Descartes’ dream at Severn Trent water supply
Mary Monro describes how advances in data analysis and presentation have enabled water supply managers at Severn Trent Water to build up a more coherent picture of performance.
One of the major changes of recent years is the vast amount of accurate data
which modern instrumentation makes available. However, it comes in a variety
of forms from a variety of sources, and can be utilised in many different
In order to transform raw data into charts, the information needs to be
plotted and statistical analysis is then required to indicate variability in
relation to standard deviations. By demonstrating how many of the readings
fall within preset limits, the charts act as a measure of performance
against specified criteria.
As part of an overall policy, Severn Trent has been setting more exacting
targets for drinking water quality which operatives must endeavour to meet.
The aim is to encourage better process control. To help them to achieve this
the company has been instituting a program through which operators can
generate their own control charts, giving them a quick and simple way of
measuring their performance. By processing their own data, staff will be
able to see how close they are to the desired results, and compare
improvements in performance over time. Regular blips might identify
recurring problems which could be solved by a change in procedures which
might otherwise have gone undetected. Comparisons with other works would
provide an added incentive or indicate the need for new or additional
It was, however, clear that if staff were to be encouraged to produce and
use their own control charts, it was important that creating them should be
made as easy and painless as possible. Working out how this could be done
has been the task of a team at Severn Trent¹s Water Supply Regional office
The Worcester team started work in June 1997. “We had vast amounts of data
relating to a wide variety of different parameters, some of which are easier
to measure than others,” said David Smith, the team leader. “One of the
first moves was to plot benefit against feasibility for a number of
different criteria to decide which elements would be give the most useful
The decision was that control charts would be produced for levels of four
parameters: fluoride, phosphate, chlorine and pH.
The second question was to decide which set of data should be used. Raw data
are available from several sources SCADA, laboratory tests, etc and
these were assessed in terms of their number, availability, accuracy,
operator involvement and reliability. The one set which fulfilled all the
criteria was produced by the operators themselves as a result of the routine
testing and monitoring which they carried out during their work.
“We then had to work out the best way to implement the program,” explained
David Smith “This meant deciding among other things how often the results
would be recorded, how target changes would be incorporated, who would
record the results and how and where the results would be displayed.” Of
even more importance was who would be responsible for the statistical
calculations required to produce the charts.
When the proposal was put to the operatives themselves this last issue was
the major hurdle. Staff saw the production of the charts as an expert job,
requiring specialised skills which had not previously been required of them.
They were understandably concerned that they might need training in
statistics and worried that if the wrong information was recorded it would
reflect on their performance.
The key factor in reassuring staff over these anxieties was the decision to
automate chart preparation by using a computer. What was needed was a
program which would carry out all the necessary calculations and, in effect,
turn the recorded readings into charts at the press of a few buttons. The
problem then turned to finding a suitable program.
“We looked into using one which already existed, but could not find anything
which we felt did what was needed,” said David Smith. “In the end a member
of the team wrote the program to match our specification.” Trials followed
at Strensham, Mythe and Draycote treatment works to sort out any problems.
After a short training period the existing staff were successfully producing
their own charts of process variability based on the information entered by
Having proved the system was practical and effective, it was then worked up
into a standardised package, both as to program and procedures, for wider
To date it is in operation for Severn Trent’s potable water operations in
the Southern Shires district, and is gradually being brought in to the East
and West Midlands districts.
Giving operators the ability to handle their own raw data and prepare their
own charts in this way means they are far more directly involved in the
processes they control. “We are bringing it all down to operator level
they produce the charts and interpret the results. They can see where
anomalies are happening, find out why, and decide themselves what to do to
improve matters,” said David Smith. And as he pointed out, by comparing the
charts month by month they can see the effect of their efforts to produce
For management, the charts can provide a coherent picture of overall
performance. They help to ensure every effort is made to optimise
processes, identify where improvements can be made through capital
investment and help to prioritise in which areas these investments will be
But above all, instituting the policy of monitoring the processes in this
way encourages the aim for constant and continuous improvement in quality
and hence greater customer satisfaction.