical support vs. Available technical support: Desktop BI is fit for those users who can hardly get any support and assistance from IT teams, for example, the sales assistant who travels a lot, business analysts who work at home, and the stock analyst as freelancer. Solution BI is fit for those users who can easily get the support and assistance from IT teams,  for example, the telephone sales assistant, and the banking report designer.

After all, we also must notice that the same users in different occasions will use different tools. For example, an accountant will login on to the enterprise portal to view the routine report, and this falls into the scope of Solution BI. Sometimes, they will use Excel to solve the temporary data computing task arranged by the finance administer, and this falls into the scope of Desktop BI. For another example, the business analyst can use esProc to handle the complex data computing independently if no support is available from the IT teams, and use SAS to perform the massive data computing if the support is available from IT teams.