Modelling Matters: In the Thinking Room

Tim Gent

The earlier years of my OR career encompassed an extensive range of events; COVID-19, Brexit and the invasion of Ukraine to name a few. I have also seen how OR is conducted within the public sector and across industry. I was never formally educated in OR except for a few modules at university that alluded to it and yet somehow, it is now my every day. How this has been possible is through several realisations and reflections that others may find useful when starting out.

Modellers need to balance the need for speed against factors such as complexity, flexibility and explainability. To do this, I find it helpful to appreciate the different levels of decision-making where models may be used, and their different requirements.

In this article, I’ll outline differences I have noted between Strategic, Tactical and Operational levels, how modelling is used by each of these, and how modelling can be aligned to those needs.

Operational – The Control Room

Let’s start at the delivery end, as it’s obviously the most important place to be! The operational level is about actually doing things, and by definition decisions must happen very quickly.

Operational decisions often happen in control rooms, e.g. for railways, power systems and factories. This could involve co-ordinating emergency services, despatching deliveries or military operations. All of these – and many more - will use models to aid rapid decision-making and optimise use of resources.

These are situations where model speed, reliability and clarity are undeniably critical. There may not be the luxury to go and make a cup of tea while a model runs, still less to spend time specifying the problem and digging around to understand the outputs. This is decision support at its purest and most time critical.

For modelling, there are two interesting points worth noting:

1. Models are used here but not developed or built here. There isn’t time!
2. All raw data (quantitative and qualitative) used in modelling originates here.

Strategic – The Boardroom

At the opposite end is the strategic level of management. The really, really big decisions about direction of travel happen here. This involves goal setting (what should we be trying to do?) and policy (how should we be going about this thing?). Often also counted in the strategic layer are the big yes/no investment decisions (should we build this thing?).

The priorities and challenges for modelling are different here. The problems are often messier and less well-defined, with more scenarios to be considered, and usually unique aspects to each decision. Arguably they are the ‘high stakes’ decisions, though clearly that’s a value judgement when one considers operational decisions made by doctors and military leaders.

Strategic decisions are often made over a longer time limit, which ideally means there is both the desire and the space to consider all aspects of the problem. However, the ministers, senior executives and other big-wigs making the decisions are busy people, and they are very rarely modellers. Their biggest ask from a model tends to be for clear answers, delivered in the timeframe that they need the information (or at least when promised).

There are two ways in which strategic and operational decisions are very similar: they both desire ready-made models, and models are not built at either level! However, whilst the operational layer outputs data, the strategic outputs goals, policies and decisions.

Tactical – The Thinking Room?

There must then be a place where modelling actually happens, and indeed this is the place I feel most of my career has been spent. It is often invisible or seen as simply part of strategy and research, but I believe it’s important to separate out clearly. I’ve come to think of it as the ‘Tactical Level’.

Tactics are often defined as the plans which implement a strategy. The goals set and big decisions made in the strategic level are turned into detailed actions in the tactical level.

An example of this from my own work has been modelling to build railway timetables. This happens after the desired service levels, capacity upgrades and goals of reduced delays and crowding are set. Building railway timetables to deliver the strategy is complex; models are vital to balance optimising use of rolling stock, station capacity and staff rosters, factoring in robustness to delays, and of course maximising revenue.

At the tactical level, decisions are less immediate, and there is more time to think and refine. The overall goals have been set, and the output is an achievable plan for the operational level to deliver. But it’s not just plans: the tactical level is also where we can spend time building tools that help make operational decisions quickly and clearly.

Tactical Modelling can go up as well as down

So, modelling at the tactical level turns strategy into operational plans and decision support. However, it also works the other way. The same skills - and often the same models - in the tactical layer are the real source of strategic modelling. It should be clear why this is the case: building and testing models takes time and focus which isn’t available in either the boardroom or the control room.

In an ideal modelling world, it might be possible to view the tactical level as where all the ‘slow and complex’ models sit. Raw data can be taken from the operational level, processed and analysed to understand what’s really going on, including relationships and trends. Complex models can be built and tested over multiple scenarios, anticipating possible strategic and operational questions. These can be further used to create optimal plans, guidelines and rules of thumb, decision support and control systems or truly strategic models for executives.

What not to say

There are a couple of common traps which I’ve strived to avoid whilst writing this article, and are worth mentioning here:

1. Strategy is long-term, operational is short-term, so tactical must be medium-term! Sometimes thinking about timeframes like this helps, and ideally strategic thinking happens months or years ahead, but not always. However, short- and medium-term are nebulous and overlapping concepts. Most of all, I think defining the tactical level as ‘medium-term’ vastly undersells it!

2. A big model is a strategic model, and a strategic model should be a big model. Everyone loves their model to be big, complex and strategic. However, big models in themselves are less helpful for strategic thinking. Being big and complex is never a goal in itself, or a condition to be useful. Quite the opposite.

Tactical Research Society, anyone?

This all brings me to the conclusion that no modelling actually happens at all at the strategic or operational levels. The boardroom and the control room are users of models and model outputs, not modellers. The middle layer which I’ve called ‘tactical’ is the real powerhouse of analysis, model design and model building. It is the ‘thinking room’ where models and modellers can take their time to address strategic, tactical and operational needs.

Of course, the names are in the end just labels: it’s the understanding of what we are really doing that matters.

I would love to hear from others whether they feel they are modelling at the strategic, tactical or operational level. With tongue firmly in cheek, could it be time for a rebrand to ‘Tactical Research Society’?

Tim Gent is an independent consultant, offering advice and support on the development, use and maintenance of all kinds of models.

Email: [email protected]