Real Numbers: Management Accounting in a Lean Organization (http://www.lean.org/BookStore/ProductDetails.cfm?SelectedProductId=139) published 13 years ago, discusses concepts of management accounting for organizations deploying a version of the Toyota Production System.
Cunningham and Fiume distill useful advice and experience; this short book is worth your time.
Based on the authors’ experience in the first wave of Lean application in the United States in the 1980s and 1990s, Cunningham and Fiume outline how standard accounting systems mislead or prevent understanding of organization performance. Information that could be useful typically arrives in reports that are hard to understand and too late to matter.
As Cunningham and Fiume say: “We believed we had more to offer than incomprehensible monthly reports; we could provide the information that give businesses a more complete picture of reality.” (p. 4).
While part of the discussion applies only to manufacturing companies—e.g. valuation of inventory and warranty reserves—Cunningham and Fiume offer practical advice for accounting and, more generally, measures and measurement for operations management.
The authors start with the basic pillars of accounting as they lay the foundation to contrast traditional accounting with accounting relevant to a Lean organization:
These four pillars offer insights for all organization measurement as you develop measures and measurements that people can use to control and improve their organization’s performance.
Materiality addresses two features. First, what is the relevance of a proposed measure—do you need to know this information to control or improve the business? Second, how precisely do you need to know a particular number? If your action won’t change given a 10% range in a reported number, then an effort to report a more precise value is immaterial (and in Lean language, waste.) For my clients, this is what I mean when I ask them if a proposed measure is “good enough” for use.
Conservatism “…means you should not overemphasize the good news or under-emphasize the bad news.” (p. 32) The use of run chart and control chart rules for any measures can help avoid claiming existence of trends where the evidence is weak.
Consistency “guides us to present facts in the same manner each time they occur.” (p. 34). Two recent client examples illustrate this pillar. First example: a project to improve performance of primary care practices found it difficult to align measure definitions across clinics, in different health systems. In many efforts to improve performance and understanding, it suffices to have the measures and measurements consistent within each clinic. “Constant bias” makes it possible to see changes over time within each clinic and provides enough consistency to gauge performance.
Second example: colleagues developing a system of patient reported measures (http://www.ihi.org/resources/Pages/Publications/PatientReportedMeasures.aspx) have to wrestle with defining a regular time interval of surveys. They aim to have a fixed interval (e.g. 12 months between screening for depression using the PHQ-2 or PHQ-9 surveys) so that any changes in measures are comparable between individuals and among groups of patients.
Matching at first view looks like it is a purely accounting concept relevant to manufacturing because it is based on an accounting rule: “All costs to manufacture the good you sell must be recognized as an expense in the month you recognize revenue.” (p. 35) On the other hand, usually costs are recognized in the period in which they are incurred. The authors point out that the interplay of these two recognition rules generates a lot of accounting discussion and transactional activity. (Of course in Lean manufacturing organizations, as cycle times from production to sale decrease dramatically and inventories shrink, most costs will just be incurred in the current month.)
Matching in fact relates to a general measurement principle that applies broadly: a measured quantity should align with the time period that spans the actions that generate the numbers. That’s so you can match patterns in the time series with the actions. For example, tracking and displaying “near miss” incidents in time order in an ambulatory surgical center is far more informative than looking at an annual summary of these incidents and then attempting to understand potential causes.