The graphic above, adapted from Rogers, Diffusion of Innovations (5th ed. 2013), shows the distribution of innovativeness among 324 German banks.  The innovativeness scale is a count of innovation adoptions from a universe of 12 interactive telecom innovations that were diffusing through the German banking sector during the early 1990s. To help distinguish the early adopters, more recent innovations were weighted more heavily.  The distribution is a textbook example of the Rogers Diffusion Curve. The long innovators tail exists because innovators are typically 2+ standard deviations from the mean on innovativeness.

Rogers uses the German bank study to illustrate numerous factors associated with higher levels of organizational innovativeness.  One factor is size.  Specifically, the largest German banks accounted for a large proportion of the innovators and early adopters.  In Rogers’ dataset, the correlation between innovativeness and total assets was a remarkable .75 (p < .01). Likewise, the correlation between innovativeness and employee headcount was an equally stunning .70 (p <.01). If readers are wondering why I am surprised, it is because the results are contrary to the standard trope that larger, more mature, and more financially successful companies — be it manufacturing, pharmaceuticals, or technology — struggle with innovation.  Indeed, this is the very problem Clayton Christiansen is trying to solve in the Innovator’s Dilemma (1997).

Post 016 is part of Legal Evolution’s foundational series on diffusion theory.  Readers seeking to influence innovation within the legal industry will be more successful if they obtain and apply this background knowledge. Care has been taken to make this information non-technical and accessible.

This post is Part II of a three-part series on innovation in organizations. See Post 015 (Part I of series).  The goal in this post (016) is to unpack the counterintuitive relationship between size and innovativeness, as the strategic takeaways are far from obvious.

Deft Minds and the Size Effect

Everett Rogers had a remarkably deft mind that could puzzle through seemingly contradictory data and, with enough time and reflection, derive the most plausible causal story. Chris Zorn, my fellow co-founder at Lawyer Metrics (now LawyerMetrix), has a similar rare ability, which is to say I have some hands-on experience in this area. Drawing upon this experience, let me gently set reader expectations: What is important in Posts 015-017 is analytically subtle in a way that is not intuitive for most lawyers.

Let’s start with the size effect, which is present in Rogers’ study of German banks.  The size effect is relevant to lawyers because (a) there is credible, recent evidence that size is correlated with innovativeness in law firms; and (b) as Rogers acknowledges, the higher levels of innovativeness are, in most cases, substantially driven by the “covariants” of size, rather than size itself.  It is this second point (b) that requires the deft researcher’s mind, albeit it can definitely be grasped by patient smart people. So hang in there.

(a) The Altman Weil Law Firms in Transition Survey

The recent credible evidence of the size effect comes from the Altman Weil Law Firms in Transition 2017 survey, The survey polled managing partners and chairs at 798 US law firms with 50 or more lawyers. The response rate was 48%, including 50% of the NLJ 350 (ranking based on lawyer headcount) and 50% of the AmLaw 200 (ranking based on gross revenues).

Ron Friedmann in his post “Law Firm Profitability + Service Delivery: What the Altman Weil Survey Says,” Prism Legal, June 21, 2017, conducted a masterful secondary analysis of the survey results. One of the survey questions asked law firm leaders, “Technology tools that incorporate artificial intelligence (AI) and machine learning — like Watson and Ross — are beginning to be adopted by some law firms.  What is your firm’s stance on the use of legal AI tools?”  From the respond data, Ron generated the chart below:

This graphic shows a very strong relationship between law firm size and the use of AI.  Over 50% of the 1000+ lawyer firms claim to have begun adoption. Further, the effect is clearly linear, with the level of use and exploration steadily declining with firm size. Drawing upon what we learned in Part I (015), the orange bars reflect the “initiation” phase and the blue bars reflect “implementation.”  (Keep in mind that the implementation phase is fraught with difficulties and often ends in failure. See Post (015).)

Further, the relationship between size and innovation is not limited to AI.  We observe the same size/innovation effect in results that show the linkage between alternative fees and changes in how work is being staffed and delivered.  The graphic below is also courtesy of Ron Friedmann:

[Query: Why is linking AFAs to staffing and service delivery so innovative? Because the real value of alternative fees is to incentivize a re-design of workflow that (i) increases quality, (ii) speeds up delivery, and (iii) decreases cost. Otherwise, alternative fees become either a price discount or a gamble with poor or unknown odds. Stated another way, there is no point in hiring a pricing specialist unless you’re also going to hire specialists in project management and process improvement.]

(b) Covariants to size, not size itself

The graphics above reveal a clear and meaningful relationship between size and innovativeness. Yet, it does not necessarily follow that increasing size will increase innovation.  Correlation, as they say, is not causation.  Instead, it may be the case that other actions or activities need to be taken to improve or enable innovation; and for a variety of reasons, those actions or activities are more likely to occur in a larger firm.

To illustrate, let’s return to Rogers’ Organizational Innovativeness model, which is reproduced below:

In category II, Internal Characteristics of Organizational Structure, there are six factors (i.e., independent variables) listed, with size being number 6.  In Diffusion of Innovations, Rogers asks the question, “Why do researchers consistently find that size is one of the best predictors of organizational innovativeness?”  The first reason, writes Rogers, it that size is easy to measure with precision and thus “is included for study in almost every organizational innovativeness investigation” (p. 411).  Rogers continues:

Second, size is probably a surrogate measure of several dimensions that lead to innovation: total resources, slack resources (defined as the degree to which an organization has more resources than those required for ongoing operations), employees’ technical expertise, organizational structure, and so on. These unidentified variables have not been clearly understood or adequately measured by most studies. These “lurking” variables may be a fundamental reason for the common finding that size and innovativeness are related (Id.).

These “lurking” variables are covariants — i.e., attributes that generally move in a linear relationship with one another, either positively (height and weight) or negatively (age and memory).  Thankfully, when we have a lot of potentially meaningful variables that are correlated with one another, we can sort out what matters, by direction of effect and magnitude, through multivariate models.  In this case, factors 1-5 in Rogers’ model reveal the more valuable insights. We will carefully review those factors in Part III (017). But before we do that, let’s make sure we know enough about the underlying statistical models to avoid very serious errors in judgment.

Size is not a strategy

In the quoted paragraph above, Rogers is describing what statisticians call “omitted variable bias.”  This occurs when we leave something out of a multivariate model — like factors 1-5 above — and get a result that inflates or deflates the predictive power of the remaining variables (like size of firm).  The risk here is that folks running and interpreting the models are not sophisticated enough to know that they might have left something out and, if so, how to correct for it.  In turn, they drawn incorrect inferences that could form the basis for disastrously wrong strategy.

An example of the subtleties at play here can be seen in a law firm profitability model that Evan Parker (Managing Director of Analytics at LawyerMetrix) and I published in The American Lawyer. See “Playbook: Top 5 Strategies of the Most Successful Firms,” January 2017. As a first-cut analysis, size is positively correlated with average partner compensation: .53 with revenues, .28 with headcount.  And among many lawyers, there is a reflexive view that bigger is better, particularly in uncertain economic times.  See, e.g., Laurence Simons, “Number of US law firm mergers rockets,” Oct. 14, 2015. Yet, many other factors are bound to matter as well, such as (a) geographic presence in particular markets, (b) geographic concentration, (c) specific practice areas, (d) practice area concentration, (e) client concentration in specific industries, and (f) measures of reputation in lucrative financial services markets.

The graphic below shows the results of a model that includes all the factors just listed.

When reliable measures of these additional factors are included in a multivariate model, size (in this case, headcount) becomes a negative predictor of profitability at a level that is statistically significant.  In other words, when all these other levers of strategy are factored into the analysis, more lawyers means lower average partner compensation.  Further, the model explains approximately 80% of the variation in average partner compensation, which makes it highly unlikely that many firms can thrive using strategies significantly at odds with the model’s results. (For the prescriptive advice that flows from this model, please read the full article.  H/T Chris Zorn and Erik Bumgardner, both PhD social scientists, who helped build the model. )

I show the profitability model to make a simple but important point:  The use of statistics to guide strategy and operational decisions only becomes valuable when deep contextual knowledge of an industry and its problems is combined with quantitative competence and the ability to effectively communicate results. By dint of his deft mind and an amazing work ethic, Everett Rogers developed this ability in the realm of diffusion theory.  Outside my former colleagues at LawyerMetrix and a few others, this combination remains all-too-rare in the legal field.  Over the next generation, this gap is destined to close.  In the meantime, beware of charlatans selling into a naive and fearful market.  Also, don’t expect an internist to perform surgery.

Strategy that works = models + reasoning ability

For lawyers, there is a silver lining to all of this complexity: High quality empirical models — like the Rogers’ rate of adoption model (Post 008), the organizational innovativeness model, or the profitability model above — are nothing more than signposts that communicate what is more, or less, likely to matter. They don’t, by themselves, produce fully baked strategy.  Rather, strategy based on models requires the application of additional reasoning ability, of which lawyers have no shortage. Fortunately, sophisticated quantitative analysis is itself an innovation that is starting to diffuse through the legal industry social system.  See Post 004 (“Rogers’ core insight – one that is absolutely foundational for Legal Evolution readers – is that the diffusion of innovation is a process that occurs through a social system” (emphasis in original)).

In Part III (017), we can finally dive into the Rogers organizational innovativeness model with confidence that we can draw the right inferences and, in turn, use the model to set and execute sound organizational strategy.

What’s next? See Innovation in Organizations, Part III (017)