‘Would you tell me, please, which way I ought to go from here?’ ‘That depends a good deal on where you want to get to,’ said the Cat. ‘I don’t much care where—’ said Alice.
‘Then it doesn’t matter which way you go.’
Lewis Carroll, Alice’s Adventures in Wonderland
Targets and trajectories
A target is a measurable goal for the system. For instance:
- reduce teacher absenteeism to 5% within 18 months
- achieve 100% enrolment of primary-age children by 2015
- increase immunization coverage to 90% within three years
- halve child mortality rates within five years
- recruit 40,000 new health workers in two years
- introduce a new examination system within 12 months
For a target to be useful it needs to be:1
Specific: the target needs to define a precise goal, not a general aspiration (for instance, ‘reduce the maternal mortality rate by 50%’ rather than ‘improve maternal health’)
Measurable: there must be a way to determine whether or not the target has been achieved (even if that measurement is difficult or imperfect)
Attainable: it must be possible (but not necessarily easy or certain) for the target to be achieved
Relevant: the target, if achieved, should have a substantial and important positive impact on the system and outcomes
Time-bound: there should be a specified time frame within which the target will be achieved
A trajectory shows the path to the target. So for instance, if the target is to be achieved in 12 months, the trajectory will show the expected progress at monthly or quarterly intervals to that date (or another appropriate time period). For instance:
|Today||3 months||6 months||9 months||12 months|
Why set targets and trajectories?
Targets and trajectories define what the system aims to achieve and when it aims to achieve it. They therefore help to:
- Guide planning
- Coordinate the different groups required to achieve the goal
- Create a picture for people of what will be achieved
- Inspire people to act
- Gain political support
- Guide calculations around investment
Targets can be harmful if they create a feeling of excessive pressure on the system or if there is a widespread sense that they are impossible to achieve. In this case, the controversy they create may outweigh their benefits. If the wrong metrics are used, or targets are set badly, they can become counter productive or simply lead to gaming of the system. Many of the most effective leaders use targets, but are careful about how they set them and how they communicate them to the system to ensure that they inspire rather than demoralize.
A trajectory enables the system to measure whether or not it is on track to reach the target.
Basic principles for setting targets in a delivery context
Many development approaches set targets based on historical tends or what the system ‘knows’ it can achieve. This produces incremental progress.
A Matrix of Delivery (adapted from Sir Michael Barber)
Delivery typically seeks a step change in performance. By extension, this means that most targets will be ambitious, expecting significantly more than historical rates of improvement. It also means that the target will generally go beyond what the system already knows how to do. At the same time, the target needs to be realistic and be grounded in evidence about what can be achieved.
The first step is to define exactly what the target metric is. For instance, the goal may be to improve immunization coverage, however, more specificity is required. For instance, we need to know:
- Exactly which immunizations are we targeting?
- At what age?
- For what target population?
- Will any groups be excluded from the target population?
Similarly, if the goal is to improve enrolment, we need to know:
- Which age group are we targeting?
- What counts as enrolment? (If a 14-year-old is in grade 1, does this count? Or a seven-year-old in pre-school? Or an enrolled child who has not attended for the last week? Does enrolment in non-formal education count? What about non- formal education that only lasts for two hours a day?)
Wherever possible targets should use existing metrics and data to ensure alignment with other programs and reduce confusion. Simple metrics, which are easy to understand, should be preferred over complex formulae or indexes.
Some targets do not lend themselves to an obvious quantitative measure. For instance, a target might be to improve school leadership, reduce corruption, or improve satisfaction with healthcare. Here, a specific metric or set of measures needs to be developed.
In some cases, these questions may touch on fundamental questions of policy. In other cases they may be more straightforward. Where possible, it makes sense to use metrics that the system is already using.
Example – measuring enrolment
In the education sector, multiple measures are used to capture enrolment:
- Participation measures the proportion of children of a given age who go to school, regardless of the type of school they attend
- Net Enrolment measures the proportion of children of a given age who go to the right school for their age group (so for instance, a 14 year old who is in primary school is excluded from net enrolment)
- Gross enrolment measures the total number of children in school (of any age) divided by the total number of children of school-age
In a low-performing system where many children enroll late or never go to school, NER will generally be the lowest (50%), participation higher (80%) and GER the highest (>100%) due to grade repetition.
Targets set using different metrics will have different effect:
Setting the baseline
The first step in setting targets is to establish a baseline – a starting point and a view of what will happen naturally.
Global average temperatures 1970-2010
This requires an understanding of the indicator, limitations of the data, natural fluctuations, and trends and their root causes.
For instance the graph shows global average temperatures for the period up to 2010. Several questions arise:
- What should be the starting point (assuming we start our work in 2011, the first year after the data series is complete)? Given the fluctuation in the data, we could take 2010 as a starting point, or an average of the last few years of data, or, better, an average weighted to favor later readings? We could plot an overall trend using data from 1970 to 2010 and then use the estimate for 2010 based on this trend, rather than the actual data, as the starting point. This would be vulnerable to later changes in the trend?
- Can we project ‘what will happen anyway’? In this case the long-term trend is clearly upwards. Reducing the rate of increase would be success, even though the trend would still be upwards. In this case, the basis for our target setting should be an assumption of the continuation of historical trends, not a straight line. In other cases, where underlying factors driving historical increases have ceased to be relevant, an assumption of no future change may be relevant.
In addition to trends, any seasonal variation needs to be factored into the baseline. For instance, the graph shows US soft drink sales between 2010 and 2012. The overall trend is declining, but any model, which fails to consider the large seasonal variation, is unlikely to be helpful.
US Soft drink sales 2010-2012
There are three potential ways to set a target
Based on aspirations: the target may be based on an aspiration set by leadership. For instance, political leaders may have committed to 100% enrolment, or provision of free basic healthcare. Aspirations may in turn be based on a view of what ‘has to be done’
Based on plans: the target may be based on the estimated impact of a set of interventions described in a plan
Based on benchmarking: benchmarking can be used to develop evidence about what it might be possible for the system to achieve
Benchmarking is the most common way to set targets and is crucial to anchor the target in evidence about what can be achieved. Even where the target is based on aspirations, benchmarking can help to check whether the target is achievable and give people confidence that it is.
The system can benchmark against four different things:
|Itself||The system can compare its average performance to the performance of the top-performing region, or to the best schools, or between genders or other divisions||What if every school was a good as the top ten percent of schools? What if boys performed as well as girls? What is the south reached the same level as the north?|
|Other systems||The system can compare itself to other similar systems. For instance, absenteeism rates of teachers might be compared to absenteeism rates of health workers. Or the public sector compared to the private sector||What if government schools were as good as private schools? What if teacher attendance matched health worker attendance?|
|Time||The system can compare itself to the past, looking at peak performance in the past or historical rates of improvement||What if we doubled the rate of improvement? What if we performed as well in summer as we do in winter?|
|The world||The system can compare itself to other countries.||What if we performed at the regional average? What if we improved at the rate of the fastest improving country?|
One consideration is how well benchmarks will resonate. People often have strong associations with different countries and regions that will determine how they react to benchmarks, even if the logic for the comparison is sound. For instance, comparing India or Pakistan to Sweden is unlikely to resonate well, because the comparison feels unfair. Conversely, Malaysia or Turkey may seem reasonable ‘targets’. Demonstrating that performance in health or education is worse than in other much poorer countries will elicit a strong reaction but may also be demoralizing. Internal benchmarks are often powerful, particularly if the top performers are unexpected (for instance, if a poorer region does particularly well on one metric, that is likely to resonate powerfully as a target).
For instance, in a school system reform, targets might be set as follows:
A trajectory shows the path to the target.
- Situations where there is a long lead-time of lag effect (for instance where extensive preparation is required before change can be implemented) will produce a low trajectory
- Situations where there are actions which can produce rapid progress but where it will then get progressively harder, will produce a low trajectory
- Situations with a combination of these will produce a linear trajectory (or, more often, an S-curve)