Consulting Best Practices
David Morganstein is a vice president and the director of Westat’s statistics unit of 65 master’s and PhD statisticians. He has consulted, published, and taught about survey sample design, weighting and analysis, and statistical quality control for more than 40 years. He is a recipient of the ASA’s Founders Award and a Fellow of the ASA.
In February, the ASA held the first of what is hoped to be an annual conference on statistical practice. This article summarizes a presentation on statistical consulting, one of the panel topics.
What are the characteristics of a good statistical consultant beyond technical expertise? A good place to start is with the characteristics suggested by practitioners and clients. Among the most important are good interpersonal skills, the ability to and interest in listening carefully and fully to the client, and an appreciation for the client’s concerns. While these may be considered low tech compared to the latest algorithm, they are necessary and important characteristics for successful consulting.
What is the purpose of statistical consulting? It might be as simple as helping understand a perceived problem or it might be as complex as identifying and solving a problem and implementing a solution. Possibly, the client might want the consultant to not only recommend, but also implement the improved method via software training or changes in the process. Each of these possibilities requires additional skills, and virtually all require understanding how to communicate effectively at and with all levels of an organization.
Many good consultants will suggest something like the following list of skills and characteristics:
- Time management
- Team player
- Good communication
- Good listening
- Roles and responsibilities
- Involving other consultants
Martin Ashford surveyed London Business School alumni to find out what skills they thought were most prized in consulting and noted the following responses:
- Skill in structuring task
- Technical knowledge/skills
- Industry experience
- Commitment to clients
- Getting along with clients
In Steven Covey’s widely read Seven Habits of Highly Effective People, you will find this advice: “Seek first to understand, then to be understood. Communication is the most important skill in life.”
An important element of good communication between the client and consultant is defining the boundaries of the work. It’s as important that you indicate what you won’t do as it is to describe what you are committing to do. Of course, once you get into the project, you may see something that wasn’t obvious from the outset. You can always provide optional steps that would extend the range of the original agreement and that may require additional resources.
In almost any consulting task, assumptions must be made, as few projects will be so obvious that no assumptions are required. It will be to your credit to clearly identify and state these assumptions and, to the extent possible, emphasize their implications. In the survey world, for example, assumptions are needed about response and eligibility rates and about respondents’ ability to recall events. These are rarely known precisely when a study is designed. As one example, deviations from your assumed values may have an impact on the total survey error, and the effective consultant will convey this to the client. This, too, is about clear communication.
Communications occur in many directions, all of which are important to the statistical consultant. The most basic line is between you and your client. When you are trying to understand the processes that affect your problem, there are also communications between various members of your client’s organization—between staff and management, between departments, between new staff and old hands, between research units and operations, between members of a team or committee, and between you and the processes you are trying to improve.
One of W. Edwards Deming’s 14 points was to break down barriers between departments. His rationale was to prevent the disruption of critical communications and to facilitate the exchange of information between coworkers who depend upon each other to improve the quality of the work and, most importantly, who are essential in implementing the changes recommended by the consultant. As statisticians, we appreciate the value of good data, which can only come from clear and accurate communications.
The most basic communication is between the statistical consultant and the problem we are addressing. We use our statistical tools to see the problem and potential solutions in ways others may not be equipped to. Irving Burr, a consultant in statistical process control, offered that the process is speaking to us all the time, but do we understand it? Statistical tools as simple as the Shewhart diagram and as complicated as multi-level modeling help the consultant understand what the process is saying.
Some of the ‘softer’ tools for improving communication include agendas, flowcharts, checklists, and minutes from meetings. Agendas and meeting minutes, when done well and in a timely fashion, can be a real help in communicating between people trying to solve problems. Flowcharts and checklists also help in clarifying change, whether through developing software or modifying processes.
Whether you are consulting within your own organization or invited in to another organization, you’re likely to spend lots of your time in meetings. You may be the planner or the participant. Either way, an effective consultant wants the meetings to be efficient. Following are several important steps in ensuring meetings are efficient:
- Include the right people
- Set an agenda
- Stick to the agenda
- Follow up the meeting with minutes
It might seem surprising, but taking time to review the problem and identifying people with important inputs, and then making sure they attend the discussions, is an easy step to overlook. Creating an agenda of the key discussion points and distributing it in advance can help encourage active participation and lead to a focused discussion. Taking and distributing minutes that summarize key points of agreement and disagreement—including action items and next steps—and identify the responsible person and a tentative due date helps move from the meeting to problem resolution.
Since costs are an inevitable component of any proposed solution, it’s reasonable to ask how much is enough and what is good enough. Morris Hansen, a respected and wise statistical consultant, had the following quote by Voltaire on his office wall: “The perfect is the enemy of the good.” Morris was all too aware that virtually no one could afford the cost or time to arrive at a ‘perfect’ solution to a problem. Knowing what is too much effort to expend for the next improvement step is a valuable attribute of a good statistical consultant.
If part of your task is to implement changes resulting from your evaluation, you will be well served to consider the reward system at work in the client’s environment. The most efficient solution may never work if the people who need to bring it about do not benefit in some way from the change. Even more basic, many have an understandable motivation to continue using tried-and-true methods. It may be important to ensure the staff is involved in the development of the new procedure and ‘buys in’ to the new methods. It may be critical that managers also are involved, agree with the approach, and convey support for the changes, demonstrating rewards for those who help in a successful transition.
Finally, never forget that ethical issues undergird all our work. Following accepted ethical practices is a necessary requirement for any statistical consultant. Our professional societies (e.g. the ASA and International Statistical Institute) have established and published acceptable codes of ethical practice. Among other characteristics, they require pursuing objectivity, clarifying objectives and roles, assessing alternatives impartially, and avoiding conflicting interests and pre-empted outcomes.
The importance of good communication and the need to clarify assumptions and understand the reward systems can be seen to underlie many of these basic principles.
A final word about a few pitfalls: Beware of the type III error, “the right answer to the wrong question.” Always remember that part of a statistician’s continued professional development is to ensure we broaden our tool kit. “If the only tool you have is a hammer, don’t be surprised if everything looks like a nail.”
And since I enthusiastically believe fairy tales are not just for kids, I remind you of the story Emperor’s New Clothes, a tale that has been retold over the ages, yet is of value in our profession as well. It’s a simple summary of how not understanding the reward system can lead to a breakdown in communications and the absence of the very data needed for good decisions.
Ashford, M. 1998. Con tricks: The shadowy world of management consultancy and how to make it work for you. UK: Simon & Schuster.
Burr, I.W. 1976. Statistical quality control methods. New York, NY: M. Dekker.
Covey, S.R. 2004. The 7 habits of highly effective people: Restoring the character ethic. New York, NY: Free Press.
Deming, W.E. 2000. Out of the crisis. Cambridge, MA: MIT Press.
Ramakrishnan, N. (n.d.). 10 winning qualities of consultants.