When the target becomes the plan

The opinions in this article are my own.

I've been interested in systems, and in particular the system that is the construction industry for quite some time. It's interesting because it is so resistant to change. I wrote a while ago about how construction needs to optimise for the whole system, not its parts. What I basically said was that incentives, and the behaviours they drive, need to be considered carefully. I was working in offsite at the time, and I have always had an interest in alliance contracting, so those were some of the suggestions I put forward that might help.

This article is more about observations than solutions, and it's about something every programme has: targets. It isn't aimed at any particular programme or person.

A target is a sensible idea. You can't manage what you can't see, and a number gives everyone something to aim at. The trouble starts when the number stops being a view of the work and becomes the work. The economist Charles Goodhart's observation, as it is usually put, is that when a measure becomes a target, it ceases to be a good measure.

Why it happens

People respond to what they are measured and rewarded on. That isn't cynicism. It is rational behaviour inside the system and constraints they've been given. Steven Kerr described the pattern in 1975 under a title that still stands up: On the folly of rewarding A, while hoping for B.

In a capital programme there are many A's and B's. Here are five patterns that are common enough to be recognisable.

• The spend profile. A target to spend to an agreed forecast by a date, used as a proxy for progress, can pull work forward, push scope into the wrong period, or reward starting things over finishing them. Money spent on time and value delivered are different things.
• Green reporting. If a red status brings scrutiny and a green one doesn't, red items get described as amber and amber ones as green. The report looks healthy until the point where it can't be hidden any more.
• Padded programmes. If finishing early earns no reward and finishing late is punished, the sensible move is to build contingency into every activity. Work then expands to fill the time it was given.
• Silent early warnings. A contract can require early warning, but if raising a problem feels like admitting failure against a target, people wait until they have a solution. By then the options are narrower and more expensive.
• Benchmark drift. Benchmark a rate against the median and everyone moves towards it, including the people whose work was legitimately different. The measure stops finding anything out.

It's important to say that these examples are not binary. People don't flat out lie in their reports about RAG ratings. It's much more nuanced. It's the adding up of small behavioural trends: rounding programmes up from three weeks to a month, or taking the worst-case scenario on costs, which effectively adds risk on top of risk. These small things have a small impact on a section of a project and a medium impact on the project, but they can cause large swings on a big capital programme.

The system, not the person

None of this needs bad people. Each team makes a reasonable choice for its own part of the programme, and the sum of those choices can be worse than any of them intended. That's the whole-system point from the other article I wrote. Optimising the part can damage the whole, and a target is a very effective way of optimising a part.

What helps

The answer is not to abandon measurement. A few habits seem to help:

1. Start with the behaviour or outcome you want, then set the target. Serious thought about targets is needed, because time spent up front saves so much later trying to fight against issues of your own making.
2. Pair measures that pull against each other. Spend alongside completed, accepted work. Programme progress alongside the quality of the information behind it.
3. Reward early bad news. If the first person to raise a problem is thanked and the last is questioned, the early warnings turn up. Contracts like NEC are designed around this, but the behaviour has to match the clauses.
4. Keep the number of targets small. Every extra target is one more thing to optimise.
5. Review the measures themselves. If a target has been met for a year, ask what it is now encouraging.

A question

Where have you seen a target produce the opposite of what it was set up to do? I'm collecting examples, and I'd like to hear them, with the names left out.