← writing
2 December 2021

A correct figure is not enough

Joining a consultancy, you learn quickly that the difficulty is less computing correctly than getting a figure accepted by people who did not follow the computation.

I joined a data consultancy in October. After a year alone with books, the change is clear: the difficulty now lies in the trust placed in results more than in the technique. The coming year will consist of client engagements, and therefore of meetings where people have to be convinced. Hence three books focused on evidence rather than on tools.

01Jonathan Gray et al., The Data Journalism Handbook
Cover of The Data Journalism Handbook

Written for journalists, useful to engineers: it treats the source, its freshness and its gaps as the main subject rather than as a preliminary. It is the discipline corporate dashboards most often lack.

02Daniel Kahneman, Olivier Sibony and Cass Sunstein, Noise
Cover of Noise

Three authors on a flaw in judgement that is rarely discussed. Two equally competent people looking at the same file do not reach the same decision, and the gap stays invisible because only one decision is ever observed. Bias is discussed in meetings, whereas noise never appears there. In committee, a figure is therefore often rejected because each person reads it differently, even when it is correct. The remedy is to set a decision rule before presenting the figure.

03David Graeber, Bullshit Jobs
Cover of Bullshit Jobs

Uncomfortable and useful when you bill by the day. Its thesis about jobs their own holders consider pointless is debatable in detail, but it raises a question to ask of every engagement: what would have been missing if it had not taken place?