A recent article suggests that, in the “near future”, data analysts will be compensated based on performance. They will receive commission-based payments, rather like salesmen, rather that being paid purely for their time. This performance will presumably be determined by the impact that the data analyst has on the key goals of the organization, e.g. profit, turnover, service level.
I’m skeptical. First of all, I’m not sure it would even be possible to assess the impact that a data analyst has on business outcomes. And, even if it were, it would be a dangerous path for an organization to tread.
The first problem is knowing what you want to measure. I remember a story about a fast-food franchise in the US. One of their performance metrics was the percentage of food that was wasted. To provide a quick service, fast-food restaurants cook items in anticipation of them being bought. If they are not bought within a set number of minutes, they are discarded as they no longer taste fresh.
One restaurant in the middle of nowhere was managing to achieve wastage rates of 0%. Business analysts from head office were dispatched to see how the restaurant was achieving this minor miracle. Of course, you’ve already guessed—they were cooking to order. Be careful what you wish for. The Law of Unintended consequences at work. If you encourage data analysts to optimize for certain metrics, they will do just that.
Assuming you could determine the metrics you wanted to measure performance against, you would be faced with the next problem—how do you separate the contribution of the analyst from the contribution of the rest of the business?
Data analysts don’t work in a vacuum. They process data to assist other parts of the business. In a recent podcast, machine learning researcher Evelina Gabasova emphasized how her role was to work closely with others to try and turn statistical insights into actionable results. She works in the bioinfomatics and statistical genomics field and relies on medical colleagues to determine whether the results of her models have any real-world value. The machine learning research and the medics work together to produce actionable results.
Domain expertise is essential when attempting to extract insights from data. The communication gap between analyst and domain expert is the biggest barrier to the effective use of data. Considerable effort in the data analysis industry is expended on producing tooling that can be used by those without access to specialist data science skills. If domain experts can process their own data, the communication gap disappears. I’ve written previously that there appears to be a trend in this direction.
Anything that encourages data scientists to become “lone wolves”—to avoid collaboration—is likely to have counterproductive results. If analysts are being compensated based on results they won’t want the credit for improvements going to their collaborators.
We must remember that data analysts don’t actually act on their analyses. They are not line managers. The only way an analyst’s results can impact the success of the organization is if someone else effects changes based on those results. Instigating and managing change is a difficult task. It typically requires a skilled leader with decent project management skills. Given that, if data-driven insights lead to business improvements, who should be praised…or compensated? The analyst? The business manager? The workforce? Try resolving that conundrum.
Of course, maybe data analysts will start to be compensated based on performance in 2015. But, I sincerely hope not.
If you are interested in learning more about data science, Learning Tree’s “Introduction to Data Science for Big Data Analytics” course cover the concepts, techniques and tools required by modern practitioners.