Metrics,
KPI’s and incentives can be dangerous things. Just ask the CEO and senior
leaders of Wells Fargo, who were proud of an incentive scheme rewarding staff
for opening new accounts until it landed them in court and in front of a baying
senate. Staff had responded to the scheme by opening false accounts en masse,
in many cases robbing real customers of their money.
I saw many
such pitfalls during my time in business, though none as openly fraudulent as
this one. Management rightly tried to use simple, targetable metrics to steer
their businesses, a practice that only became more widespread once scorecards
and dashboards came into vogue. Sometimes the real target was the leader’s own
bonus package, but more often than not I witnessed a genuine attempt to improve
a complex business.
But the
attempt backfired almost as often as it succeeded. A common example is the
trade off between volume and margin. Sales people love volume targets and
respond well to incentives on volume, but the risk is that they chase down
prices in their zeal or target customers of low value. In the short term,
margin incentives can make more sense, but not if they result in collusion with
competitors or in pulling back from profitable, if marginal, clients. Further,
margin is usually more complicated to measure and influence, and sales people
value simplicity.
The most
dangerous indicators are usually ratios, KPI’s measures as X per Y. A margin
target is usually a ratio, something like dollars per litre. An acid test of a
good ratio is that it is always beneficial to both increase the numerator and
to decrease the denominator. In the case of a margin KPI, it will almost always
be good to grow the dollars, but often not good to reduce the litres. Further, reducing
the litres is the easiest way to shift the indicator in the short term. So such
a margin target runs a severe risk of a department follow a “golden litre”,
that is shrinking the business down just to the most profitable clients, a sure
recipe for a business death spiral.
Shell and
others tried to get around the volume/margin dilemma by targeting a concept
called contribution. This was sound, and generally successful, but it lacked
simplicity and often left sales people confused.
In call
centers, a common incentive KPI is time per call. True, dealing with complaints
and enquiries quickly is a good general idea, but not at the expense of being
thorough. In some cases, employees have been known to put the phone down on
customers where the call showed signs of dragging on.
In Shell
retail, another example of a ratio metric was called efficiency index, defined
as sales volume per station compared with the market average. This had the same
downside, namely that while growing volume is certainly good, reducing station
numbers can be easier and unhelpful to the business. What if those stations
still made a good contribution, and had low capital and costs to serve, low
risks and low effort intensity? That was usually the case with dealer owned
stations. I preferred a metric of company owned efficiency index, once again
with the downside of loss of simplicity. I guess the moral is that business can
be complex.
And if
business is complex, so is macroeconomics. Central banks and governments face
the same pitfalls in trying to find and influence levers to improve
performance. And most politicians make the CEO of Wells Fargo seem honest.
Maggie Thatcher relentlessly tinkered with definitions to try to obscure the
scale of unemployment her policies caused. The Kirchner’s in Argentina gutted
their statistics. We can trust very little data coming out of China, and the
investors know it, but fear shouting out for the same reasons they pretended
all was well with sub-prime mortgages eight years ago – their own golden eggs
depend on the deception.
Even honest
Economists have difficulties, one notable one being that times change. Forty
years ago policy was all about avoiding devaluation, then it became a helpful
tool, only we didn’t see how helpful until the Euro zone discarded it. Much of
the current weaponry was designed to control inflation, but now we realise that
we need more of it not less.
There is a
faction trying to redefine GDP, or at least to replace the current definition
of GDP in many key indicators. Partly this is because GDP itself has become
slightly out-dated, because it doesn’t really reflect things that have low
manufacturing component or are almost free like internet access or Airbnb
bookings.
One thing
we can still do is to choose to use GDP total or GDP per capita. Businesses
like GDP total, because that drives activity and demand for them. But for
wellbeing of citizens, GDP per capita matters more. Japanese GDP is declining,
but mainly because the population is shrinking. That carries challenges, but
one of those is not immediate poverty, since GDP per capita is still rising.
An
indicator that seems to have survived the test of time is productivity, defined
as some measure output per a chosen unit of input, usually labour hour.
I first studied Economics in the 1970’s, and
at the time UK labour productivity was suffering compared with that of Germany
and others. Pundits had easy explanations. British workers were lazy, and
usually on strike. British managers lacked drive. Germany benefited from having
lost the war, because it meant its infrastructure could be replaced.
Even back
then, it was possible to see the problem with the indicator. Higher
productivity was generally good, but which of these explanations was most
likely to be right, or was it a combination or something else? And what exactly
could policymakers do to improve things? In practice, the politicians chose
their own causes and remedies – so for example the right demonised trade
unions.
Since then
bigger problems with the indicator have emerged. It seems difficult to predict
and sometimes moves in surprising ways. But worse than that, it suffers from
the classic problem with ratios – increasing the numerator is good, but
reducing the denominator is not unless there is full employment, which there
rarely is nowadays. Indeed, much government policy is designed to create jobs,
irrespective of any output generated.
So we have
an indicator that has lost its lustre. The numerator seems poorly measured, due
to the trend away from manufacturing. The denominator is not something
governments want to reduce. And the metric is not intuitive or simple or even
particularly actionable.
So why do
we still measure it? I suppose there is some potential merit to trying to
improve output per unit of input, if only the output could be measured and the
input was somehow constrained. I’ve read many explanations for its trend, but
not my own pet theory, which is that it measures the sort of portfolio of work
an economy specialises in. Policy in the UK, and increasingly the US, drives
more people into low wage service jobs. As this constitutes a greater share of
the mix, productivity goes down. But so what?
Productivity
fails so many tests of a good indicator. Perhaps the only reason it is still popular
is to support a pet theory of a politician or partisan economist, such as enabling
union bashing or lobbying for infrastructure investment. It is time to consign
this indicator to the dustbin.
No comments:
Post a Comment