Last week, I began my article by extolling the virtues of knafa. This week it is the turn of baklawa, another popular Middle Eastern dessert. It’s made in many countries in that region and it varies somewhat depending on the countries. In Turkey, a much larger amount of syrup is used (and possibly honey, Google seemed a bit inconclusive on this!). In Lebanon, Syria and Jordan by contrast, it tends to be made with less syrup, and it tends to be made into smaller pieces. The general approach is consistent though, involving many layers of filo pastry, with lashings of butter and a variety of different types of nuts (often pistachio, but also other types such as cashew). The whole point of course is the combination of the flavours, but also the layers of pastry. If it were just a single layer, it would be a totally different dessert and have a totally different texture, when eating it.
We can use this type of layered baklawa approach to understand how the data science function can be utilised (bear with me for a moment.. there is a vague analogy here!). There are many ways to deploy data science teams in financial organisations. One way is the single layer approach. In other words parachute in a team of data scientists into a firm, and somehow they will get involved. For a startup, this type of build first approach can work. However, in established companies this may not work so well.
Instead in established firms, the “baklawa” approach could work better. Rather than suddenly deploying a massive data science team, think about it layer by layer. Bring in data scientists slowly into the organisation, and hopefully over time they will show the value such approaches can be valuable. Of course, it would quite be baklawa, without the layer of pistachio. This is the rest of the organisation, that needs to be brought on board to make the recipe to work. If data scientists are totally independent entities within a firm, then it might be tough for them to have an impact.
By contrast, if they can work together with other members of the organisation, such as portfolio managers their output can be an integral part of the decision making process. It’s only when the right questions are being asked of data scientists, that it can truly work. If their insights are not relevant to a financial firm, then no-one is getting the true value of this data driven approach. The layers or “data science stack”, which can eventually come about with time, consists not only of the team members working together in different functions, like trading, data science and IT, but it can also include the actual tech stack being used (eg. Python, Spark, databases etc.) It all takes time to build, and needs buy in from management and the organisation as a whole. If financial firms fail to develop an effective data science function, it is likely they could fall behind their competition.
I suppose that only leaves one thing left to do: search for some good baklawa in London!