Can we radically improve service contract management?

The government spends more than $500 billion each year on contracts — an enormous sum. So deriving savings from better contracting practices is an obvious target for deficit-cutting efforts.

Like the business leaders two centuries ago who dreamed about how rich they would be if they could sell only one of their products to each person in China, it is tempting to make a statement such as, “If we could save only 5 percent from better contracting, that would cut $25 billion from the deficit.”

We all know anecdotally about unnecessary work performed under service contracts and insufficient cost controls, even if the popular belief that the government spends $600 on a hammer is ludicrous.

The problem is not with the claim that better contracting could save 5 percent. The challenge is figuring out how to do that, particularly in service contracting. For commercial products, there are straightforward ways, such as strategic sourcing and reverse auctions. But achieving such savings in services is the public management equivalent of house-to-house fighting in war: It is not a question of grand strategy or magical directives from the top but lots of individual battles down in the trenches.

To help us make progress on that house-to-house fighting, I propose that we provide more resources to the people doing the fighting — the contracting officers and program officials who manage the contracts.

Many contracting professionals believe that providing more contract management employees can pay for itself — perhaps many times over — in cost savings and, ideally, performance improvement. I am inclined to agree, but nobody knows whether that assumption is correct.

How can we determine whether such a claim is true? We can run experiments. And that’s what I propose we do with the idea of providing more resources to contract management.

Let’s choose some number of service contracts — let's say 20, so we have some chance of having a large enough sample from which to draw conclusions. They should not be IT application development contracts, which have special issues that are already the object of scrutiny by the Office of Management and Budget and others.

They should be contracts that cover an ongoing service and have been around for a while, long enough to have established a track record of cost and performance. They should be cost-type contracts. I think an ideal dollar value would be $5 million to $10 million a year — large enough for cost savings to make a difference but not so large that it would become too expensive to provide them with more management resources.

We should experiment by doubling or tripling the level of contract management resources currently provided. With only 20 contracts in the experimental group, it would be difficult to have too many variations on how the additional resources are used or whether resources are doubled, tripled or quadrupled. But we should think about whether there are specific techniques we’d like to test beyond simply providing more people.

We should then follow those experimental contracts for two years and compare what happens to their costs — and satisfaction with contractor performance or performance metrics, if they exist — with 20 similar contracts that receive no additional resources. If costs go down and/or performance goes up on the contracts receiving additional resources, we should use that approach more broadly in the future.

If there is no improvement, we will need to re-examine assumptions about the impact of adding resources and how easy it might be to cut contracting costs through better management.

Obviously, many details of the experiment would still need to be worked out. But if we’re serious about saving money on contracting, it is an experiment worth trying.

About the Author

Kelman is professor of public management at Harvard University’s Kennedy School of Government and former administrator of the Office of Federal Procurement Policy. Connect with him on Twitter: @kelmansteve

Who's Fed 100-worthy?

Nominations are now open for the 2015 Federal 100 awards. Get the details and submit your picks!

Featured

Reader comments

Sat, Mar 12, 2011

In my opinion, adding more contracting workforce will not solve the problem. It appears to be we already have sufficient numbers of workforces are in place. The extent of problem I see is that half of the current workforces do not know what they are doing. Also, there are no mentors and/or on the job training is in place to support those untrained and/or inexperienced workforce. The other problem is that an individual's commitment and passionate about their work. I believe those are the main problems.

Sat, Feb 5, 2011 Jaime Gracia Washington, DC

Adding bodies to the issue I do not think is a viable solution. The issue to improving service contracts is at the beginning with sounds requirements. If a contract has been improperly structured, if it is the wrong contract type, if proper metrics are not established, then no amount of resources will solve this problem. I think the better experiment is to improve training, and ensure acquisition officials can properly structure, negotiate, and manage contracts.

Tue, Feb 1, 2011

Adding more contract officers may, but probably will not lead to cost savings. Cost savings accrue when Program Managers/Contract Officers have the leadership capacity necessary to create and sustain a team focused on effectively and efficiently managing the contractor relationship. Investments in leadership capacity will yield cost savings, they will be maximized with a combination of leadership capacity plus additional personnel.

Tue, Feb 1, 2011 Larry Allen NoVA

Steve- Good article. I know that Shay Assad and his colleagues are particularly looking at this. I'd be interested in seeing what he says here.

Sat, Jan 29, 2011 Todd

Steve, this makes a lot of sense! This general approach has taken off in a wide range of industries, from development economics, to political consulting, to direct marketing and web retailing. The returns on investment have tended to be large. I think a proof of concept is feasible with reasonably quick turnaround. Two brief comments about the expense and duration of possible experiments in this domain. 1. In terms of cost, I am not sure that useful experiments have to involve devoting a large portion of budget toward non-program costs. There are several elements of cost. First, if the treatment being studied entails expenses that would be paid for anyway (i.e., funding a program, paying for staff, etc.) then it is not fair to account for these costs as purely the result of conducting an experiment. Of course, additional staff, above and beyond what you have currently budgeted and planned, should count as marginal cost associated with the experiment, but those costs are spent on actual program: they are not purely research costs. That said, there are other purely research costs like project design and management, and organizational costs of treating treatment and control differently. In the hypothetical experiment you described those costs would not seem to be prohibitive. An aspect of your concern about cost is the sheer scale of many field experiments. The scale of an experiment depends on the expected effect size and the level at which one randomizes (individual-level, office level, region-level, etc.). The smaller the effect size, or the larger the level at which randomization, occurs, the larger the scale needs to be. In terms of the experiment you described, if you believe the effect size will be relatively large, the scale can be relatively small. Of course, to determine how large an experiment needs to be to reliably detect effects of the size you expect you can conduct a "power calculation." Here's a nice overview of how that works: http://en.wikipedia.org/wiki/Statistical_power 2. Experiments last a long time when the outcome of interest takes a long time to emerge (i.e., lifetime income from subsidizing housing, or college matriculation from elementary school vouchers). It seems like the outcomes of interest in the experiment you described are expected to emerge relatively quickly, suggesting that your experiment could be conducted relatively quickly. Lots of other outcomes of interest likely emerge even more quickly. I hope this approach takes off. Thanks for the post!

Please post your comments here. Comments are moderated, so they may not appear immediately after submitting. We will not post comments that we consider abusive or off-topic.

Please type the letters/numbers you see above