![]()
Part One
THE FIVE PRINCIPLES OF EVIDENCE-BASED CHANGE
![]()
CHAPTER 1
INFORMATION OVERLOAD OR PERSUASIVE ANALYTICS?
Not long ago, it was widely agreed that a lack of metrics hindered the HR profession’s ability to demonstrate its value, influence key decision makers, and uncover insights into the effects of human capital on strategic success. Today—according to Boudreau and Ramstad (2007), and as observed by Cascio and Boudreau (2010)—the hard work of the HR profession, and of the thought leaders on HR measurement, has led to an embarrassment of riches. Information overload is now far more prevalent than the lack of data. Just consider the extensive numbers produced by the typical HR information system. It is often possible to generate statistics—such as turnover rates, salary costs, demographic distributions, competency inventories, and employee-opinion levels—at the touch of a key, and to “cut” those statistics to focus in on business units, individual leaders, and specific employee groups, product lines, or regions.
This is not to say that the data HR wants are always just sitting in the HR information system, waiting to be used. Many HR and business leaders find that the available information is often not suited to their strategic questions, and they invest heavily in developing new measures that better illuminate key relationships. It is clear, however, that the future of HR will be characterized less by the lack of data than by questions about how to use data well and generate data judiciously.
Figure 1.1 shows, in three dimensions, how the HR consulting firm Towers Watson conveys the potential arenas of measurement. This figure illustrates the myriad ways in which HR can tackle its data. Being drowned in a sea of numbers is the problem; logic-driven analytics is the solution. The z-axis denotes the alternative business strategies an organization might pursue, whereas the x-axis captures the elements of the talent life cycle and the y-axis focuses on the four common categories of metrics. Each cell thus captures a unique set of metrics that are specific to the strategy, life cycle element, and type of measurement focus for an organization. For example, a consumer goods company with an innovation-based strategy that is interested in employee metrics related to sourcing and selection might focus on the turnover of new hires in its product development group. A data cube like this helps organize data and metrics and identify the areas of focus.
Understanding Logic-Driven Analytics
The real crux of logic-driven analytics is that it is not enough to have numbers, and it is not enough to do an analysis of those numbers—there has to be an underlying logic guiding the analysis.
A typical HR metric is turnover. Turnover numbers are interesting. Add in an analysis of the cost of turnover, and the business begins to pay attention. Yet HR really begins to add value when it builds a logic around what is good turnover, what is bad turnover, and how the costs and benefits surrounding turnover can be optimized in light of business needs. This logic transforms the turnover metric from an interesting number into meaningful evidence that can guide the organization in making the right kinds of changes.
By logic, we are not implying the dry, formal methods taught in first-year philosophy. The term logic simply points to some context, some line of reasoning, that guides the analysis. In the example of the hospital (see the Introduction), the turnover numbers needed to be looked at in the context of what each department was delivering. There was a logic behind the relatively high turnover in the food service division. In light of that logic, it was possible to know whether the turnover was good or bad and whether any changes needed to be made. Without that logic, HR could not have seen what the numbers meant.
To help guide HR toward logic-driven analytics, Boudreau and Ramstad (2007) developed the logic, analytics, measures, and process (LAMP) mnemonic. The important point to remember is that logic comes first. Later in this book (see Chapter Nine), we will see how IBM captured numerous measures about the competencies of its consultants. This data capture was guided by the logic of viewing the talent pipeline as similar to a supply-chain pipeline, with the goal of having the right talent available to fill the business needs at the right time. Without the logical framework built on the supply-chain concept, IBM would have had a lot of metrics but no powerful metaphor to guide talent-management decisions on the basis of those data. High-quality measures and sophisticated analytics can reveal important insights, but evidence-based change often hinges on using the appropriate logic to target the analysis to the most promising and important questions.
In the Ameriprise case that also appears later in this book (see Chapter Ten), the issue was which HR services to continue to provide. One logical framework that was deployed was a marketing concept called Kano analysis; at its simplest, it involves distinguishing among the features that customers must have, the ones that they would like to have, and the ones that they do not expect but would be delighted to have. The logic used in this case guided thinking on what questions to ask and what data to collect about existing HR services. The company was able to collect all kinds of data about HR services—usage rates, user-satisfaction rates, figures for estimated impacts on sales, and so on—but in the absence of a logical framework such as Kano analysis, the company would have had just a collection of numbers, not a guide to decision making.
The process part of the LAMP framework is a critical but frequently overlooked aspect of logic-driven analytics. Good measures and analysis driven by a logical framework are not enough; there must be a process for communicating and framing the information in such a way that key constituents outside HR will act on it. Valid and insightful statistical analysis can add great value, but there is a danger that sophisticated statistical analysis by competent HR professionals can fall flat if it is seen as “HR speak.” The same analysis can be a source of breakthrough change if carefully constructed to answer the right questions, and if presented in a way that makes it actionable by key decision makers. The latter outcome requires skills well beyond those of the typical data analyst or systems expert. IBM brought in a high-level supply-chain expert to help HR with the logic and assist with the process of communicating the implications to stakeholders.
Data and sophisticated analysis translate into true change leadership only when combined with the human touch. This calls for an individual or a group that brings an awareness of the audience, an ability to find the most significant relationships, an expert’s eye for the organization’s prominent mental models, and the insight of a good storyteller. Gebauer and Lowman (2008) share the compelling story of how McKesson combined data and analytics from multiple perspectives with leadership action to engage and involve employees in significantly enhancing the performance of the organization. It is this ability to use metrics thoughtfully in motivating and informing change that is at the heart of evidence-based change.
The “Analytics” in Logic-Driven Analytics
HR leaders are embracing the fact that skills in data collection and analysis are fundamental competencies for the HR profession. Analysis is sometimes the domain of a few specialists (note they are not necessarily in HR) with advanced degrees in such areas as psychology or economics. Nevertheless, the future of HR depends not only on such specialists but also on what is still the rare capacity for HR professionals to be at home with basic principles of data analysis, research design, and the statistical inferences that can and cannot be made from a set of data.
Here are a few principles of statistical and research design that are fundamental to a wide variety of human capital analyses but are often sources of error when they are ignored:
- Sampling, which makes it possible to generalize findings to the relevant situations
- Correlation (an instance of two phenomena tending to move in the same direction) as distinguished from causality (an instance of two phenomena moving together because one causes the other)
- Elimination of alternative explanations through careful design of experiments and quasi-experiments
None of these techniques requires an advanced degree in order to be used, and HR professionals need to take enough interest in these ideas to begin applying them to their work. Interested readers can learn more about these techniques by reading Cascio and Boudreau (2010) or by partnering with people who have experience in this area to grasp how these techniques are applied to HR analytics. However HR approaches learning these basic analytical tools, the takeaway is that they now belong in the toolkit of the typical HR professional; they are no longer just the purview of expert data analysts.
Another useful set of ideas about analytics is shown in Figure 1.2. In this figure, we see how Towers Watson thinks about the analytical domain in terms of the results of the analysis (the four rows) as they are applied to the main elements of the talent life cycle (the five columns). There are four sorts of analytical outcomes, represented by the rows labeled Optimize, Predict, Correlate, and Describe and Benchmark. With respect to the element labeled Source and Select, a starting point is simply to collect data, such as “cost per hire”—a Describe and Benchmark outcome. A more sophisticated approach is to undertake some analysis, such as correlating the quality of hires to business performance. As just noted, correlation should not be confused with causation, and a higher level of analysis seeks to predict causal outcomes on the basis of metrics (for example, one can predict the impact on productivity of a reduction in turnover among new hires). Finally, the most sophisticated outcome of analysis would be a model in which, for example, HR simulates the impact of investment in programs for new hires. The point is not that every analysis must strive for the upper part of the matrix, but rather that useful and change-inducing results can occur at every level of analysis. Where logic-driven analytics is concerned, HR needs to start applying frameworks like LAMP, learn the basics of data analysis, and be aware of the different levels of analysis, such as those shown in Figure 1.2, in order to act at the appropriate level.
HR analysts should also develop facility with a number of analytical concepts from economics and finance, including those listed here:
- The differences among fixed, variable, and operating costs
- The time value of money
- Present value and discounting
- The difference between cost-benefit analysis and cost-effectiveness analysis
- The notion of utility as the perceived value of something, where perceived value depends on the value of individual attributes, their probability, and their relationships
- The notion of break-even analysis and inflection points as opportunities for optimization and simplification
It is not our purpose to teach these concepts here but to draw attention to the idea that logic-driven analytics rests, in part, on the strength and validity of the analysis itself. These concepts are not overly difficult, and the easiest way to learn how to apply them to HR issues is to partner with people in the business to whom these concepts are second nature.
That said, as we noted earlier, even the most rigorous analysis that uses the most valid measures can fail to induce change if it is not carefully crafted with the use of logical frameworks that engage the target audiences.
Using Logic to Find the Right “Story”
Organizations have the potential to generate thousands of reports on such issues as turnover, employee attitudes, and skill levels, and to parse the data so as to give personal reports to leaders in every unit, reports that can be further analyzed for all sorts of associations. It might be discovered, for instance, that turnover in one unit is higher among females early in their careers, or that employee attitudes are below national benchmark levels in units that fail to meet their financial goals. Again, however, the goal is not simply to generate interesting HR metrics and analysis but to use analytics to make good decisions and drive change. That distinction bears repeating when the HR function in so many organizations makes the mistake of simply throwing data out to its constituents, in the hope that they will make the necessary connections to important outcomes, root causes, and actions that will produce results.
Thus one important use of logic is to discern which data and which analytics are likely to be most pivotal to the vital issues facing the organization. That’s why HR analysts have to be good at analyzing strategy and business issues and understandin...
