Goal-driven knowledge discovery

Linking corporate objectives to information infrastructure

A significant challenge faced by many managers is turning board-mandated strategic objectives into operational reality. Andrew Boyd presents a framework for achieving competitive advantage by focusing on goal-driven knowledge discovery. His method requires that functional areas set targets and initiatives based on corporate objectives, laying the groundwork for information retrieval and iterative evaluation.

Researchers and practitioners alike agree that organisations and individuals are overburdened with information. Now more than ever, managers, whether responsible for the organisational knowledge management function or not, need a systematic and repeatable way to identify and gather useful information from their systems, and to link high level corporate objectives to the information architecture.

The method outlined in this article is based on a model developed by researchers at the Software Engineering Institute at Carnegie Mellon University. The Goal-Driven Software Measurement Guidebook was developed as a way for organisations to map software development measurement to business objectives. The goals, questions, indicators, measures (GQIM) approach does not start by asking, ‘what metrics should I use?’ but rather focuses on, ‘what do I want to learn?’ This approach is also particularly useful for mapping corporate objectives to the underlying information architecture. The result is a simply explained taxonomy that both IT staff and managers can understand.

The process of mapping objectives to data sources also produces an undisputable gap analysis that managers can use as justification to further develop systems, or to justify existing time consuming and expensive information management/analysis efforts. In addition, if the data doesn’t exist in the system, it gives management a way to justify the cost of collection, processing and maintenance of new data.

To build a system that links corporate objectives to the information architecture, a manger must:

1. Identify business objectives or desired outcomes;

2. Identify goals and sub-goals for each objective;

3. Identify business entities and attributes;

4. Identify questions that address business goals;

5. Define measures that will be used to construct indicators;

6. Identify and construct indicators that will answer questions;

7. Identify data sources;

8. Implement a measurement programme.

Step 1 – Identify business objectives and desired outcomes

Most organisations that have bothered to write and publish a mission statement produce some variant of: ‘Maximise stakeholder, customer and employee satisfaction while producing the best possible product, in the most efficient manner, using the latest technology.’ Obviously, this is a gross generalisation and a significant amount of time goes into determining the precise wording that will best affect the various stakeholder constituencies. However, if the above mission statement is accepted as a given, a forward-thinking manager can begin to anticipate the organisation’s information needs.

Generally, literature and experience has shown that organisations are interested in maximising or optimising the following business outcomes: productivity, quality, customer satisfaction, innovation and learning, and financial factors such as sales, revenue and profits. If the organisation is publicly traded, shareholder satisfaction also becomes important.

Step 2 – Identify goals and sub-goals for each objective

Before any information is collected or analysed, managers must determine what their business goals are for each entity. For this method to be successful, a goal must be a quantifiable, achievable and measurable target that will drive a desired business outcome. ‘Raise sales in the next quarter’ is not quantifiable. ‘Raise sales in the next quarter by 1000 per cent’ is certainly a goal, but it may not be achievable. Setting goals too high or too low will undermine the programme.

The board of directors, in conjunction with their direct reports, should set goals for each business objective. However, an organisation should try to limit the overall number of goals to about seven. (In 1956, George Miller’s seminal research found that human beings have severe limitations on the amount of information they are able to receive, process and remember. Any more than seven [plus or minus two] goals, and employees won’t remember what they are striving to achieve. Any less may focus efforts too narrowly.)

Step 3 – Identify business entities and attributes

In step two the entities that influence the desired business outcomes are identified. Identifying entities and related attributes helps to refine the questions. In many companies, the word ‘customer’ can mean different things. To the marketing department, it could mean anyone that has placed an order; to finance, anyone who has placed and paid for an order; and to customer services, it could mean anyone that has contacted that department. By identifying entities and determining their related attributes, the organisation is forced to operate within the boundaries of a common language.

Generally, the head of the department or functional area will identify the entities that drive the business outcomes that he or she is most concerned with. Operations may identify business process entities and attributes, whereas the head of marketing would most likely be concerned with the customer entities. It is up to the programme sponsor (or the GQIM moderator within the organisation) to make sure that all entities have been accounted for and defined.

In most businesses or business units, there are generally four primary entity groupings that influence business outcomes: employees, technology, business processes and customers. Each of these entities interacts with one or more of the others to influence a business outcome. For example, employees interact with customers to produce sales. A positive interaction has a positive effect on customer satisfaction. That interaction may also result in ‘learning’ for the employee. However, if the employee spent an inordinate amount of time pleasing the customer, customer satisfaction may be very high, but at the expense of productivity. Similarly, rigid business processes and technology can produce high quality products in a productive way, but innovation and learning may suffer.

To achieve a sustainable competitive advantage, managers must learn to balance these often conflicting goals. To do so, managers must have the information and metrics that support their goals, while the underlying information architecture must underpin these efforts. The OEI (objectives, entities and infrastructure) framework highlights how the information architecture underpins business entities and corporate objectives.

Step 4 – Identify questions that address business goals

This phase is where managers identify all of the information that they need to know to address the business goals. Gathering of questions should be done iteratively, ideally in a workshop environment. When managers know what their colleagues are asking, the process tends to build upon itself. Also, many organisations may be shocked to learn about the cross-departmental duplication of effort that occurs in data collection and analysis. If the required information is not provided through existing management information systems, resourceful managers will often find a way to get their hands on it, although data collected in this manner is rarely shared.

Traditional GQIM asks questions before the identification of entities and attributes, but adds a step that link entities, questions and goals. Experience has shown that identifying the relevant entities and attributes up front gives that organisation a better grounding in asking the right questions, and less time is wasted arguing about definitions and semantics.

The manager of the GQIM effort will find that many of the questions will address multiple goals. For example, the question, ‘how fast are we responding to customer inquiries?’ may address both customer satisfaction and productivity goals. Knowing this may save duplication of effort when constructing indicators.

Step 5 – Define measures that will be used to construct indicators

This step should be considered one of the most important in the exercise. Each entity, as identified in step three, consists of both facts and measures. In this step, facts and measures of the elements that are used in the construction of the indicators must be defined. Facts define what it is that will be measured, such as customers, orders, units, cancellations, etc. Measures, on the other hand, are the formulaic definitions of the units of measurement, such as gross sales, net sales, units in, charge backs, cancellation percentage, etc.

For example, if a ‘gross sales by customers, grouped by segment’ report is required, then ‘sales’, ‘customer’ and ‘segment’ need to be defined and documented. Organisationally, this can be quite a chore. Does ‘gross sales’ mean gross requests for a product, or requests that can be fulfilled? Most online businesses are familiar with the tendency for jokers to order 700 units of the same product to be sent to a particularly hated neighbour. Although fraud detection methods will identify these orders as suspect, it can be embarrassing when a sales manager has to explain that a record sales day is a result of poorly defined reporting measures. As mentioned above, existing departmental nomenclature complicates matters; historically, ‘sales’ could mean two different things to two different divisions.

Step 6 – Identify and construct indicators

Indicators are graphs, tables, charts and reports that managers and executives will use to answer questions and monitor the progress in achieving goals. This part of the exercise can be the most frustrating. Some managers may have the tendency to argue for hours about the best way to present data. Again, seven (plus or minus two) indicators should be chosen to measure progress towards the goals for each entity, with a single overall indicator representing the entity grouping horizontally. For example, the customer entity may have several questions that the sales and marketing manager needs answered. Thus he may need several indicators on a daily basis. However, if the organisational objective is to increase customer satisfaction, only the overall satisfaction score need be regularly reported as a single primary indicator contributing to the organisational seven.

Both dimensions and selections influence indicators. Dimensions define how you want your data sliced (by product, channel, source, etc) and selections define what is included and what is excluded in the indicator.

From experience, a good way to disseminate the indicators to the organisation is to push the information via e-mail or text message to mobile phones. This makes sure that everyone gets the information at roughly the same time. More detailed graphs, charts and tables can be provided on an intranet. Datasets should be provided for analytically inclined managers to do their own ad hoc reporting.

Step 7 – Identify data sources

The next step is to figure out where in the organisation the data needed to construct the indicators resides. Generally, there are two types of databases in an organisation: transactional and analytical. Transactional data is constantly in a state of flux, whereas analytical data is controlled through prescribed updates and cleaning procedures. A critical mistake that smaller organisations often make is to try to derive analytical data from transactional systems. Since the data is always changing, using transactional databases for analytical purposes can lead to inaccurate, confusing and unverifiable reports. Furthermore, large multi-table queries of transactional databases can slow down production systems.

It is important not to overlook two significant data sources: private data stores and paper. Most organisations have an unbelievable amount of data stored on the private drives of employees’ PCs. Resourceful managers will not go without the information they need to do their jobs and will often go to great lengths to create their own private stores. Another, often overlooked, data source is paper. Many organisations still receive quite a bit of information in paper form, including telephone bills, credit card bills and bank statements. If mined properly, this data contains an incredible source of knowledge. Moreover, if asked for, much of this information is available in an electronic format that can easily be imported into the reporting system.

Data refreshing and the timing of reports should also be considered. Prior to developing timed reports, the timing of critical business processes must be mapped and understood. Certain transactional events may occur on a scheduled basis, which can have a serious impact on reporting. For example, if all credit cards are batch processed from 3am, net revenue reports shouldn’t be run at midnight. Also, some indicators may need to be refreshed more often than others, particularly during critical business events. If marketing is running a promotion on green widgets, the green widget report may be all of a sudden be in significantly higher demand than was originally planned. The information system needs to be designed in such a way that the frequency of refreshing information can be changed easily.

Step 8 – Implement a measurement programme

After the first seven steps are completed, the programme manager’s job is far from over. Developing a useful and sustainable measurement programme that links corporate goals to the information architecture is an ongoing process. Corporate objectives, the information infrastructure and the manager’s information needs all change, and a successful measurement programme will change with the organisation. Dave Shuman, EVP of Management Information Systems for Enews Inc, an Arlington, Virginia-based magazine agent, offers three tips for success when developing an information and reporting system:

Get board level commitment. Although it need not be expensive, a good measurement programme will command both financial and human resources. Having a very senior executive sponsor will help in negotiating budgets and resources from other departments;
Start with a clean pot. There is a tendency to start with an old, likely inadequate, system and build the new one on top of it. But a good chef will always start with a clean pot. Certainly elements from older systems can be used in building the new one, but only if appropriate. But a word of warning: ‘starting with a clean pot’ can create political battles with the designers and managers of the older systems;
Identify and deliver short wins. Without tangible benefits or a high degree of visibility into the development process, organisational support for the measurement programme will wane. To mitigate this risk, the programme manager should identify and frequently deliver small but noticeable benefits to end-users.

Andrew Boyd is responsible for business development at Aspect Proteus. He can be contacted at andrew.boyd@aspectproteus.com