A while ago, we worked with a leading law firm to help them become more customer-centric. A key component of this work was understanding the networks of partners within the firm, and how natural working relationships mapped onto the formal organization structure. To do this, we used existing data and social network analysis.

The problem

Our client is a successful firm whose work spans both multiple client industries and multiple legal disciplines (e.g., litigation, commercial, intellectual property). The firm was subdivided into smaller groups to ease management and collaboration, but there was a feeling that the current structure might be capable of improvement, especially as regards putting teams together that can more readily do two things:

Thus, we needed to find a way to objectively evaluate the current organization with respect to its ability to serve the needs of customers, not just the internal capabilities of different parts of the law firm.

The approach

To attack this challenge quickly and efficiently, we used existing data, rather than embarking on a lengthy process of interviewing or data gathering. We used a 5-year dump of the firm's time records, which contained records of all projects, who worked on those projects, and how much time each person billed to each project.

We needed to use this data to find connections between people, in a way that was possible to analyse. To do this, after trying various cuts, we decided to use hours worked in common sectors as the connecting metric. For example, if you and I both worked a total of 200 hours last year on three telecoms projects, and also 100 hours on five media projects, the strength of our relationship could be quantified as 300. We used our automated tools to trawl through the time record database and generate these connections between all combinations of several hundred partners.

To make sense of all this data, we used our web-enabled social network analysis software to generate graphical maps of selected departments, which we analysed visually. Some examples follow (click on an image to see a larger version):

1. Raw connections

This initial view, arranged using a "spring layout" algorithm, contained all connections above a certain threshold, with lines weighted in proportional to the number of hours billed in common sectors.

This showed that most partners were connected within well-defined clusters, with some partners acting as "bridges" between clusters, whilst a few partners tended to be only loosely connected.

2. Sector focus

We next colour-coded the partners by sector (each colour corresponds to a different industry), and this revealed that the clusters did indeed correspond to strong shared activity within common sectors (not surprising, since the connections were defined in terms of time spent in different sectors).

3. Organizational units

Where it got surprising was when we changed the colours to represent sub-departments within the larger group (each colour now represents a sub-department).

What we found here was that there was no rhyme or reason for the current organization, since it did not correspond to any shared activity, and thus did not contribute to the formation of teams that could serve a customer any better. In fact, the random distribution of the colours suggests that the allocation of people to departments was quite arbitrary.

4. Billing heavy-hitters

When we turned-on the feature of our software that changes the size of the circles to represent an attribute, in this case total billings over the last year.

What we found here was that some partners were diverse, falling somewhere between two extremes. At one end were exceedingly well connected and productive partners, with a high level of activity spanning multiple clusters; at the other, were a small number that generated few billable hours and were highly isolated.

Implications

This analysis showed that current organization did not match the informal networks and the working patterns that actually took place in the firm. The structure in fact may have inhibited some of the collaboration necessary to effectively anticipate and serve client needs.

The clusters revealed by the social network maps suggested a new organization, based more closely on how people actually worked, and how they actually collaborated. It also identified a number of special cases of partners who were particularly important in bridging clusters, and others who were more isolated, data that proved helpful in improving the overall effectiveness of the group.

Other applications

As our page on social network analysis suggests, this approach is useful for finding patterns in interactions within your organization, and with clients, customers, or partners in other organizations. Looking at these interactions can help you change organization or working patterns, and improve productivity and collaboration.

It should be emphasised that this can often be done quite quickly using existing data. In this case we used time records, but these maps can also be generated from e-mail traffic, data from diary management systems, project work order systems, and the like.

Add a comment