We would all like to know what will happen in the future: life would be easier if we could get a warning about problems before they happen. This is also true for managers - especially when they are working on strategy implementation projects. Finding a way to give managers this early warning has been a key theme driving performance management system design in recent years.
Sadly more often what happens is that bad things happen, and only afterwards do we realise that if we had paid attention to the information we had, we could have avoided the issue altogether. One way to think about the value of a strategic performance management system is that it is designed to make this kind of ‘hindsight’ a thing of the past…
The good news is that it is now easier than ever to construct these ‘strategic early warning’ systems. They work by using one measure (of past performance) to help managers predict future performance of something else. A commonly used term for the measure used to do the predicting is a leading measure or KPI.
Although the terms leading measure or leading KPI are familiar to many, it seems that they are not well understood. Questions about how the terms are defined and how they are chosen come up frequently during our training courses - an experience apparently shared by many others if the preambles to definitions of the terms on the internet are to be believed. Widespread use of the phrases leading measures and leading KPIs in the context of performance management is a relatively new thing. The key concept they are based on is much older - dating from research carried out in the USA in the 1940s/50s on the use of time series data to forecast changes in the US economy - but modern usage stems mostly from their inclusion in articles and books associated with the Balanced Scorecard. The terms first appear in some of the materials about Balanced Scorecard published during the 1990s, but it really wasn’t until about 2000 that the term broke out of the specialist performance management space into common usage.
A modern concept
Publications featuring these terms that have been published during the last five years are particularly informative - they suggest that whilst there is good awareness of the need for predictive measures, people are not so clear about how to identify them. This is understandable: although the concept of predictive measures is easy enough to grasp, the practicalities of defining and using them are more complex.
But hopefully help is at hand - in this feature we outline how 2GC approaches the design of leading measures and KPIs, an approach that is practical and easy to implement.
To understand why, we first need to go back and understand more about where the leading measures idea came from.
Causality is the key
The earliest Balanced Scorecard designs simply comprised sets of financial and non-financial measures chosen to be informative about the progress made towards implementing a strategy. Measures can only ever track things that happened in the past, but the designers of these early Balanced Scorecards recognised the need for a strategic management tool to have some kind of predictive ability. Their solution was to build a Balanced Scorecard with a mix of leading and lagging measures: Leading measures were chosen to be predictive of changes in other measures on the Balanced Scorecard, the prediction coming from the application of a simple causal model to the base measure. For example, in the cable TV industry it was initially thought that increasing the number of houses that could physically be connected to a cable TV network would quickly lead to an increase in subscribers on that cable network. So, it was proposed that a good leading measure for subscriber growth in future months would be the measure of ‘additional homes passed’ by the cable network in a month.
This use of these simple causal models presented two problems, one for Balanced Scorecard designers and one for the Balanced Scorecard users: the designers needed to find the right model; the users needed to know what it was.
Because the causal models were usually unique to a specific organisation it was hard for designers to use general knowledge or consult ‘compendiums’ of measures to work out what they should be - the cable TV leading measure of ‘additional homes passed’ worked to forecast future subscriber levels for a cable TV business during its infrastructure development phase, but would be of limited use to one whose network was complete, and no use at all to a competing company offering satellite based TV services. The choice also depended on the designers understanding the organisations future strategic goals. Because of this, leading measures on early Balanced Scorecards often had obscure selection processes that were hard to reflect in the definition of the leading measure - and so users of the measures were expected to understand this ‘measure backstory’ by some other means.
When users were unclear about what causal model was being applied (or just guessed wrongly) misunderstandings happened. In the cable TV example, the cable laying teams responded to the KPI given to them (to maximise the number of houses passed by new cable laying each month) by focusing on laying cable in areas of highest population density - typically the poorest parts of cities. Laying cable in these areas met the KPI target, but did not result in as many new subscribers (residents often could not afford the subscriptions) and also increased the cost and complexity of the cable laying operations (due to a mixture of vandalism and theft of cable laying equipment left on-site overnight).
If the cable TV teams had understood better why they were given the KPI (i.e. to maximise new subscribers, rather than simply pass as many houses as possible) they might have approached their task in a more useful way. So we see that the challenge with leading measures stems not from the measures themselves, but from peoples’ ability to understand the the strategic reasons that informs how they should be chosen. Those selecting measures have to gain a deep understanding of an organisation’s operations and strategy; those using the measures have to understand the thought processes the designers used in order to decide how best to act to achieve it. In both cases the challenge is to understand the ‘causality’ that allows the base measure to predict the outcome desired. In traditional cases (such as the cable TV example) this causality was not defined - it was implicit in the measure selection. Unsurprisingly the best-practice fix is simply to make this causality explicit.
Communication of the causal model enables success
So we see that the challenge with leading measures stems not from the measures themselves, but from peoples’ ability to understand the the strategic reasons that informs how they should be chosen. Those selecting measures have to gain a deep understanding of an organisation’s operations and strategy; those using the measures have to understand the thought processes the designers used in order to decide how best to act to achieve it. In both cases the challenge is to understand the ‘causality’ that allows the base measure to predict the outcome desired. In traditional cases (such as the cable TV example) this causality was not defined - it was implicit in the measure selection. Unsurprisingly the best-practice fix is simply to make this causality explicit.
As the cable TV example shows, a leading measure tracks our activity on something that has a direct impact on a key outcome for the organisation (in that example the activity was cable laying, the outcome was increased subscriber numbers).
In practice any measure of an activity an organisation carries out can be used as a leading measure if we can match it via a causal path to a key organisational outcome. We also don’t want to have lots of measures, so our selection process needs to be able to identify which activities are the most important ones to focus on. Fortunately there is a proven and robust process that does exactly these two things - it is part of the modern Balanced Scorecard design process.
Modern Balanced Scorecard design methods are built around documenting the causal links between a critical set of strategic implementation activities and the key organisational outcomes that these are expected to influence in the future. A modern Balanced Scorecard does a lot more than this, but at its core is this reliable process for identifying and communicating priority activities - and it can be used simply to develop a set of leading and lagging measures or KPIs - even if you don’t then go on to use the Balanced Scorecard report itself.
The other useful by-product from developing your leading and lagging KPIs using a Balanced Scorecard design method is that it naturally leads you into considering all fours steps of the 2GC ACME strategy implementation process - in particular encouraging you to review the measures periodically (the Engage step) - which brings us back to where we started. Getting an early warning of problems before they occur is only useful if you also take steps to do something with this information… and the awful feeling of ‘hindsight’ you are left with knowing that you could have done something if only you had been paying attention.