Where the line is

Not the volume.
The consequence.

Schools are told, roughly once a year, that they should be more data-driven, and roughly once a year somebody responds that children are not data. Both positions contain something true and neither is usable on a Monday. The workable distinction is narrower than either: a measurement is legitimate when it exists to change what an adult does for a child, and becomes surveillance when it exists to record what the child did.

That test sorts the usual cases quickly, and it does not track the quantity of data at all — an attendance register has been collected in schools for a century and is squarely on the right side of the line, while a single continuously-updated screen-time figure is on the wrong one.

The measurement
Which side
Why
A flag that a student missed four sessions
Care
Acts on absence the school already records, and ends in a phone call
A class average and its score bands
Care
Aggregate. Tells a teacher where to re-teach, names nobody
Minutes a child spent on screen
Surveillance
Requires continuous monitoring and rewards slowness
A ranked list of children by average
Surveillance
Changes nothing a teacher does; changes how a child is seen
Keystroke or webcam monitoring in assessment
Surveillance
Treats every child as a suspect to catch a handful

This article is about the operational design of reporting. The separate question of what the law requires of you — lawful basis, notices, retention, rights — is the subject of the data protection cluster, and nothing here is legal advice.

Decision one

Aggregate is a privacy control,
not a performance one.

The most effective single protection in learning analytics is refusing to let the reporting layer see the content of work. A school page can be genuinely useful knowing only that a class averages 71, that its scores fall into four bands, and that eleven answers are waiting to be graded. It does not need to know what any child wrote to tell a head of department that one set is a long way behind another.

SprintUp is built this way deliberately: the analytics never load an answer payload or any question content. They count and average records that already exist. The practical effect is that the reporting system cannot become a window into individual children's work even for someone with access to it, because the content is not in what it reads.

Individual work is not hidden from the people who should see it — a teacher reads and grades their own students' answers, which is the job. The point is that reporting and teaching have different access needs, and conflating them is how a dashboard quietly becomes a monitoring tool for whoever has the widest login.

Decision two

Publish the rule
that flags a child.

If a system is going to mark a student as needing attention, the rule that did it should be legible to the staff acting on it and, in substance, to the family. Hidden scoring is where analytics loses schools their trust: a parent told their child has been flagged by the system, and a school unable to say what the system measured, is a bad conversation with no good ending.

The flags SprintUp uses are deliberately plain and arithmetic rather than inferred: attendance below 90% once there are at least three records to judge on, an average below 50, a fall of 15 points or more against the previous period, a fortnight with no activity, and an account never signed into. The thresholds are sent back with the data, so the page describing what a flag means is reading the same numbers the calculation used and the explanation cannot drift away from the behaviour.

Two details make this humane rather than merely transparent. The minimum of three attendance records stops a student being labelled on the basis of one missed session in their first week. The fortnight rule serves as a grace period for new accounts, so nobody has to learn to ignore the list every September. Both are the sort of thing that only shows up once real schools use a feature, and both are worth asking a vendor about explicitly.

🔍A question for any vendor
"Show me the exact rule that puts a student on your at-risk list." If the answer is a model, ask what it was trained on, who can see its output, and what a school is expected to tell a parent who asks why their child is on it.
Decision three

Decide now
what you will not build.

Every school reaches a moment where somebody — a board member, an ambitious member of staff, a vendor's sales engineer — proposes a report that the data would technically support and that the school should not keep. It is much easier to decline in advance, as policy, than in the moment, as an opinion.

The proposal
The argument for it
Why it does not survive
Student league tables
"It motivates them"
It motivates the top three and teaches the rest their place
Engagement scores per child
"It flags disengagement"
A composite hides which input moved; absence already flags it
Parent-facing live activity feeds
"Parents want visibility"
Real-time monitoring of a child by a parent is a different product
Teacher performance rankings
"It drives standards"
Class averages measure intake at least as much as teaching

The last row is the one most likely to be unpopular internally and it matters most for staff trust. Class averages reflect who is in the class at least as much as who teaches it, and a school that ranks teachers on them will find, within a term, that its best teachers stop volunteering for the hardest groups. Workload and grading turnaround are fair things to look at with a teacher; a league table of averages is not.

Decision four

Write it down
before anyone asks.

A short, readable statement of what the school measures does more for family trust than any amount of careful practice nobody knows about. It does not need to be a policy document. Half a page at enrolment, in the language families actually speak, covering four things: what the school records about a child, what it uses each thing for, who inside the school can see it, and what the school does not do.

That last part is the one that earns the trust, and it is the part almost nobody writes. "We do not track how long your child spends on screen, we do not rank children against each other, and no member of staff outside your child's teaching team sees their work" is a paragraph most schools could write truthfully today and very few have.

It also has a useful internal effect: a school that has published what it does not do has a ready answer the next time something is proposed that would contradict it.

The staff half

Teachers notice
which way the dashboard points.

Teachers work out very quickly whether a reporting system exists to help them or to check on them, and their answer determines whether your data is any good. A school whose staff believe the analytics are a monitoring tool will get careful, defensive data entry — attendance marked in bulk at the end of the week, grading timed to look responsive — and every number downstream becomes fiction.

Two practices keep this honest. Show teachers their own figures before anybody else sees them, so nothing arrives as a surprise in a meeting. And make sure a figure means the same thing everywhere: in SprintUp a class average on the school page is produced by exactly the same routine as the one on the teacher's own page, so the two cannot disagree. Reporting that contradicts a teacher's own screen loses the argument permanently, whichever number happens to be right.

Before the term

The line
checklist.

✅Seven items
Every measurement passes the "changes what an adult does" test · the reporting layer cannot read the content of children's work · the rule behind any at-risk flag is written down and shareable · minimums and grace periods exist so nobody is flagged on one data point · a written list of reports the school will not build · half a page for families covering what you do not do · teachers see their own numbers first, and every figure means the same thing on every screen.
Next cluster: data protection and GDPR →← Back to A2