Schools have given pupils "wildly inaccurate" predictions of their exam performance under a data monitoring system now widely used, teachers have claimed.
The achievements of pupils against these estimates are then used in many teachers' performance management reviews, and in local authority judgments on school effectiveness.
The Government is also urging schools and local authorities to use the data system, run by the Fischer Family Trust charity, in target-setting decisions, even though the trust cautions against this.
However, its data system is enthusiastically backed by some teachers, who say it helps set achievable goals for pupils.
The Fischer Family Trust, based in South Wales, provides all local authorities in England and Wales with data analysis.
The system provides estimates of what a child might be expected to achieve at later key stages, based on their past performance and the average achievements of children judged by the trust to have similar characteristics.
But it is generating unhappiness in some staffrooms. Secondary teachers of non-core subjects such as drama, music and PE have told The TES of their concerns.
At secondary level, Fischer Family Trust bases its estimates of what pupils might be expected to achieve in GCSEs in each subject, given their test and teacher assessment performance in English, maths and science at key stage 2 and 3. But teachers say that these are no gauge of ability in drama, music, art and PE. Under the trust's system, the estimates of children's likely achievement in GCSE drama are the same as those used for GCSE English, The TES has learned. Clive Hulme, a drama teacher who has been looking into the use of the data and has spoken to teachers about it, made the discovery and wrote about it in Teaching Drama magazine. He has spoken to heads of department who find the system very useful.
But he also said that the data was widely misused. In its guidance, the trust says that the grades generated by its systems should be seen as "estimates" rather than predicted marks or targets.
Teachers can interpret the data taking into account what they know about their pupils to produce their own predictions or target grades, it says.
However, Mr Hulme said that in many schools, the trust's scores were simply being transferred into performance management systems, without any checks. He said: "Most of the teachers that we spoke to say the trust's predicted grades are being used to set performance management objectives, as part of upper pay spine (threshold) progression.
"In some schools, performance management objectives have been set along the lines of 'all of your students will meet predicted grades'."
One music teacher told The TES: "Our school uses the trust as a basis for all data. If we don't reach its estimates, we are required to explain ourselves to the governors."
A teacher from a large urban primary in the North said: "We are under increasing pressure to make sure the results match the predictions. If we get any that are above expectations, we are told to downgrade children to benefit the next teacher up the school next year."
Campbell Russell, a former secretary of the National Union of Teachers' Cheshire division, has been collecting evidence of teachers being denied progression on the upper pay spine "entirely on the basis of Fischer Family Trust applied outside of core subjects".
Schools are also at risk of being labelled as underperforming if they fail to hit estimates generated by the data, he said.
The latest guidance on target-setting from the Department for Children, Schools and Families which was sent to local authorities earlier this term, appears to put pressure on schools to set targets based on the most stretching estimate of future performance generated by the trust's data. Under this system, four predictions of pupil performance are made, designed to be more or less stretching.
The most demanding is its model D estimate, which schools would match if, on average, their pupils performed as well as the top-performing 25 per cent in England.
The guidance says: "Where trust estimates are used to inform target-setting, schools should be directed towards model D, provided the school is not exceeding model D estimates." To sceptics such as Mr Russell, this suggests the Government is not taking seriously the trust's own statement that the data must be used cautiously.
Mike Treadaway, the trust's research director, said that what the DCSF was saying was "not identical" to the trust's own guidance on the use of data. Asked which one was right, he said: "It depends what you mean by right. We have been consistent in what we have been saying all along."
He said, however, that the Government's advice was slightly weaker than it had been two years ago, when it said schools should "use" model D, rather than be "directed towards" it.
Mr Treadaway said if data was being misused, all those working in and with schools had to bear some of the blame. This was a wider issue than simply the trust's data, but applied to other systems, such as the Government's Raise, a self-evaluation scheme similar to the trust.
Mr Treadaway denied that the trust's estimates were "useless" in non-core subjects. He said that it was made clear to teachers on the documents in which they were made that they were not accurate for every child. Estimates of secondary performance had a 70 per cent accuracy rate, he said.
Mr Hulme said often the problem was that those using the data had not been trained properly in how it should be interpreted.
"It's not the trust's fault. They are trying to give schools information to improve results," Mr Hulme added. "Unfortunately, what should be used as a carrot to make children strive harder is often being used as a stick to beat teachers with."
- Do you have views on data use in schools? Email firstname.lastname@example.org or email@example.com
HOW TRUST WORKS
The Fischer Family Trust is a charity set up by Mike Fischer, co-founder and former chief executive of RM, the education computer company. It provides data analysis for local authorities, and has expanded rapidly in recent years. In 2001, only a third of authorities used its services. Now they all do.
The trust analyses how groups of children have performed in the past to provide estimates of how similar pupils might do in the future.
It looks at how each pupil has already performed - mainly using test and teacher assessment data in core subjects - to offer estimates of likely performance at the end of future key stages.
These estimates are then added up to come up with ones for each school, and for each local authority.