From 64 signals to a single score.
Your score isn’t guesswork. It’s the output of a transparent model that combines questionnaire data with consultant judgement. Below, we walk through every step. The plain-English version comes first, and the math is there for anyone who wants it.
Four steps from raw answers to a published score.
Every report follows the same four steps. The first two are about gathering raw evidence. The last two turn that evidence into the numbers you see on your cover page.
Gather Information
Two inputs feed the model: your pre-assessment questionnaire and your consultation call.
The questionnaire captures hard facts — systems in use, backup processes, headcount, integrations. The consultation captures context, nuance, and the things you would not think to write down. Together they cover the 64 signals the model looks at.
Score Each Signal
Every signal gets a value between 0.0 and 1.0. There are three ways that value is set.
Tier 1 signals are extracted deterministically from questionnaire answers (e.g. backups in place: yes / no). Tier 2 signals are counted with diminishing returns (e.g. number of digital systems). Tier 3 signals are scored by your consultant against a five-level rubric using the consultation transcript.
Roll Up To Six Focus Areas
Signals are weighted within their focus area, then rolled up into a 1.0 to 5.0 score.
Each focus area has its own weights. A backup process matters more than a single storage system, so it carries a higher weight. The result is six dimension scores on a 1.0 to 5.0 scale, which are also shown as 0 to 100 for easy comparison.
Combine Into An Overall Score
The six dimension scores are combined using a weighted geometric mean, then gate checks run.
A geometric mean penalises weak dimensions more heavily than a simple average, so a single major gap is reflected honestly in the overall number. After the mean is calculated, a gate check caps the score if any dimension falls below a safety threshold.
Six dimensions, weighted by impact on AI readiness.
Not every focus area carries the same weight. Data Quality is worth a quarter of your score because a model is only as good as the data it learns from. Strategy is lower-weighted because, while important, it follows once the foundations are in place.
Data Quality
25%How accurate, complete, and accessible your data is across the business. AI is only as good as the data it learns from.
Technology
20%Your core systems, integrations, and cloud setup. Your tech stack sets the ceiling on what AI can realistically do.
People & Skills
15%Technical know-how, training, and how confidently your team adopts new tools. AI only delivers value when the people using it can get results.
Governance
15%Policies on data ownership, access, and risk. Clear rules keep things safe and accountable as AI takes on more decisions.
Use Cases
15%Whether you have identified realistic places AI could help, and which to tackle first. Picking the right projects turns AI spend into results.
Strategy
10%Leadership commitment, budget, and how AI fits a longer-term plan. A clear direction stops effort being wasted.
The weights are intentional. They reflect what we consistently see drives AI readiness in UK SMEs at different stages, and they stay the same for every report so scores remain comparable across clients.
Every signal that contributes to your score.
Each focus area is made up of between 9 and 16 signals. Some are extracted directly from your questionnaire. Others are scored by your consultant using a five-level rubric. The full catalogue is below.
- T1Data backup process in placew 0.15
- T2Digital vs paper data storagew 0.15
- T2Storage method diversityw 0.10
- T2Data quality frustrationsw 0.10
- T3Backup testing maturityw 0.10
- T3Data knowledge dependencyw 0.08
- T3Data entry duplicationw 0.08
- T3Version control maturityw 0.07
- T3AI-readiness of data qualityw 0.06
- T3Disaster recovery preparednessw 0.06
- T3Infrastructure currencyw 0.05
Why we use a weighted geometric mean.
A simple average can hide a critical weakness. An organisation with excellent technology but poor governance would look healthier than it really is. The weighted geometric mean penalises imbalance, so any single weak dimension pulls the overall score down honestly.
It is the same approach used in the UN Human Development Index, for the same reason: a country cannot offset poor education with excellent income and call itself developed.
Imagine a company scoring a perfect 5 on five focus areas but only 1 on Data Quality (the highest-weighted area). Here is what each formula returns.
Simple average
4.3Looks great. Hides the critical gap.
Weighted geometric mean
3.3Pulls the score down to reflect the gap.
A safety net for critical gaps.
Even after the geometric mean, there are some gaps so critical that we cap the overall score until they are addressed. No backup process, for example, is the kind of thing that can sink a business overnight. The gate system stops a strong overall number from glossing over that.
| Focus area | If dimension score falls below | Overall score is capped at |
|---|---|---|
| Data Quality | 1.5 | 40/100 |
| Technology | 1.5 | 40/100 |
| People & Skills | 1.5 | 45/100 |
| Governance | 1.5 | 35/100 |
| Use Cases | 1.0 | 50/100 |
| Strategy | 1.0 | 55/100 |
Worked example
Imagine a company scores well overall (geometric mean works out at 68/100) but their Governance score is 1.2, below the 1.5 threshold. The gate triggers and caps the overall score at 35/100 until governance improves. The cap is not a punishment. It is honest signal that one weakness is severe enough to dominate the picture.
What the numbers actually mean.
Two scales appear in your report. The overall score sits on 0 to 100. Individual focus areas sit on 1.0 to 5.0 (also shown as 0 to 100). The labels below are how we describe each band in plain English.
Overall score (0 to 100)
| Label | Range | Meaning |
|---|---|---|
| Initial | 0–25 | Little or no structured approach. |
| Developing | 26–50 | Some awareness, early efforts underway but inconsistent. |
| Established | 51–70 | Foundations in place but gaps remain in key areas. |
| Advanced | 71–85 | Good practices established and consistently followed. |
| Optimised | 86–100 | Data and AI deeply embedded, continuous improvement. |
Focus area score (1.0 to 5.0)
| Label | Raw 1.0–5.0 | Scaled 0–100 |
|---|---|---|
| Ad Hoc | < 1.5 | < 13 |
| Emerging | 1.5–2.49 | 13–37 |
| Defined | 2.5–3.49 | 38–62 |
| Managed | 3.5–4.49 | 63–87 |
| Optimised | 4.5+ | 88–100 |
RAG status
Each focus area also gets a RAG label so you can see at a glance where attention is needed.
Significant gaps.
Needs attention.
On track.
Why your score has a ±3 margin.
Any assessment is a snapshot. The same business answered on a different day, or interpreted a question slightly differently, could land a few points either side. We acknowledge that explicitly with a ±3 point band shown on your score circle.
The band is calibrated from the natural variation we observe in how organisations interpret questionnaire items and how consultants apply rubric scoring. A wider band would dilute the signal. A narrower one would overstate the precision. Three points is the honest middle.
If your score is 62 with a ±3 band, treat it as “Established, with real shape to it” rather than “exactly 62 out of 100.” The recommendations in your report do not change within the band.
How the pipeline produces a single number.
Meet a fictional UK SME we’ll call Acme Logistics. They are an early-stage candidate for AI: digital-first, but with patchy data hygiene and no formal governance. Here is what the model does with their answers.
Geometric mean
Weighted geometric mean of the six scores comes out at 2.47 on the 1.0–5.0 scale.
Scale to 0–100
Mapped linearly via (2.47 − 1) ÷ 4 × 100 gives a provisional headline of 37.
Gate check
All dimensions sit above their gate thresholds, so no cap applies. Final overall score: 37/100. That puts Acme in the Developing band.
The same six numbers under a simple average would have produced 2.55. The geometric mean knocks 0.08 off precisely because People & Skills (1.8) and Governance (2.0) drag the picture down honestly.
Ready to see your own number?
Every report on this site is built using the model above. No black box, no proprietary magic. If you want a deeper look first, the sample report walks through a complete worked example.