Education & EdTech

Everyone logs in at nine o’clock on Monday.

Education platforms have the least forgiving load profile in software. Usage is not distributed. It arrives in a wall at the start of a lesson, from devices that are older than the students using them, over bandwidth the school cannot upgrade, and if it fails the failure happens to thirty children at once in front of a teacher who now has no lesson.

What buyers here open with

“Will it work on a school laptop, on school wifi, at the start of a lesson?”

1,000 2,000 BELL BELL BELL 08:30 09:00 09:30 10:00 10:30 11:00 ZERO TO 2,400 IN UNDER 20 SECONDS CONCURRENT SESSIONS, ONE SCHOOL MORNING AVERAGE CONCURRENCY IS A NUMBER THAT DESCRIBES NOBODY.
0 Acceptable failures during a live lesson

The pressure

Four constraints that are specific to this sector.

01

Synchronised load

Concurrency is not spread across the day. It is a spike at the top of each period, and your capacity plan has to be built for that shape.

02

Constrained devices

Chromebooks with four gigabytes of memory, tablets two operating systems behind, and a network shared with the whole building.

03

Two sets of users

Learners and administrators want opposite things. One needs a system that gets out of the way, the other needs reporting, permissions and oversight.

04

Procurement reality

The person who buys it is not the person who uses it, budgets are annual, and accessibility and data protection are gating questions rather than nice-to-haves.

The work

What we build for education

Four systems, all of which have to work on the worst device in the building.

Learning platforms

Course delivery, assessment and progress that work offline-tolerant and resume cleanly when a connection drops mid-lesson.

Assessment and marking

Timed assessments that survive a dropped connection without losing a student’s work, and marking tools that give teachers time back.

Administration systems

Enrolment, timetabling, attendance and reporting, including the returns you are statutorily required to file.

Parent and guardian portals

A different audience again, on personal devices, checking one thing quickly.

The device floor

The hardware we test against, not the hardware we would like you to have.

Every performance budget needs a worst case that is real. This is ours, and we publish it so you can hold us to it during procurement.

TargetWhat it isBudgetWhy this one
Baseline laptopChromebook, 4GB RAM, 2019 class silicon< 2.5s to interactiveThe most common device in state schools, and the one that will still be in the cupboard in three years.
Baseline tabletiPad 7th generation, two OS versions behind< 3.0s to interactiveShared trolley devices are rarely current and almost never charged.
Network4 Mbps shared, 120ms latency, 2% packet loss< 400KB initial payloadA whole year group on one connection at nine in the morning.
Concurrency2,500 sessions arriving inside 20 seconds0 dropped sessionsThe bell is the load test. There is no gradual ramp.
InterruptionConnection lost for 90 seconds mid-assessment0 lost answersA student who loses work once will not trust the platform again, and neither will their teacher.
AccessibilityWCAG 2.2 AA, keyboard and screen readerPass before releaseA procurement gate in this sector, checked rather than asserted.

Non-negotiable

What we will not ship without

Accessibility is a gate
WCAG 2.2 AA is not a polish item in this sector. It is a procurement requirement and it is checked.
Children’s data
FERPA, COPPA and GDPR where relevant, with a clear answer to what is collected, why, and how long it is kept.
Works on the worst device
The performance budget is set by the oldest supported device on the slowest supported network, not by the developer’s laptop.
Data loss is unacceptable
A student who loses forty minutes of work to a connection drop will not use the platform again, and neither will their teacher.

Where AI actually lands

Four places it earns its cost, and one line we do not cross.

Worth doing

  • Formative feedback on drafts and practice work, where fast imperfect feedback beats slow perfect feedback that arrives after the topic moved on.
  • Lesson and resource preparation, giving teachers back the hours currently spent adapting materials.
  • Early identification of students falling behind, surfaced to a teacher rather than acted on automatically.
  • Accessibility support, including live captioning, reading support and translation.

Where we stop

Summative grading that affects a student’s record without a human deciding. The accuracy argument is beside the point; the accountability question is who answers to the parent, and it cannot be a model.

Evidence

We build platforms designed for thousands of concurrent users on school-grade bandwidth, load-tested against the start-of-lesson spike rather than an average day.

A person talking through an interface on a laptop during a research session

Next step

Talk to the engineer who would run your build.

No discovery call with a salesperson, no deck. A senior engineer reads what you send and replies with a real assessment, including when we think you shouldn’t build it.

Tell us what you’re building

We reply within one business day. No sales sequence, no newsletter.

Certified, partnered and awarded