Organizing Business Knowledge The Mit Process Handbook [Electronic resources] نسخه متنی

12.4 Case Example — Generating Innovative Ideas for the Hiring Process

To illustrate the use of our methodology, we will use examples based on a field study we conducted in collaboration with one of our corporate research sponsors, the AT Kearney consulting firm, and one of their clients, which we call Firm A to preserve the client's anonymity (for more detailed descriptions of this study, see Malone et al. 1999; Kruschwitz and Roth 1999).

Firm A was experiencing increasing problems with their hiring process. They were growing rapidly in a tightening labor market, and they had a culture of independent, competitive business units. Together, these factors led to increases in the time and cost to hire people and to increasingly frequent instances of business units ''hoarding''candidates or bidding against each other for the same candidate. In an effort to improve their hiring process, the organization had invested a great deal of time and energy into ''as is''process analysis using conventional techniques such as flow-charting. But they also wanted some way to come up with highly innovative ideas about how to improve their process. We next investigate how each of the stages of our methodology can be used in this situation.

12.4.1 Analyzing Deep Structure

The purpose of deep structure analysis is to identify the core activities and key dependencies of the process one wants to (re-)design. In the absence of the Handbook database, one could find a process deep structure by starting with a candidate process (e.g., the process that is currently used, if one is re-designing an existing process) and repeatedly replacing its activities and dependencies by more abstract versions that represent ''why''the surface activities are there. Eventually one will end up with a set of core activities that appear essential to the process, connected by a minimally suffcient set of dependencies.

The Handbook database can greatly simplify this procedure by allowing us to find the process(es) in the database that are most similar to the initial target process. The processes to the left of that point then represent increasingly deep structures for the original process, and those to the right represent alternative surface structures.

In Firm A's case an ''as is''model of their hiring process had already been developed before we began working with them. Their ''as is''model consisted of the four steps ''identify need,''''source and select,''''enrollment,''and ''physical installation.''Since our Handbook database already has a generic ''hire''process (see figure 12.5), it is straightforward for us to treat Firm A's process as a specialization of this more generic one. One immediate insight from looking at this representation of the deep structure of the hiring process was that the initial ''as is''diagrams developed by Firm A had left out the step of 'Pay employee'. When the employees of Firm A saw this representation in the Handbook, they agreed that they should have included this in their initial ''as is''analysis.

Figure 12.5: Deep structure for 'Hire'. The arrows represent the flow dependency among the components.

12.4.2 Generating Alternative Surface Structures

The 'Generate alternatives'step involves generating the set of surface structures that represent potentially viable candidates for achieving a particular deep structure. This is done by identifying the dimensions along which the surface structures can be varied, identifying all the values for each dimension, and then considering some or possibly even all combinations of these values. We thereby define a multidimensional design space in which every point represents a potential surface structure.

Recall that a process consists of sequenced activities inter-related by dependencies, and that dependencies are managed by coordination mechanisms. There are thus three main dimensions along which alternative surfaces structures can be generated: (1) alternative specializations for a given activity, (2) alternative coordination mechanisms for managing a given dependency, and (3) alternative sequencings for the activities in a process:

Alternative Activity Specializations Specializations of an activity can be found by generating answers to key questions about the activity such as who performs the activity (i.e., the actor), how the activity is performed (i.e., the activity decomposition), where the activity is performed, and when the activity is performed. Using the Handbook, one does not have to imagine all the answers to these questions unaided. Instead, one could browse the specialization tree to the right of the activity to uncover a potentially large number of alternative variations and examples of the activity. One could also look at the alternative specializations for all the parts (and subparts) of the activity that is of interest.

For example, table 12.2 shows a number of examples of ''interesting practices''represented in the Handbook as specializations of the 'Select human resources'part of hiring. The employees of Firm A found several of these examples to be quite intriguing stimuli for innovations they might try.

Alternative Coordination Mechanisms The space of alternative coordination mechanisms for a dependency can be found, as with activity specializations, by generating potential answers to a set of key questions. In this case, however, the key questions include:

What type of dependency is involved (i.e., flow vs. sharing vs. fit) (Malone et al. 1999).

When are the activities performed? Options here include the source activity for the dependency must end before the target activity starts, or the source and target activities can overlap in time. These timing options can be formalized using a temporal logic such as that proposed in (Allen 1981; Lee et al. 1998).

Where do the activities occur?

What type of resource gives rise to the dependency? The alternatives can be selected from a resource type taxonomy (Fadel et al. 1994; Lenat 1995) and can include such options as divisible or nondivisible, consumable or nonconsumable, and so on.

How much of the resource is involved?

As with activity specializations, a preexisting knowledge base of coordination mechanisms can greatly simplify the identification of alternatives by providing a set of possibilities. In the Firm A hiring process, for example, there is a dependency between identifying the staffng need (typically done by a manager) and finding candidates that can satisfy that need (done by recruiters). The coordination mechanism typically used is a requisition form sent by the manager to the recruiters on an ad hoc basis, and thus is a kind of 'Make-to-order'process.

An alternative coordination mechanism, accessible as a sibling of 'Make to order'in the Handbook database, is 'Make to forecast'. This process suggests that we create staffng requisitions in response to an overall business plan instead of based on individual manager's requests. Another possibility is suggested by looking at ''options markets,''which is a specialization of 'Make to forecast'. The notion here is that we can requisition items (in this case employees) when the item is inexpensive, in anticipation of needing the item later when it may be more expensive due to greater general demand.

Alternative Orderings Another dimension for generating new processes is to reorder the sequence in which the activities occur. The ordering of the activities must, of course, satisfy the ''core''prerequisite dependencies inherited from the process deep structure, but aside from this we have complete latitude to change the surface structure dependencies and, therefore, the activity ordering. In the hiring process, for example, the only core dependency is that selection occurs after sourcing. We can re-order the other activities to suggest novel alternatives; we can, for example, place the 'Install employee'step before the 'Enroll'step, which implies that we install the employee (i.e., for a trial period) before deciding to hire him or her.

Keeping Track of the Alternatives Generated These methods for generating process alternatives can be applied recursively, in the sense that the new activities and dependencies generated can in turn have further subalternatives defined in the same way. We wind up, in any case, with several dimensions of variation (one for each dependency and activity, plus one representing alternative re-orderings of the activities) plus one or more values for each dimension.

We have found that a multicolumn menu represents a convenient metaphor for deriving all the surface structures that can be defined given the alternatives defined by the steps above. An example is given in table 12.3. Each column represents a dimension of variability, and the items in the columns represent choices along that dimension. Process alternatives are then generated by selecting one (or sometimes more) choices from each column. If one does this exhaustively, one can typically generate very many alternatives from all the possible combinations involved.

While these tables can always be constructed manually using knowledge from the Handbook, we have also developed a specialized software tool, called the Process Recombinator that automatically generates such tables and then automatically creates specific combinations based on the user's selections (Bernstein, Klein, and Malone 1999). Figure 12.6 shows a screen shot from this tool showing the selection made for each of the dimensions.

Figure 12.6: Subactivity recombinator user interface

12.4.3 Looking Even ''Deeper''

So far we have seen how one can go to the most obvious deep structure representation of a process (''hiring''in this case) and then generate alternative surface structures. Sometimes the most interesting ideas, however, come from looking at deep structure representations that are even ''deeper''than the most obvious ones. In the case of hiring, for example, if we look further left in the specialization tree we see that ''hiring''is classified as a specialization of ''buying''(figure 12.7).

Figure 12.7: Specialization tree generalizations for hiring process

Based on this observation, we can then get even more potentially interesting ideas by examining some alternative specializations (surface structures) of the generic ''buy''process. For example, table 12.4 shows some alternative ''interesting practices''for buying represented in the Handbook. These examples suggest, for instance, the possibilities of having different hiring processes depending on the kind of employee (by analogy with Acer's strategy) or using corporatewide hiring standards ( la Motorola).

One can go even farther, for example by looking at specializations of 'Sell', the sibling to 'buy', and the great-uncle to 'Hire'in the specialization tree (table 12.5). These suggest such interesting ideas as letting a company 'Test-drive'a potential employee ( la Cessna/BMW) before making a full-fledged hiring commitment, or data-mining a database of comments from people who applied to work at a company in order to develop a more effective process of ''selling''the company to potential employees ( la New Pig). This process can be carried all the way left in the specialization tree to extremely generic activities: such as 'Create', 'Destroy', 'Modify', 'Preserve', 'Combine', and 'Separate'. If we consider these as means of hiring, we come up with options like:

Hire by creation. Breed employees (as in family-owned businesses or monarchies).

Hire by destruction. Eliminate unusable employees. Hire everyone, and let go any who don't pass muster (as in the Armed services).

We refer to examples like these as ''distant analogies''; such analogies represent a particularly powerful property of specialization trees as applied to process design. Generally speaking, the more distant an analogy we consider, the more creative the innovations we are likely to uncover, but the greater the risk that the process idea may prove inapplicable to our problem.

In this case, when we originally did this analysis (several years ago), the idea of buying via the Internet suggested the possibility of hiring via the Internet. Even though this now seems obvious, at the time it was a very novel and interesting idea. Some time later, we generated another—even wilder—idea based on this case example: What if, just as there were on-line auctions to ''buy''things, there could be on-line auctions to ''hire''people? For several months, we used this in presentations as an example of a wild idea, generated by our methodology, that might someday be useful.

Within less than a year, however, we began to hear about a whole new category of very well-funded Internet startup companies to do exactly this: to provide on-line auctions for employees and contractors! This example therefore provides at least an ''existence proof''that our methodology can be used to generate new ideas that have significant practical potential.

12.4.4 Selecting Alternatives

Once a number of alternatives have been generated, one needs to select the ones that appear best suited to one's particular challenges. There are many ways this can be done, but we found that the following approach works well. First, we identify the requirements we are attempting to satisfy by uncovering the key variables and their desired values. In the hiring domain, for example, we found that the key variables included the type of person we want to hire (one with widely available ''commodity''skills rather than a senior person with highly unique skills), the quality of the employee hired, the cost and speed of the hiring process, and so on. This analysis thus uncovers the different requirements sets (in this case there are two), each of which can potentially be achieved using different hiring processes. We then use trade-off tables and other tools to help assess the relative worth of the alternative surface structures implied by the multicolumn tables.

For example, let us consider the case of hiring senior employees. Since we do not want to explore the many potential alternatives blindly, it makes sense to identify the top candidates from each column and then consider at first just the different combinations of those options. We can do this using trade-off tables. The trade-off table for the 'Identify potential sources'activity, for example, is shown in table 12.6. This table suggests that relying on the Search Firm maximizes candidate quality, but using the Internet has advantages in terms of the speed of search and the breadth of access.

If we continue this analysis for all the columns in the table, we come up with two top candidates for surface structures for the senior employee hiring process, one where the applicant approaches the firm and the other where we are finding replacements for an existing position (figure 12.8). We can then produce a trade-off table that compares our two candidate hiring processes with respect to the requirements we identified earlier (figure 12.9).

Figure 12.8: Two alternative surface structures for the senior employee hire process

The entries in the trade-off tables can be generated by combining information from the trade-off tables entered for the component alternatives in the multicolumn tables, by soliciting the judgment of specialists in the process domain (Human Resources experts, in this case), and potentially even by simulating the candidate surface structures using simulation tools. The Handbook includes an evolving interchange format called PIF (the Process Interchange Format) (see chapter 21 in this volume) whose intent is to allow Handbook process descriptions to be exported easily into existing simulation and other process analysis tools. Note, however, that this approach currently produces potentially applicable but not necessarily viable combinations; there may be interactions between choices in the different columns that make a given set of choices unworkable. The methodology does not automatically prune out invalid combinations for you—this requires human judgment.

Figure 12.9: Trade-off table for 'Hire'process alternatives