Causation versus correlation: A primary criticism is that correlation does not imply causation. The program identified a statistically significant relationship between profitability and market share, but this correlation does not definitively establish a "true" causal relationship. High market share may yield high profitability, but high profitability could also enable acquisition of market share, or a third factor could cause both. In the multivariate correlation analysis, high market share was associated with high profits, but high profits could have been associated with high market share, or a third factor common to both could have caused the correlation. Many analysts believe that it is possible to use a statistical causality test to determine causation, but if the whole problem is that correlation is insufficient to determine causation in the first place, then how can using another correlation, which is what is used in the tests, determine causation. The program has addressed these concerns through longitudinal time-series analysis enabled by multi-year data collection per SBU, which can provide stronger evidence for causal relationships than cross-sectional data alone, though this remains a methodological limitation of empirical research. The PIMS Cause-Effect Model explores the causal relationships between the relationships between PIMS variables in further detail. → ''In connection with the market share, already indicated and frequent allegations that correlations are used in the PIMS investigations to draw conclusions about causal relationships, i.e. correlation is equated with causality. However, this problem is too obvious not to have been examined in detail during the development of the PIMS program. Backhaus et al. formulate this aptly: "The primary field of application of regression analysis is the investigation of causal relationships (cause-effect relationships), which we can also refer to as 'The more the' relationships". Backhaus et al. (2006), p. 46 (Emphasis in the original.) These authors then add the following: "It should be emphasized here that neither regression analysis nor other statistical methods can prove causalities beyond doubt. Rather, regression analysis can only prove correlations between variables. This is a necessary but not yet sufficient condition for causality." Backhaus et al. (2006), p.48 f. Within the framework of the PIMS studies, it was thus possible to determine causalities with the help of time series analyses due to the availability of data over longer periods. See, for example, Barylite (1994), p. 61. Correlations in this sense, including in the PIMS program, initially give nothing other than a reason to investigate possible causalities substantiated and intensively.''''. →
The PIMS master database at the heart of the PIMS program now includes more than 25,000 years of business experience across a broad spectrum of industries worldwide. These are more than 90% of the companies to be processed. About one-third of them manufacture consumer goods, 15% manufacture capital goods. The remaining business units are suppliers of raw materials and semi-finished products, components or accessories for industry and commerce. Trade and services companies account for less than 10% of total companies and yet represent a fairly large sample (over 250) of strategic business units in this category. About half of the business units in the PIMS database market their products or services nationally in the United States or Canada, while 11% serve regional markets in North America. European companies are also numerous today, with around 1,000 business units from continental European countries and 600 from the UK. Data collection bias: Critics have argued that the database is weighted toward large companies, as smaller entrepreneurial firms are less likely to pay the associated consulting fees and provide detailed survey data.
Mintzberg (1998) claims that because the database features a large representation of large established firms, it is more suitable as a technique for assessing the state of "being there rather than getting there" (page 99). Given the program's ambitious goal of identifying 'laws of the marketplace' that apply across industries, the sampling strategy's representativeness is a key consideration for interpreting results.
Market definition and survivor bias: Tellis and Golder (1996) argue that respondents can describe their markets narrowly to give the appearance of high market share. They believe this self-reporting bias makes conclusions suspect. They also highlighted concerns that no defunct companies were included at the time of their writing, leading to "survivor bias". PIMS practitioners counter that, aside from the act of inflating market share figures being counter-productive for participants to receive useful insights and thus not popular, PIMS consultants have continuously been involved in the data gathering process, ensuring data quality and veracity.
Homogeneity assumption: A criticism of the PIMS program concerns its reliance on a homogeneity assumption — the idea that all firms, regardless of industry, are drawn from a single underlying statistical population. Critics argue that cross-sectional econometric models implicitly assume identical distributions across industries. If this assumption is false, then pooling firms from heterogeneous industries obscures structural differences in competitive dynamics, and the estimated “laws of the marketplace” may be invalid. However, this critique overlooks the central purpose of the PIMS database: to identify broad, cross-industry regularities in the determinants of strategic performance, using variables that exist and apply to all industries. Empirically, PIMS has consistently shown that firms with above-average market share, product quality, labour productivity (and others) achieve higher returns across a wide spectrum of sectors. These relationships hold in industries as diverse as automotive, food manufacturing, and heavy industrial production, suggesting the existence of market-agnostic strategic principles grounded in universal economic mechanisms such as economies of scale, perceived quality advantages, and productivity-driven cost efficiencies. Thus, while industry heterogeneity undeniably exists, the stable cross-industry trends identified by PIMS weaken the claim that pooling data is inherently inappropriate, and instead support the view that PIMS captures strategic patterns that are sufficiently generalisable to be analytically meaningful. ==PIMS and pims.ai==