How a Budgeting Problem May Not Really Be a Budgeting Problem.

 E-mail this page

Elsewhere on our site, we criticize firm-wide, over-reliance on a single, uniform performance measure like profit-sharing or share price.  The intention may be good, but the result need not be.  The basis of our criticism is the presumption that most employees are risk averse, and for most of them there exists other, better performance measure that are statistically more significant than a global one for the firm.  Using more sensitive and precise measures allows the organization to either: (1) induce the same types of effort or decisions at a lower compensation cost (by reducing the necessary risk premium in compensation) or (2) induce more goal-compatible actions and decisions at the same expected compensation cost.

Combining that criticism with this current essay reminds us of Goldilocks and of the benefits of moderation—i.e., not too few, not too many.

We recall a meeting and conversation with a senior finance officer a few years ago, where he complained about the budgeting process at his firm, MegaCorp, with its obviously made-up name.

In his industry, mergers and consolidations were frequent because there were substantial economies of scale for most service processing activities.  In fact, the promised savings from those economies were the main justifications for the wave of mergers.  By centralizing the services, if MegaCorp had accurate demand planning, it could realize huge cost savings and equally large increases in profits and market value.

Unfortunately, the firm’s actual demand was rather volatile and often very different than planned.  There were periods of substantial excess capacity and periods of severe capacity constraints, and in both extremes the benefits of consolidation were eliminated if not greatly reduced.  To say that coordination was lacking was an understatement.

Our acquaintance blamed the firm’s regional presidents (RPs) for this predicament.  Many RPs had run the smaller organizations that had been acquired by Mega.  Because of their equity stakes in the acquired companies, they had benefited from the mergers, but most other MegaCorp employees hadn’t, yet.

He noted that the RPs were particularly poor predictors of both new customers and of the resource usage growth due to existing customers.

He had his own theory about why they were poor prognosticators.  He stated that many of these individuals had become quite wealthy when they sold their small, regional firms to MegaCorp and had assumed their current positions.  He also mentioned that many of these folks were visible members of their communities. He argued that since they had acquired their wealth, their goals had changed.  They were now focused on maintaining their personal prestige rather than running their businesses as dedicated MegaCorp employees.

Curiously, he did admit that personal relationships were crucial to the firm’s success in these smaller markets, and that caused us to wonder whether any of his incentive hypothesis was based upon envy.   At the time, the regional president had made their millions, and our acquaintance had not.

Given his theory, he was convinced the budgeting process was broken and needed to be fixed.  He wanted our advice about how to fix it, and we agreed to help because we knew that regardless of the validity of his theory, the budgeting process could still be broken.

We understand that there are different types of people: high types, low types—effective and ineffective ones.  We aren’t known for viewing the world through rose-colored glass, and we can easily attribute any level of success—outside of our own—to luck rather than to ability.  However, many of the regional presidents had successfully run their smaller organizations in the past without the benefit of the low costs resulting from national economies of scales.  So, it seemed unlikely that the inaccurate forecasts were due to sheer incompetence.  Moreover, if they were wealthy enough not to care about work, they could likely quit; so, neglect did not seem to be a compelling reason either.

It could have been an issue with the budgeting process.  Poorly designed cost allocation schemes or transfer pricing systems can cause wild swings in internal demand estimates—even in stable environments.  Plus, the industry was growing slightly, not shrinking.  We wondered about the usual problems of eliciting private information, but decided to start where we always do—at the beginning—by asking, “What do you want the regional presidents to do?”

We thought that given the justification for MegaCorp’s multiple recent mergers and acquisitions, the answer would be short and simple (and crucial): make accurate forecasts…of the volume of shared processing resources, either directly or indirectly by forecasting sales.

We were wrong—very wrong—by two orders of magnitude.  (No, we were not wrong about how he should have answered the question, but our prediction of his answer was incorrect.)

His initial answer was, “It depends.”  Having sat through a few of our seminars, he thought it was clever response—his mimicry of our classroom mantra.  As it turns out, his response was truthful but his reasoning was not wealth-maximizing.

MegaCorp used a management-by-exception policy for the regions.  Every month corporate headquarters would inform the regional presidents of the three-to-five performance measures that appeared most out-of-control in their locations and would instruct them to fix the problems causing these variations.

We asked about the total number of performance measures.  He replied that he hadn’t counted them all, but there were many.  So, we asked to see a comprehensive list of performance measures.

We received six pages with almost 70 rows of measures (of small fonts) on each page. ¹ Yes, there were approximately 400 “performance” measures, and many of these measured various aspects of the region’s operating performance and efficiencies—there were many ratios and variance calculations.

Now, before continuing it is worth noting that the probability is very low that one could find 400 independent and relevant variables measuring distinct aspects of a person’s performance.  It was more likely that many of these variables were highly correlated with one another or that many variables were actually dependent variables—linear combinations of others, particularly if they were extracted from accounting reports.  However, this point isn’t crucial to the case.

We asked, “Which of these do you emphasize?”

His response: “That depends, there are usually three or four—sometimes five—that we emphasize each month. Usually, we won’t have four different issues.  Instead, we’ll get multiple signals of one or two underlying problems.”

“Are they the same signals for everyone?”

“No.”

“Are they the same measures for the same people every month?”

“Uh, no”

“So, every month, you are telling most of these folks to focus on different aspects of their region’s performance?”

“Well, yes, we’re just showing areas for improve.” 

We asked, “How do the regional presidents know what you want them to do?”

We continued, “There is a good chance that they are confused by headquarters monthly drawing of concern factors.  Moreover, there is a good chance that they would do what you want them to do if they knew what it was, but to them it seems that the organization cannot make up its mind.”

“But that is not true.  We know what we want.  We want them to make accurate forecasts and take care of problems that arise.”

“You might say that, but your actions speak louder than words.  Twelve times per year you are asking the RPs to (possibly) redirect their energies to “new” problems.  In fact, if there are no glaring weaknesses in a region, then it is more likely that you are inducing them to run willy-nilly to different illusory problems each month.  (We’ll explain that later with our recommendations.)”

He disagreed.  “That doesn’t make sense.  We spent a lot of time and money developing our reports, and every month we employ substantial resources to monitor performance.  In fact, we have several analysts compile and review the reports.”

“The Soviet Union did, too, but that’s not the point. First, note that if there were glaring problems, then you would likely see serial correlations among the out-of-control measures.  That means that problems would likely persist through time; this month’s troubles would be like last month’s.”

“We concluded that we saw no persistence because we were managing effectively.  In other words, through our data mining, we would identify problems, and the regional managers would rapidly fix them.”

“Well, supposing that your conclusion is true, your measurement system is still suboptimal.”

“How can you say that?”

“Very easily by asking our first question, again: what do you want them to do?”

“Make accurate forecasts and fix problems.”

“So, among the three-to-five performance variables, where is the emphasis on forecasts?”

“Well, it is not explicit, but it is in there.”

“Yeah, it probably is in there, but with 400 variables everything else is in there, too, and it appears that you are giving those other factors equal if not more-than-equal weight.

As you know, many of the factors relate to operating efficiencies.  Let’s suppose that each month your analysts identify real problems, and as we will explain later, that is a supposition at this point, but let’s go with it.

At the margin, correcting those problems will increase Mega’s profitability but only marginally.  Your current scheme makes the firm pennywise, but pound foolish.”

He interrupted, “You know, someone did mention we don’t seem to make a lot of progress despite all of the activity.”

“Well, yeah.“  We continued, “Could the acquisitions have been justified on the basis of eliminating all regional inefficiencies—excluding the gains from the centralization of processing services?”

Having been involved in the analysis of each of the acquisitions, he knew the answer was no, and replied, “No, the crucial savings involved the centralization of service processing, but the regions still need to be well-managed.  So, what do you recommend?  How do we get accurate forecasts for central activities and still manage regional performance?”

“We would split the RP positions into two jobs: the one called regional president dealing with sales, forecasting, and community involvement and one called regional chief operating officers (RCOOs) devoted to day-to-day operating issues.  Given what you told us, it’s likely that most regions already have someone serving in this role, albeit informally.”

In addition, we recommended that the current RPs remain in their current positions, but that RCOOs report directly to Mega’s headquarters.  We did recommend that the RCOOs report on a dotted-line basis to the RPs, too.  We viewed this formality as an overt and necessary method to ensure that the RPs had access to any relevant information possessed by the RCOOs and to ensure that the RCOO’s recognized their ongoing responsibility along this dimension. ²

The regional presidents could then focus on their key tasks, and we could develop ways to elicit their private information that included a careful analysis of cost behavior, existing capacity utilization, and anticipated changes in technology to determine how MegaCorp wanted the RPs to act.  Eliminating most, if not all, operating performance measures already greatly simplified the planning problem. Lastly, with respect to the RPs, we mentioned that Mega could offer different rewards depending upon the market characteristics of the regions, e.g., in smaller and stagnant markets, gaining the RPs’ differential information was generally less valuable and likely not worth the cost.

In addition, we recommended a more careful analysis of the important tasks performed by RCOOs and a stream-lining of their performance measures: “Suppose that the existing 400 operating performance measures are independent random variables and each one is influenced by a separate, independent action taken by the employee performing 400 different, worthwhile, non-repetitive actions per month—seems ambitious, but just suppose.  Furthermore, suppose that the employee takes the 400 actions that MegaCorp prefers.  ³

If the monthly review investigated any measure that was exceptionally bad, say, in the worst one percent of that measure’s possible values, then, on average, there will be four exceptional measures per month.  ⁴  So, the worker is asked to change his action despite behaving appropriately in the first place.  ⁵

Now, suppose there are only 100 independent random variables and the other 300 are deterministic functions of these variables.  For simplicity, assume that each of the 100 has three related, dependent variables.  In this case, there would be one true exception per month and three violations of related measures.  The additional three provide no additional information; so, 75% of your measures are a waste of time.  If the relationships aren’t deterministic, there is a possibility of more noise and worse inferences, which we would observe as recommendations to fix non-existent problems.

So, the current shotgun approach may be comprehensive, but a more careful approach that is consistent with Mega’s priorities will be more profitable.  That means start by asking: what does Mega want the RCOOs to do and what signals exist of this effort?

Also, given differences in the markets that Mega serves, there is no reason to believe that the answer to this question is constant across regions.  MegaCorp should prefer the flexibility of using different measures or different weights on the same measures depending upon regional market characteristics.

Fortunately, with 400 existing measures, we have a solid platform from which to start.

So, we are now able to solve your two most pressing problems.  The first involves eliciting the RPs’ private information for planning purposes and that is a much smaller problem now—in the sense that we can now trade efficiency gains of managing centralized capacity against opportunity costs of insufficient capacity and our estimates will be more precise.  The second involves preventing operating problems within the regions and quickly fixing those that do occur.  By elevating individuals to formal RCOO roles, your compensation costs may increase slightly, but that marginal cost is a small fraction of the benefits that should be derived by improving the coordinated use of shared resources.”