How the number is made
Everything the toolkit outputs comes from one model with published rules. The language model classifies; code scores and sizes. This page is the whole method, so an AE or a delivery lead can check the number rather than trust it.
Fitment: the call and the score
Each requirement gets one of five calls. The model makes the call and names the sub-module and feature; the score is computed in code from the calls, so the same input gives the same number.
| Call | Means | Weight |
|---|---|---|
| OOB | Ships in the product. Turn it on, populate it. | 1.0 |
| Config | No-code or low-code setup: rules, flows, catalog items, SLA definitions, templates. | 0.9 |
| Custom | Script includes, custom tables, scheduled jobs, or integrations with logic. Priced as effort. | 0.4 |
| Add-on | Needs a separately licensed product or store app. 0.4 by default, 0.9 if the customer confirms the licence. | 0.4 / 0.9 |
| Another tool | Not the platform's job. Say so; suggest what is. | 0 |
bands: ≥85 strong fit · ≥70 good fit · ≥50 partial · below 50 reconsider
cap: specialist mathematics (actuarial, capital, pricing engines) never scores above 55
Sub-module and feature
"ServiceNow can do this" is not an answer. Every call names the sub-module and the feature that does the work, because that is what decides the licence and the effort.
| Requirement | Sub-module · feature | Call |
|---|---|---|
| Approvers approve changes from Microsoft Teams | Change Management · approvals in Microsoft Teams (Collaboration Services) | Add-on |
| Track Oracle licence entitlements discovered on-prem | SAM Professional · software discovery + entitlement reconciliation | Add-on |
| Auto-categorise incidents from the affected CI | Incident Management · business rule mapping CI class → category | Config |
Effort: one model
Effort is in ES. One ES is 80 person-hours: two people for a week, or one for two. Every module has a base ES from the deals I delivered, and a multiplier from five neutral factors that is capped so nobody can inflate a number by stacking adjectives.
M = f(data volume, process maturity, integration depth, customisation appetite, stakeholder spread) · M ≤ 2.2
overhead: PM + solution architect at 25% of build · QA sits inside base
Integrations are priced by tier (file or REST with no logic · API with mapping · orchestration with state), not by name. A "Teams integration" is a tier-2 integration, whatever the logo.
The reconciliation block
Every estimate prints three views of the same number and they have to agree. If they don't, the estimate is wrong and the tool says so rather than rounding.
team 7.25 FTE × 43 weeks × 16 h/wk build = 4,988 h · Δ 0.6%
price ÷ blended rate = 4,960 h ✓
Heuristic or computed
| Output | How |
|---|---|
| The call per requirement | Language model, grounded on the knowledge base. A judgement; check the ones marked custom. |
| Fitment score and band | Computed from the calls. Deterministic. |
| Scoping questions and field implications | Looked up per module. No model, no cost. |
| Effort, team, calendar, reconciliation | Computed from the model file. Deterministic and versioned; every result carries its kb version. |
| Licence tier and quantities | Rules from the packaging catalogue plus your persona counts. Computed; the traps are heuristics and say so. |
Licences: a position, never a price
The toolkit tells you the minimum tier each capability needs, the true quantity per metric (fulfillers, approvers, requesters, units), which known traps apply, and what to confirm with the ServiceNow AE. It never quotes a price and never mixes licences into the services number. Pricing is ServiceNow's or your reseller's.
Calibration
The model was fitted to two delivered deals, a SEA IT-services programme (6.5 FTE at peak over 36 weeks) and a SEA telco programme (44 weeks), and reproduces both inside 10%. It is re-tested against them on every change, and the fitment model is re-run against 20 test cases. The results are on the proof page.