Mockup Tilli × EEF · Response to ITT: Developing an EdTech practitioner tool · July 2026

The EdTech Implementation Companion, shown through two of the ITT's user journeys

Indicative UI produced for tender evaluation: not an EEF product. Each screen is shown as it would appear embedded natively in the EEF website via Craft CMS. Numbered notes in the right-hand margin map design decisions to the ITT and to our proposal.

Use case 1: Reviewing an existing EdTech investment Use case 3: Exploring GenAI in KS2 English Technical architecture & PII handling →
Use case 1 · ITT §User journeys

A school leader reviewing an existing EdTech investment

The journey below moves from a structured intake, through evidence and an assessment plan built on data the school already holds, to an implementation plan and a presentation the head can take to their leadership team.

Fictional persona
Margaret Osei: Headteacher
St Cuthbert's C of E Primary, Wakefield · 412 pupils · 31% Pupil Premium. Adopted NumberBoost, a maths tutoring platform, in Sept 2024: quickly, without a planned implementation strategy. Two years in, she is unsure of its impact.
Screen 1.1: Choosing a starting point on the EEF website
educationendowmentfoundation.org.uk/support-for-schools/edtech-implementation-companion
EEFEducation Endowment
Foundation
About usEducation evidenceSupport for schoolsProjects & evaluationSearch
Support for schools › EdTech Implementation Companion

Make an evidence-informed decision about technology in your school

A guided companion built on the EEF's implementation guidance. Work through the questions that matter, at your pace: and leave with a written plan you can share.

This tool never recommends or compares products. It helps you work on the decision: whether technology is the right response, what would need to be true for it to succeed, and how you'll know.
Review an EdTech approach we already use
✓
Is it achieving what we hoped? How would we know?
Plan a new approach to a school priority
From an identified need to an implementation plan.
Explore AI in my classroom
For teachers curious about GenAI, with or without a specific idea.
Support disadvantaged pupils
Plan and monitor equitable implementation of Pupil Premium spending.
No account needed · nothing you type is stored with your name · how we handle your data
Start reviewing →
1
Delivered as a native Craft CMS feature of the EEF website: no separate product, login or domain. RFP §4.6
2
The four entry routes are the four ITT user journeys. One conversational engine, four framings. ITT use cases 1–4
3
Product-agnostic by design, stated before the first question. The tool holds no product database, so it cannot recommend one. RFP §3.2
Screen 1.2: Explore: structured questions with conversational moments
EEFEdTech Implementation Companion
Explore Prepare Deliver Sustain
EEF
Let's start with why NumberBoost was adopted in the first place. What problem was it meant to solve? It's fine if this was never written down: that's common, and worth surfacing now.
Low attainment in maths ✓ Teacher workload Catch-up after absence It felt expected of us Something else…
Honestly, a mix. Maths results in Years 3–4 had dipped, but there was also pressure to be seen doing something with technology. We rolled it out to everyone in one term.
EEF
That's a really useful distinction: a dip in Years 3–4 is a specific, checkable problem; "being seen to use technology" isn't one the evidence can help with. Let's anchor the review to the first. Before we look at any data, one question the EEF asks every school to sit with:
Equity check · required
Who in your school is NumberBoost least likely to be working for: and why?
Your answers build your plan: nothing is scored Continue →
4
The journey is controlled by the application, not the model: the four EEF implementation stages are always visible and always in order. RFP §4.1
5
Hybrid interaction: structured choice chips keep the diagnosis on track; free-text moments capture the school's real context: and gently challenge assumptions.
6
The equity check cannot be skipped. Barriers named here resurface as monitoring lines in the final plan. RFP §3.4
Screen 1.3: What the evidence says (and doesn't)
EEFEdTech Implementation Companion
ExplorePrepareDeliverSustain
EEF
Here's what the EEF's evidence says about approaches like NumberBoost: an online tutoring programme offering individualised maths practice. I'll be direct about where the evidence is strong and where it isn't.
Individualised instructionStrong evidence base
Moderate impact (+4 months) for very low cost: but effects depend heavily on how it's implemented, not on the software itself.
Teaching & Learning Toolkit → Individualised instruction
Technology-delivered tutoring specificallyMixed evidence
Results vary widely between programmes and settings. Usage time alone is a poor predictor of impact; quality and integration with teaching matter more.
EdTech Evidence Review (Haßler et al., 2025), §5.2
Whole-cohort rollout without a pilotLimited evidence
The evidence can't tell you whether your rollout worked: but the DIY Evaluation Guide can help you find out with data you already hold. That's our next step.
EEF DIY Evaluation Guide → Before-and-after comparisons
Where we're honest about uncertainty: the evidence does not support a confident answer on whether NumberBoost is working in your school. No summary can. What it supports is a way of checking: which is what we'll build now.
7
Every substantive claim carries a link back to the EEF publication it was retrieved from: responses are generated only from retrieved EEF evidence. RFP §4.4
8
Evidence-strength labels (strong / mixed / limited) use consistent language for uncertainty. A tool that never says "the evidence is limited here" is not an evidence tool. RFP §3.3
Screen 1.4: Building the assessment plan from data the school already holds
EEFEdTech Implementation Companion
Explore ✓PrepareDeliverSustain
EEF
You mentioned you run PUMA standardised maths assessments each term. That's enough to look at before-and-after patterns for the year groups using NumberBoost: no new testing needed. I've laid out your figures the way the DIY Evaluation Guide suggests reading them.
Year 4 cohort · PUMA standardised score (national average = 100)
entered by school · aggregates only
Summer 2024: before NumberBoost
97.8 all pupils
94.0 Pupil Premium
Summer 2025: after one year
99.1 all pupils
94.4 Pupil Premium
Spring 2026: latest
99.6 all pupils
94.9 Pupil Premium
What this pattern can and can't tell you: scores have risen for all pupils: but the gap for Pupil Premium pupils has barely narrowed (3.8 → 4.7 points). And without a comparison, a rise can't be attributed to NumberBoost: curriculum changes, staffing and cohort differences could all explain it.
Mixed confidence. Before-and-after comparisons are the weakest evaluation design in the DIY Evaluation Guide: useful for spotting patterns, not for proving impact. We'll strengthen it below.
Your assessment plan so far
1Existing outcome data: termly PUMA scores, Years 3–5, tracked separately for Pupil Premium pupils. Already collected; zero new workload.
2A comparison: Year 5 uses NumberBoost 3×/week, Year 6 doesn't. Comparing their trajectories is a stronger design than before-and-after alone.
3How it's actually used: a short sentiment survey for parents, staff and pupils (next step), plus NumberBoost's own usage reports read critically.
9
The assessment plan is scaffolded directly from the EEF DIY Evaluation Guide: using data the school already holds, before asking for anything new.
10
Aggregate figures only. The tool never asks for pupil-level data; anything pasted is scanned and PII is stripped before it reaches the model: see the architecture document. RFP §3.2
11
The tool is honest that before-and-after can't prove impact: and responds by strengthening the design, not softening the claim. The PP gap is surfaced, not hidden.
Screen 1.5: Designing the sentiment survey for parents, staff and pupils
EEFEdTech Implementation Companion
Explore ✓PrepareDeliverSustain
EEF
Test scores tell you whether things changed; people tell you why. Here's a draft five-question survey for each group. Everything is editable: and I've kept the parent version to a two-minute read.
Parents & carers Teaching staff Pupils (Y3–5)
1. My child talks about what they do in NumberBoost sessions.
NeverSometimesOften
2. My child seems more confident with maths this year.
DisagreeNot sureAgree
3. My child can use NumberBoost easily at home.
If not: what gets in the way? device · internet · time · knowing how to help
4. How does your child feel about maths lessons at the moment? (really dislikes → really enjoys)
5. Anything about your child's maths learning you'd like to tell us? (open answer)
Plain English · reading-age checked · question 3 exists because of your equity answers
Edit questions Print version Copy to Google/Microsoft Forms
12
Multi-informant by design: parents, staff and pupils each get an audience-appropriate version, drawing on Tilli's multi-informant assessment platform experience. RFP §8.2
13
Surveys export to the tools schools already use (print, Google/Microsoft Forms). Responses stay with the school: the companion never collects them.
Screen 1.6: The implementation plan, and a version for the leadership team
EEFEdTech Implementation Companion
Explore ✓Prepare ✓DeliverSustain
Implementation review plan · draft 1 · 12 Feb 2027
Reviewing NumberBoost at St Cuthbert's: is it improving maths outcomes for all our pupils?
1 · The problem we adopted it for
Dip in maths attainment in Years 3–4 (2023–24). Rollout was whole-school, in one term, without a written implementation plan.
2 · The decision question
Continue (and improve), scale back, or replace: decided on evidence by July 2027, ahead of contract renewal.
3 · Assessment plan
Termly PUMA scores (all pupils + PP separately) · Year 5 vs Year 6 comparison · usage reports read against outcomes.
4 · What people tell us
Sentiment surveys: parents, staff, pupils: spring term. Home-access barriers tracked explicitly.
5 · Risks & equity watchpoints
PP gap not closing (4.7 pts) · home device access · text-heavy instructions for weaker readers.
6 · Monitoring & next review
Half-termly check-ins; intermediate outcomes defined per the EEF implementation guidance; full review July 2027.
Download PDF Edit in Word
Need to bring your SLT with you? I can turn this plan into a short briefing deck for your leadership meeting: same content, presentational format.
NumberBoost: is it working for our pupils?
SLT briefing · Feb 2027
What the data shows
Scores up; PP gap not closing
What parents, staff and pupils tell us
Decision points for SLT
Continue · scale back · replace
Generate briefing deck Download .pptx
Every slide keeps its evidence links and uncertainty labels: the deck can't overclaim what the plan doesn't.
14
Every conversation leaves something useful behind: a written plan with problem, approach, risks and monitoring: not a chat transcript. RFP §3.2
15
One click converts the plan into a briefing deck for the leadership team: the "share with governors and colleagues" moment, designed in. Uncertainty labels survive the format change.
Use case 3 · ITT §User journeys

A classroom practitioner exploring GenAI: without a specific idea yet

The companion meets the teacher where they are: curious but unfocused. It narrows curiosity into a teaching problem, checks whether technology is the right response, and ends with a small, safe classroom trial.

Fictional persona
Daniel Price: Year 5 teacher
Brackenfield Primary, Shropshire. Generally aware of GenAI's potential to support teaching KS2 English, but has no specific ideas for how to apply it: and limited time to find out.
Screen 3.1: From vague interest to a real teaching problem
EEFEdTech Implementation Companion · Exploring AI in my classroom
ExplorePrepareDeliverSustain
I keep hearing about GenAI and feel like I should be using it in my English lessons, but I honestly don't know where I'd start with Year 5.
EEF
"Feeling like you should" is where a lot of teachers are: and it's not a reason to adopt anything yet. Let's flip it: instead of starting from the technology, let's start from your teaching. In your English lessons, what takes the most of your time, or where do pupils struggle most?
Giving useful feedback on writing ✓ Finding good model texts Vocabulary teaching Differentiating tasks Planning time
Feedback, definitely. Thirty books after every extended write: my comments end up rushed, and I'm not sure the children act on them anyway.
EEF
Now we have a real problem: feedback on extended writing is slow to produce and pupils don't always use it. That's something the evidence speaks to directly: and it gives us a fair test for whether GenAI helps. Shall we look at what the evidence says before deciding anything?
16
The tool is designed for the teacher who doesn't yet have a specific idea: exactly the ITT's use case 3 framing. Curiosity is a legitimate entry point.
17
Problem before product: the conversation deliberately inverts "what tool should I use?" into "what teaching problem am I solving?": the EEF Explore stage, applied.
Screen 3.2: Is technology the right response? The evidence, with its limits
EEFEdTech Implementation Companion · Exploring AI in my classroom
ExplorePrepareDeliverSustain
EEF
Two honest findings. The evidence on what makes feedback work is strong and doesn't depend on any technology. The evidence on GenAI producing that feedback is early and limited. Both matter for your decision.
What makes feedback improve learningStrong evidence base
Feedback works when it's specific, task-focused, and pupils get time to act on it. Timing and pupil response matter more than volume of comments.
Guidance report → Teacher Feedback to Improve Pupil Learning, recs 2–5
GenAI for generating pupil feedbackLimited evidence
Early studies suggest GenAI can draft plausible feedback and reduce teacher time: but accuracy varies, and there's no reliable evidence yet on pupil outcomes. Teacher review of every output is essential.
EdTech Evidence Review (Haßler et al., 2025) → Emerging technologies
A non-tech option, considered seriouslyStrong evidence base
Whole-class feedback and structured redrafting time address the same problem with no new technology. If GenAI doesn't earn its place in the trial below, this is the fallback: not "more of the same".
Guidance report → Improving Literacy in KS2, rec 4
18
Strong pedagogy evidence and limited GenAI evidence are visibly distinguished: the tool doesn't let novelty borrow credibility from the feedback literature. RFP §3.3
19
A credible non-technology alternative is always on the table: "is technology the right response?" is asked, not assumed.
Screen 3.3: A small, safe classroom trial: with data hygiene built in
EEFEdTech Implementation Companion · Exploring AI in my classroom
Explore ✓PrepareDeliverSustain
EEF
Here's a six-week trial sized for one class and one unit: teacher-only use first, so no pupil ever interacts with the AI. Small enough to stop without cost, structured enough to learn from.
Classroom trial plan · Year 5 English
Can GenAI-drafted feedback (teacher-reviewed) improve how pupils redraft persuasive writing?
Scope: Persuasive writing unit, 6 weeks, one class. Teacher drafts feedback with GenAI, reviews and edits every comment before pupils see it. Pupils never use the tool.
What stays the same: Success criteria, marking policy, redrafting lesson structure.
Success measures: Time per book (self-logged) · quality of pupil redrafts against criteria · pupil survey: "did the comments help you improve?"
Stop conditions: Feedback quality below teacher's own standard · time savings don't materialise by week 3.
If it doesn't earn its place: Switch to whole-class feedback + structured redrafting (evidence-backed, no tech).
Before you paste any pupil writing anywhere:
✓ Remove names and anything identifying: first names in stories count.
✓ Use your school-approved tool, never a personal account.
✓ Check your school's data protection policy covers this use.
Inside this companion, anything you paste is automatically scanned and personal details are stripped before any AI processing: see how this works.
Download trial plan Share with my head Set a week-3 check-in reminder
20
The output is a small, reversible trial with stop conditions: proportionate to limited evidence. Teacher-first use means no pupil interacts with AI during the trial.
21
Data hygiene is part of the pedagogy: practical PII rules for the teacher, plus automatic PII stripping inside the tool itself: detailed in the technical architecture. RFP §2.2
Tilli × EEF · mockup for tender evaluation · all schools, staff and data fictional · EEF branding used indicatively
Continue to the technical architecture & PII handling →