Database Management Systems and SQL

Data-processing techniques, processing power, and enterprise performance management capabilities have undergone revolutionary advances in recent years for reasons you are already familiar with—big data, mobility, and cloud computing. The last decade, however, has seen the emergence of new approaches, first in data warehousing and, more recently, for transaction processing. Given the huge number of transactions that occur daily in an organization, the data in databases are constantly in use or being updated. The volatility of databases makes it impossible to use them for complex decision-making and problem-solving tasks. For this reason, data are extracted from the database, transformed (processed to standardize the data), and then loaded into a data warehouse.

Database management systems (DBMSs) integrate with data collection systems such as TPS and business applications; store the data in an organized way; and provide facilities for accessing and managing that data. Factors to consider when evaluating the performance of a database management system are listed in Tech Note 3.1. Over the past 25 years, the relational database has been the standard database model adopted by most enterprises. Relational databases store data in tables consisting of columns and rows, similar to the format of a spreadsheet, as shown in Figure 3.3.

Relational management systems (RDBMSs) provide access to data using a declarative language—structured query language (SQL). Declarative languages simplify data access by requiring that users only specify what data they want to access without defining how access will be achieved. The format of a basic SQL statement is

• SELECT column_name(s)
• FROM table_name
• WHERE condition

An instance of SQL is shown in Figure 3.4.

DBMS Functions

An accurate and consistent view of data throughout the enterprise is needed so one can make informed, actionable decisions that support the business strategy. Functions performed by a DBMS to help create such a view are shown in Figure 3.5.

Select the caption to view an interactive version of this figure online.

DBMS and Data Warehousing Vendors Respond to Latest Data Demands

One of the major drivers of change in the data management market is the increased amount of data to be managed. Enterprises need powerful DBMSs and data warehousing solutions, analytics, and reporting. The four vendors that dominate this market—Oracle, IBM, Microsoft, and Teradata—continue to respond to evolving data management needs with more intelligent and advanced software and hardware. Advanced hardware technology enables scaling to much higher data volumes and workloads than previously possible, or it can handle specific workloads. Older general-purpose relational databases DBMSs lack the scalability or flexibility for specialized or very large workloads, but are very good at what they do.

Concept Check 3.1

Garbage In, Garbage Out

Data collection is a highly complex process that can create problems concerning the quality of the data being collected. Therefore, regardless of how the data are collected, they need to be validated so users know they can trust them. Classic expressions that sum up the situation are “garbage in, garbage out” (GIGO) and the potentially riskier “garbage in, gospel out.” In the latter case, poor-quality data are trusted and used as the basis for planning. For example, you have probably encountered data safeguards, such as integrity checks, to help improve data quality when you fill in an online form, such as when the form will not accept an e-mail address or a credit card number that is not formatted correctly.

Table 3.2 lists the characteristics typically associated with dirty or poor-quality data.

Data Ownership and Organizational Politics

Compliance with numerous federal and state regulations relies on rock-solid data and trusted metrics used for regulatory reporting. Data ownership, data quality, and formally managed data are high priorities on the agenda of CFOs and CEOs who are held personally accountable if their company is found to be in violation of regulations.

Despite the need for high-quality data, organizational politics and technical issues make that difficult to achieve. The source of the problem is data ownership—that is, who owns or is responsible for the data. Data ownership problems exist when there are no policies defining responsibility and accountability for managing data. Inconsistent data formats of various departments create an additional set of problems as organizations try to combine individual applications into integrated enterprise systems.

The tendency to delegate data-quality responsibilities to the technical teams who have no control over data quality, as opposed to business users who do have such control, is another common pitfall that stands in the way of accumulating high-quality data.

Those who manage a business or part of a business are tasked with trying to improve business performance and retain customers. Compensation is tied to improving profitability, driving revenue growth, and improving the quality of customer service. These key performance indicators (KPIs) are monitored closely by senior managers who want to find and eliminate defects that harm performance. It is strange then that so few managers take the time to understand how performance is impacted by poor-quality data. Two examples make a strong case for investment in high-quality data.

Retail banking: For retail bank executives, risk management is the number one issue. Disregard for risk contributed to the 2008 financial services meltdown. Despite risk management strategies, many banks still incur huge losses. Part of the problem in many banks is that their ISs enable them to monitor risk only at the product level—mortgages, loans, or credit cards. Product-level risk management ISs monitor a customer’s risk exposure for mortgages, or for loans, or for credit cards, and so forth—but not for a customer for all products. With product-level ISs, a bank cannot see the full risk exposure of a customer. The limitations of these siloed product-level risks have serious implications for business performance because bad-risk customers cannot be identified easily, and customer data in the various ISs may differ. However, banks are beginning to use big data to analyze risk more effectively. Although they are still very limited to credit card, loan, and mortgage risk data, cheaper and faster computing power allows them to keep better and more inclusive records of customer data. Portfolio monitoring offers earlier detection and predictive analytics for potential customers, and more advanced risk models show intricate patterns unseen by the naked eye in large data sets. Also, more fact-based inputs and standardized organizational methods are being implemented to reduce loan and credit officer bias to take risks on undesirable customers.

Marketing: Consider what happens when each product-level risk management IS feeds data to marketing ISs. Marketing may offer bad-risk customers incentives to take out another credit card or loan that they cannot repay. And since the bank cannot identify its best customers either, they may be ignored and enticed away by better deals offered by competitors. This scenario illustrates how data ownership and data-quality management are critical to risk management. Data defects and incomplete data can quickly trigger inaccurate marketing and mounting losses. Banks’ increasing dependence on business modeling requires that risk managers understand and manage model risk better. Although losses often go unreported, the consequences of errors in the model can be extreme. For instance, a large Asia–Pacific bank lost $4 billion when it applied interest-rate models that contained incorrect assumptions and data-entry errors. Risk mitigation will entail rigorous guidelines and processes for developing and validating models, as well as the constant monitoring and improvement of them (Harle et al., 2016).

Manufacturing: Many manufacturers are at the mercy of a powerful customer base—large retailers. Manufacturers want to align their processes with those of large retail customers to keep them happy. This alignment makes it possible for a retailer to order centrally for all stores or to order locally from a specific manufacturer. Supporting both central and local ordering makes it difficult to plan production runs. For example, each manufacturing site has to collect order data from central ordering and local ordering systems to get a complete picture of what to manufacture at each site. Without accurate, up-to-date data, orders may go unfilled, or manufacturers may have excess inventory. One manufacturer who tried to keep its key retailer happy by implementing central and local ordering could not process orders correctly at each manufacturing site. No data ownership and lack of control over how order data flowed throughout business operations had negative impacts. Conflicting and duplicate business processes at each manufacturing site caused data errors, leading to mistakes in manufacturing, packing, and shipments. Customers were very dissatisfied.

These examples demonstrate the consequences of a lack of data ownership and data quality. Understanding the impact mismanaged data can have on business performance highlights the need to make data ownership and data accuracy a high priority.

Data Life Cycle and Data Principles

The data life cycle is a model that illustrates the way data travel through an organization as shown in Figure 3.8. The data life cycle begins with storage in a database, to being loaded into a data warehouse for analysis, then reported to knowledge workers or used in business apps. Supply chain management (SCM), customer relationship management (CRM), and e-commerce are enterprise applications that require up-to-date, readily accessible data to function properly.

Three general data principles relate to the data life cycle perspective and help to guide IT investment decisions:

1. Principle of diminishing data value The value of data diminishes as they age. This is a simple, yet powerful principle. Most organizations cannot operate at peak performance with blind spots (lack of data availability) of 30 days or longer. Global financial services institutions rely on near real-time data for peak performance.
2. Principle of 90/90 data use According to the 90/90 data-use principle, a majority of stored data, as high as 90%, is seldom accessed after 90 days (except for auditing purposes). That is, roughly 90% of data lose most of their value after three months.
3. Principle of data in context The capability to capture, process, format, and distribute data in near real time or faster requires a huge investment in data architecture (Chapter 2) and infrastructure to link remote POS systems to data storage, data analysis systems, and reporting apps. The investment can be justified on the principle that data must be integrated, processed, analyzed, and formatted into “actionable information.”

Master Data and Master Data Management

As data become more complex and their volumes explode, database performance degrades. One solution is the use of master data and master data management (MDM) as introduced in Chapter 2. MDM processes integrate data from various sources or enterprise applications to create a more complete (unified) view of a customer, product, or other entity. Figure 3.9 shows how master data serve as a layer between transactional data in a database and analytical data in a data warehouse. Although vendors may claim that their MDM solution creates “a single version of the truth,” this claim is probably not true. In reality, MDM cannot create a single unified version of the data because constructing a completely unified view of all master data is simply not possible.

Select the caption to view an interactive version of this figure online.

Concept Check 3.2

Procedures to Prepare EDW Data for Analytics

Consider a bank’s database. Every deposit, withdrawal, loan payment, or other transaction adds or changes data. The volatility caused by constant transaction processing makes data analysis difficult—and the demands to process millions of transactions per second consume the database’s processing power. In contrast, data in warehouses are relatively stable, as needed for analysis. Therefore, select data are moved from databases to a warehouse. Specifically, data are as follows:

1. Extracted from designated databases.
2. Transformed by standardizing formats, cleaning the data, integrating them.
3. Loaded into a data warehouse.

These three procedures—extract, transform, and load—are referred to by their initials ETL (Figure 3.10). In a warehouse, data are read-only; that is, they do not change until the next ETL.

Three technologies involved in preparing raw data for analytics include ETL, change data capture (CDC), and data deduplication (“deduping the data”). CDC processes capture the changes made at data sources and then apply those changes throughout enterprise data stores to keep data synchronized. CDC minimizes the resources required for ETL processes by only dealing with data changes. Deduping processes remove duplicates and standardize data formats, which helps to minimize storage and data synch.

Building a Data Warehouse

Figure 3.11 diagrams the process of building and using a data warehouse. The organization’s data from operational transaction processes systems are stored in operational databases (left side of the figure). Not all data are transferred to the data warehouse. Frequently, only summary data are transferred. The warehouse organizes the data in multiple ways—by subject, functional area, vendor, and product. As shown, the data warehouse architecture defines the flow of data that starts when data are captured by transaction systems; the source data are stored in transactional (operational) databases; ETL processes move data from databases into data warehouses or data marts, where the data are available for access, reports, and analysis.

Real-Time Support from an Active Data Warehouse

Early data warehouse technology primarily supported strategic applications that did not require instant response time, direct customer interaction, or integration with operational systems. ETL might have been done once per week or once per month. But, demand for information to support real time customer interaction and operations leads to real-time data warehousing and analytics—known as an active data warehouse (ADW). Massive increases in computing power, processing speeds, and memory made ADW possible. ADW are not designed to support executives’ strategic decision-making, but rather to support operations. For example, shipping companies like DHL use huge fleets of trucks to move millions of packages. Every day and all day, operational managers make thousands of decisions that affect the bottom line, such as: “Do we need four trucks for this run?” “With two drivers delayed by bad weather, do we need to bring in extra help?” Traditional data warehousing is not suited for immediate operational support, but active data warehousing is. For example, companies with an ADW are able to:

• Interact with a customer to provide superior customer service.
• Respond to business events in near real time.
• Share up-to-date status data among merchants, vendors, customers, and associates.

Here are some examples of how two companies use ADW.

Capital One. Capital One uses its ADW to track each customer’s “profitability score” to determine the level of customer service to provide for that person. Higher-cost personalized service is only given to those with high scores. For instance, when a customer calls Capital One, he or she is asked to enter a credit card number, which is linked to a profitability score. Low-profit customers get a voice response unit only; high-profit customers are connected to a live customer service representative (CSR) because the company wants to minimize the risk of losing those customers.

Travelocity. If you use Travelocity, an ADW is finding the best travel deals especially for you. The goal is to use “today’s data today” instead of “yesterday’s data today.” The online travel agency’s ADW analyzes your search history and destinations of interest; then predicts travel offers that you would most likely purchase. Offers are both relevant and timely to enhance your experience, which helps close the sale in a very competitive market. For example, when a customer is searching flights and hotels in Las Vegas, Travelocity recognizes the interest—the customer wants to go to Vegas. The ADW searches for the best-priced flights from all carriers, builds a few package deals, and presents them in real time to the customer. When customers see a personalized offer they are already interested in, the ADW helps generate a better customer experience. The real-time data-driven experience increases the conversion rate and sales.

Data warehouse content can be delivered to decision-makers throughout the enterprise via the cloud or company-owned intranets. Users can view, query, and analyze the data and produce reports using Web browsers. These are extremely economical and effective data delivery methods.

Concept Check 3.3

Human Expertise and Judgment are Needed

Human expertise and judgment are needed to interpret the output of analytics (refer to Figure 3.13). Data are worthless if you cannot analyze, interpret, understand, and apply the results in context. This brings up several challenges:

• Data need to be prepared for analysis For example, data that are incomplete or duplicated need to be fixed.
• Dirty data degrade the value of analytics The “cleanliness” of data is very important to data mining and analysis projects. Analysts have complained that data analytics is like janitorial work because they spend so much time on manual, error-prone processes to clean the data. Large data volumes and variety mean more data that are dirty and harder to handle.
• Data must be put into meaningful context If the wrong analysis or datasets are used, the output would be nonsense, as in the example of the Super Bowl winners and stock market performance. Stated in reverse, managers need context in order to understand how to interpret traditional and big data.

IT at Work 3.2 describes how big data analytics, collaboration, and human expertise have transformed the new drug development process.

Machine-generated sensor data are becoming a larger proportion of big data (Figure 3.14), according to a research report by IDC (2015). It is predicted that these data will increase to two-thirds of all data by 2020, representing a significant increase from the 11% level of 2005. In addition to its growth as a portion of analyzed data, the market for sensor data will increase to $1.7 trillion in 2020.

On the consumer side, a significant factor in this market is the boom in wearable technology—products like FitBit and the Apple Watch. Users no longer even have to input data to these devices as it is automatically gathered and tracked in real time. On the public sector and enterprise side, sensor data and the Internet of Things (IoT) are being used in the advancement of IT-enabled business processes like automated factories and distribution centers and IT-enabled products like the wearable tech (IDC, 2015). Federal health reform efforts have pushed health-care organizations toward big data and analytics. These organizations are planning to use big data analytics to support revenue cycle management, resource utilization, fraud prevention, health management, and quality improvement.

Hadoop and MapReduce

Big data volumes exceed the processing capacity of conventional database infrastructures. A widely used processing platform is Apache Hadoop. It places no conditions on the structure of the data it can process. Hadoop distributes computing problems across a number of servers. Hadoop implements MapReduce in two stages:

1. Map stage MapReduce breaks up the huge dataset into smaller subsets; then distributes the subsets among multiple servers where they are partially processed.
2. Reduce stage The partial results from the map stage are then recombined and made available for analytic tools.

To store data, Hadoop has its own distributed file system, Hadoop File System (HDFS), which functions in three stages:

• Loads data into HDFS.
• Performs the MapReduce operations.
• Retrieves results from HDFS.

Figure 3.15 diagrams how Facebook uses database technology and Hadoop. IT at Work 3.3 describes how First Wind has applied big data analytics to improve the operations of its wind farms and to support sustainability of the planet by reducing environmentally damaging carbon emissions.

Data and Text Mining

Data and text mining are different from DBMS and data analytics. As you have read earlier in this chapter, a DBMS supports queries to extract data or get answers from huge databases. But, in order to perform queries in a DBMS you must first know the question you want to be answered. You also have read that Data Analytics describes the entire function of applying technologies, algorithms, human expertise, and judgment. Data and text mining are specific analytic techniques that allow users to discover knowledge that they didn’t know existed in the databases.

Data mining software enables users to analyze data from various dimensions or angles, categorize them, and find correlations or patterns among fields in the data warehouse. Up to 75% of an organization’s data are nonstructured word-processing documents, social media, text messages, audio, video, images and diagrams, faxes and memos, call center or claims notes, and so on.

IT at Work 3.4 describes one example of how the U.S. government is using data mining software to continuously improve its detection and deterrence systems.

Text mining is a broad category that involves interpreting words and concepts in context. Any customer becomes a brand advocate or adversary by freely expressing opinions and attitudes that reach millions of other current or prospective customers on social media. Text mining helps companies tap into the explosion of customer opinions expressed online. Social commentary and social media are being mined for sentiment analysis or to understand consumer intent. Innovative companies know they could be more successful in meeting their customers’ needs, if they just understood them better. Tools and techniques for analyzing text, documents, and other nonstructured content are available from several vendors.

Combining data and text mining can create even greater value. Burns (2016) pointed out that mining text or nonstructural data enables organizations to forecast the future instead of merely reporting the past. He also noted that forecasting methods using existing structured data and nonstructured text from both internal and external sources provide the best view of what lies ahead.

Creating Business Value

Enterprises invest in data mining tools to add business value. Business value falls into three categories, as shown in Figure 3.16.

Here are some brief cases illustrating the types of business value created by data and text mining.

1. Using pattern analysis, Argo Corporation, an agricultural equipment manufacturer based in Georgia, was able to optimize product configuration options for farm machinery and real-time customer demand to determine the optimal base configurations for its machines. As a result, Argo reduced product variety by 61% and cut days of inventory by 81% while still maintaining its service levels.
2. The mega-retailer Walmart wanted its online shoppers to find what they were looking for faster. Walmart analyzed clickstream data from its 45 million monthly online shoppers; then combined that data with product- and category-related popularity scores. The popularity scores had been generated by text mining the retailer’s social media streams. Lessons learned from the analysis were integrated into the Polaris search engine used by customers on the company’s website. Polaris has yielded a 10% to 15% increase in online shoppers completing a purchase, which equals roughly $1 billion in incremental online sales.
3. McDonald’s bakery operation replaced manual equipment with high-speed photo analyses to inspect thousands of buns per minute for color, size, and sesame seed distribution. Automatically, ovens and baking processes adjust instantly to create uniform buns and reduce thousands of pounds of waste each year. Another food products company also uses photo analyses to sort every french fry produced in order to optimize quality.
4. Infinity Insurance discovered new insights that it applied to improve the performance of its fraud operation. The insurance company text mined years of adjuster reports to look for key drivers of fraudulent claims. As a result, the company reduced fraud by 75%, and eliminated marketing to customers with a high likelihood of fraudulent claims.

Text Analytics Procedure

With text analytics, information is extracted from large quantities of various types of textual information. The basic steps involved in text analytics include the following:

1. Exploring First, documents are explored. This might occur in the form of simple word counts in a document collection, or by manually creating topic areas to categorize documents after reading a sample of them. For example, what are the major types of issues (brake or engine failure) that have been identified in recent automobile warranty claims? A challenge of the exploration effort is misspelled or abbreviated words, acronyms, or slang.
2. Preprocessing Before analysis or the automated categorization of content, the text may need to be preprocessed to standardize it to the extent possible. As in traditional analysis, up to 80% of preprocessing time can be spent preparing and standardizing the data. Misspelled words, abbreviations, and slang may need to be transformed into consistent terms. For instance, BTW would be standardized to “by the way” and “left voice message” could be tagged as “lvm.”
3. Categorizing and modeling Content is then ready to be categorized. Categorizing messages or documents from information contained within them can be achieved using statistical models and business rules. As with traditional model development, sample documents are examined to train the models. Additional documents are then processed to validate the accuracy and precision of the model, and finally new documents are evaluated using the final model (scored). Models can then be put into production for the automated processing of new documents as they arrive.

Analytics Vendor Rankings

Analytics applications cover business intelligence functions sold as a standalone application for decision support or embedded in an integrated solution. The introduction of intuitive decision support tools, dashboards, and data visualization (discussed in detail in Chapter 11) have added some interesting interactive components to big data analytics to bring the data to life and enable nonexperts to use it.

Organizations invest in analytics, BI, and data/text mining applications based on new features and capabilities beyond those offered by their legacy systems. Analytics vendors offer everything from simple-to-use reporting tools to highly sophisticated software for tackling the most complex data analysis problems. A list of the top five analytics and BI application vendors are shown in Table 3.4.

Concept Check 3.4

Business Benefits of BI

BI provides data at the moment of value to a decision-maker—enabling it to extract crucial facts from enterprise data in real time or near real time. A BI solution with a well-designed dashboard, for example, provides retailers with better visibility into inventory to make better decisions about what to order, how much, and when in order to prevent stock-outs or minimize inventory that sits on warehouse shelves.

Companies use BI solutions to determine what questions to ask and find answers to them. BI tools integrate and consolidate data from various internal and external sources and then process them into information to make smart decisions. BI answers questions such as these: Which products have the highest repeat sales rate in the last six months? Do customer likes on Facebook relate to product purchase? How does the sales trend break down by product group over the last five years? What do daily sales look like in each of my sales regions?

According to The Data Warehousing Institute, BI “unites data, technology, analytics, and human knowledge to optimize business decisions and ultimately drive an enterprise’s success. BI programs usually combine an enterprise data warehouse and a BI platform or tool set to transform data into usable, actionable business information” (The Data Warehousing Institute, 2014). For many years, managers have relied on business analytics to make better-informed decisions. Multiple surveys and studies agree on BI’s growing importance in analyzing past performance and identifying opportunities to improve future performance.

Common Challenges: Data Selection and Quality

Companies cannot analyze all of their data—and much of them would not add value. Therefore, an unending challenge is how to determine which data to use for BI from what seems like unlimited options (Oliphant, 2016). One purpose of a BI strategy is to provide a framework for selecting the most relevant data without limiting options to integrate new data sources. Information overload is a major problem for executives and for employees. Another common challenge is data quality, particularly with regard to online information, because the source and accuracy might not be verifiable.

Aligning BI Strategy with Business Strategy

Reports and dashboards are delivery tools, but they may not be delivering business intelligence. To get the greatest value out of BI, the CIO needs to work with the CFO and other business leaders to create a BI governance program whose mission is to achieve the following (Ladley, 2016):

1. Clearly articulate business strategies.
2. Deconstruct the business strategies into a set of specific goals and objectives—the targets.
3. Identify the key performance indicators (KPIs) that will be used to measure progress toward each target.
4. Prioritize the list of KPIs.
5. Create a plan to achieve goals and objectives based on the priorities.
6. Estimate the costs needed to implement the BI plan.
7. Assess and update the priorities based on business results and changes in business strategy.

After completing these activities, BI analysts can identify the data to use in BI and the source systems. This is a business-driven development approach that starts with a business strategy and work backward to identify data sources and the data that need to be acquired and analyzed.

Businesses want KPIs that can be utilized by both departmental users and management. In addition, users want real-time access to these data so that they can monitor processes with the smallest possible latency and take corrective action whenever KPIs deviate from their target values. To link strategic and operational perspectives, users must be able to drill down from highly consolidated or summarized figures into the detailed numbers from which they were derived to perform in-depth analyses.

BI Architecture and Analytics

BI architecture is undergoing technological advances in response to big data and the performance demands of end-users (Wise, 2016). BI vendors are facing the challenges of social, sensor, and other newer data types that must be managed and analyzed. One technology advance that can help handle big data is BI in the cloud. Figure 3.17 lists the key factors contributing to the increased use of BI. It can be hosted on a public or private cloud. Although cloud services come with more upkeep, optimizing the service and customizing it for one’s company brings undeniable benefits in data security. With a public cloud, a service provider hosts the data and/or software that are accessed via an Internet connection. For private clouds, the company hosts its own data and software, but uses cloud-based technologies.

For cloud-based BI, a popular option offered by a growing number of BI tool vendors is software as a service (SaaS). MicroStrategy offers MicroStrategy Cloud, which provides fast deployment with reduced project risks and costs. This cloud approach appeals to small and midsized companies that have limited IT staff and want to carefully control costs. The potential downsides include slower response times, security risks, and backup risks.

Electronic Records Management

All organizations create and retain business records. A record is documentation of a business event, action, decision, or transaction. Examples are contracts, research and development, accounting source documents, memos, customer/client communications, hiring and promotion decisions, meeting minutes, social posts, texts, e-mails, website content, database records, and paper and electronic files. Business documents such as spreadsheets, e-mail messages, and word-processing documents are a type of record. Most records are kept in electronic format and maintained throughout their life cycle—from creation to final archiving or destruction by an electronic records management system (ERMS).

One application of an ERMS would be in a company that is required by law to retain financial documents for at least seven years, product designs for many decades, and e-mail messages about marketing promotions for a year. The major ERM tools are workflow software, authoring tools, scanners, and databases. ERM systems have query and search capabilities so documents can be identified and accessed like data in a database. These systems range from those designed to support a small workgroup to full-featured, Web-enabled enterprisewide systems.

Legal Duty to Retain Business Records

Companies need to be prepared to respond to an audit, federal investigation, lawsuit, or any other legal action against them. Types of lawsuits against companies include patent violations, product safety negligence, theft of intellectual property, breach of contract, wrongful termination, harassment, discrimination, and many more.

Because senior management must ensure that their companies comply with legal and regulatory duties, managing electronic records (e-records) is a strategic issue for organizations in both the public and private sectors. The success of ERM depends greatly on a partnership of many key players, namely, senior management, users, records managers, archivists, administrators, and most importantly, IT personnel. Properly managed, records are strategic assets. Improperly managed or destroyed, they become liabilities.

ERM Best Practices

Effective ERM systems capture all business data and documents at their first touchpoint—data centers, laptops, the mailroom, at customer sites, or remote offices. Records enter the enterprise in multiple ways—from online forms, bar codes, sensors, websites, social sites, copiers, e-mails, and more. In addition to capturing the entire document as a whole, important data from within a document can be captured and stored in a central, searchable repository. In this way, the data are accessible to support informed and timely business decisions.

In recent years, organizations such as the Association for Information and Image Management (AIIM), National Archives and Records Administration (NARA), and ARMA International (formerly the Association of Records Managers and Administrators) have created and published industry standards for document and records management. Numerous best practices articles, and links to valuable sources of information about document and records management, are available on their websites. The IT Toolbox describes ARMA’s eight generally accepted recordkeeping principles framework.

ERM Benefits

Departments or companies whose employees spend most of their day filing or retrieving documents or warehousing paper records can reduce costs significantly with ERM. These systems minimize the inefficiencies and frustration associated with managing paper documents and workflows. However, they do not create a paperless office as had been predicted.

An ERM can help a business to become more efficient and productive by the following:

• Enabling the company to access and use the content contained in documents.
• Cutting labor costs by automating business processes.
• Reducing the time and effort required to locate information the business needs to support decision-making.
• Improving the security of content, thereby reducing the risk of intellectual property theft.
• Minimizing the costs associated with printing, storing, and searching for content.

When workflows are digital, productivity increases, costs decrease, compliance obligations are easier to verify, and green computing becomes possible. Green computing is an initiative to conserve our valuable natural resources by reducing the effects of our computer usage on the environment. You can read about green computing and the related topics of reducing an organization’s carbon footprint, sustainability, and ethical and social responsibilities in Chapter 14.

ERM for Disaster Recovery, Business Continuity, and Compliance

Businesses also rely on their ERM system for disaster recovery and business continuity, security, knowledge sharing and collaboration, and remote and controlled access to documents. Because ERM systems have multilayered access capabilities, employees can access and change only the documents they are authorized to handle.

When companies select an ERM to meet compliance requirements, they should ask the following questions:

1. Does the software meet the organization’s needs? For example, can the DMS be installed on the existing network? Can it be purchased as a service?
2. Is the software easy to use and accessible from Web browsers, office applications, and e-mail applications? If not, people will not use it.
3. Does the software have lightweight, modern Web and graphical user interfaces that effectively support remote users?
4. Before selecting a vendor, it is important to examine workflows and how data, documents, and communications flow throughout the company. For example, know which information on documents is used in business decisions. Once those needs and requirements are identified, they guide the selection of technology that can support the input types—that is, capture and index them so they can be archived consistently and retrieved on-demand.

IT at Work 3.5 describes how several companies currently use ERM. Simply creating backups of records is not sufficient because the content would not be organized and indexed to retrieve them accurately and easily. The requirement to manage records—regardless of whether they are physical or digital—is not new.

Concept Check 3.5