ICP Refiner from Closed-Won Cohort
Stop guessing your ICP. Walk your closed-won cohort through HG firmographic plus technographic plus spend and surface the features that actually distinguish them.
ICP Refiner: Cloud Cost Optimization — German Industrial Enterprise Cohort
ICP Paragraph
Your ideal customer is a large-scale, EMEA-headquartered German manufacturer with revenues above €20 billion and a workforce of 100,000+ employees, operating across diversified industrial or advanced-technology sectors (electronics/instruments manufacturing, automotive systems, chemicals). These companies run mature, multi-cloud footprints — Microsoft Azure and AWS both verified at high intensity across hundreds of global locations — and carry modeled cloud services spend in the $270M–$580M range, representing roughly 19–24% of their external IT budget. They exhibit strong AI and GenAI forward posture (AI maturity scores of 56–77, classified as "AI-leader accelerating" with growing investment) and an "autonomous" automation stage, signaling that cloud cost complexity is actively expanding as AI/ML workloads grow. Critically, the won cohort shows private/hybrid cloud posture (not pure public cloud) combined with verified multi-cloud installs, creating exactly the cost sprawl where optimization tools deliver value. The contrast set (thyssenkrupp, Evonik) sits in smaller revenue bands ($16–37B), lower AI maturity (scores 27–33, plateau trajectory), and significantly thinner cloud estates, reducing both the pain point and the budget authority for optimization tools.
Firmographic Distinguishing Features
(HG firmographic, May 2026)
| Feature | Won Cohort (Siemens, Bosch, Continental, BASF) | Contrast Cohort (thyssenkrupp, Evonik) | Signal Source |
|---|---|---|---|
| Revenue band | $22B–$103B; median ~$80B — all Forbes 2000 top-600 | $16B–$37B; both Forbes 2000 1100–1400 range | HG firmographic revenue |
| Employee band | 92K–418K; 3 of 4 exceed 100K | 31K–93K; both below 100K | HG firmographic employeeCount |
| Industry | Two automotive/electronics (NAICS 3363/3344), one chemicals (NAICS 3251), one mixed manufacturing (NAICS 3363) — all complex, multi-BU operations | Steel/metals (NAICS 3312) and specialty chemicals (NAICS 3259) — narrower vertical footprint | HG firmographic industryCodes |
| Geography / HQ | All Germany/EMEA; global multi-region tech footprints verified in US, IN, DE, SG | Same Germany/EMEA HQ, but lower international tech install density | HG firmographic location + technographic locations |
| Modeled IT spend (total) | $391M–$4.0B; cloud services $52M–$581M (19–24% of external IT) | $318M–$347M; cloud services $62M–$75M (19–22% of external IT) | HG modeled spend company_spend |
Note on the cloud spend % ratio: The cloud-as-a-share-of-IT figure is similar across cohorts (~20%), but the absolute dollar scale in the won cohort is 4–8× larger — making cost optimization ROI proportionally more compelling and easier to justify procurement.
Technographic Distinguishing Patterns
(HG technographic, May 2026 — sorted by intensity)
| Pattern | Won Cohort Adoption | Contrast Cohort Adoption | Signal Source |
|---|---|---|---|
| Multi-cloud depth: Azure + AWS both at high/medium intensity | Siemens: Azure intensity 3,788 (189 locations) + AWS 3,633 (198 locations). Bosch: Azure 1,777 (26 locs IN) + AWS 1,252 (45 locs US). BASF: Azure 629 (34 locs US) + AWS 1,329 (21 locs US). Continental: Azure 108 (7 locs DE, stale Mar 2025) + AWS 250 (7 locs DE). | thyssenkrupp: Azure 607 (19 locs US), AWS 322 (10 locs US) — medium intensity only. Evonik: Azure top install intensity just 118 (2 locs DE, stale Mar 2025), AWS 66 (1 loc DE). | HG technographic productId 781, 6426 |
| Kubernetes / container orchestration | Bosch: Kubernetes intensity 1,426 at 15 IN locations — fresh (Apr 2026). Siemens: Red Hat (OpenShift carrier) high intensity 2,455 at 99 locations. BASF: not explicitly surfaced but Azure+SAP stack implies container use. | thyssenkrupp: No Kubernetes signal in top-30. Evonik: Kubernetes intensity 63, 2 DE locations (stale Mar 2025). | HG technographic productId 14664, 819 |
| SAP + Cloud co-existence (ERP-cloud hybrid cost drivers) | BASF: SAP BW intensity 889 US + 524 DE (fresh). Continental: SAP BW 202 DE. Siemens: SAP BW 2,064 US (Nov 2025). All won accounts run heavy SAP alongside Azure/AWS, creating complex cost attribution. | thyssenkrupp: SAP Basis 197 DE (fresh Apr 2026). Evonik: SAP BW 243 US + SAP Basis 72 US. SAP present but thinner cloud co-existence footprint. | HG technographic productId 20297, 22000 |
Operating Signals Overlay
(HG operating signals, May 2026)
| Signal | Won Cohort | Contrast Cohort |
|---|---|---|
| AI trajectory | Siemens: "AI-leader-accelerating" (score 77.3, +3.2 Δ). Bosch: "AI-leader-accelerating" (score 56.8, +4.0 Δ) | thyssenkrupp: "AI-leader-plateau" (score 33.4, −0.4 Δ). Evonik: "AI-leader-plateau" (score 27.4, −1.4 Δ) |
| Cloud depth (GenAI) | Siemens: cloud-native (Azure 4,578 / AWS 4,126 / GCP 1,253). Bosch: cloud-native (Azure 2,865 / AWS 2,320 / GCP 725). BASF: cloud-native (Azure 1,287 / AWS 1,576). Continental: cloud-native (Azure 220 / AWS 472). | thyssenkrupp: cloud-native but lighter (Azure 1,121 / AWS 529). Evonik: intensity 82/100 — Azure 302 / AWS 87 only. |
| Cloud posture (mentions) | Siemens: "private-first" (hybrid + edge + VPC signals). Bosch: "hybrid." Continental/BASF: no explicit cloud-posture signal — large on-prem implied. | thyssenkrupp + Evonik: no cloud posture signal — suggests earlier-stage or predominantly on-prem orientation. |
| GenAI readiness | Siemens: "genai-ready" (intent 42,669). Bosch: "genai-ready" (intent 34,854). Continental/BASF: "genai-interested" (intent 37K–42K but lower data maturity). | thyssenkrupp: "genai-interested" (intent 25,166). Evonik: "genai-interested" (intent 30,128). Lower intent and lower AI product adoption velocity. |
Confidence Note
- Won cohort: n=4 accounts. Signal is rich for Siemens and Bosch (high-intensity technographic, fresh verifications, robust operating signals). Continental and BASF have patchier cloud technographic signals — Continental's Azure is stale (Mar 2025), suggesting the cloud install may be undercounted. Treat Continental's cloud intensity scores as directional rather than definitive.
- Contrast cohort: n=2 accounts — below the minimum threshold of 10 for high-confidence statistical claims. All deltas above are directional; with only two contrast accounts there is material risk that thyssenkrupp and Evonik are atypical of German industrial firms in their revenue band. A stronger contrast set would be built by querying
search_companiesfor NAICS 3312–3259, revenue $10B–$40B, EMEA, to pull 8–10 firms. - Spend data is HG-modeled, not invoiced. Cloud services figures reflect operating footprint estimates, not purchase orders.
- Technographic category note: The query returned broad infrastructure results — "Cloud cost optimization" is not a named HG product category; signals were extracted from cloud IaaS/PaaS/managed-cloud installs and container orchestration products as the nearest proxies.
Recommended ICP Refinement Actions
- Tighten the revenue filter to >$20B — Continental ($22B) is the floor; below that, the absolute cloud budget shrinks below the ROI threshold for most optimization tools.
- Add a multi-cloud requirement — Azure + AWS both present (verified within 18 months) as a mandatory qualifier. Single-cloud accounts rarely face the cost-attribution complexity that drives platform purchases.
- Use AI trajectory as a forward-looking filter — "AI-leader-accelerating" with a positive 6-month delta predicts expanding cloud workloads (and therefore expanding cost complexity). Plateau or declining scores suggest a stagnating cloud footprint.
- Expand the contrast set via
search_companies(NAICS 3312–3363, revenue $5B–$20B, EMEA, min employees 10K) to reach ≥10 contrast accounts before locking ICP thresholds.
Overview
Sharpen your ideal-customer definition from the cohort that actually buys. Marketing Ops feeds a list of closed-won accounts; the workflow runs them through HG firmographic plus technographic plus spend, finds the 5 firmographic features and 3 technographic patterns that distinguish them from a closed-lost or peer cohort, and returns an ICP definition every claim of which cites the signal it came from.
Use cases
An ICP definition you can defend in the QBR
Walk your last N closed-won accounts through HG firmographic, technographic, and spend, contrasted against your closed-lost cohort or a matched peer set. The output is a five-feature firmographic table and a three-pattern technographic table, every cell citing the HG signal it came from. The CRO challenge becomes 'why these features' and the answer is in the source column.
View workflow prompt
# ICP Refiner from Closed-Won Cohort
## Parameters
- `{{closed_won_domains}}` *(required)* — Comma-separated domains of accounts that closed-won in the relevant window. Example: `siemens.com,bosch.com,continental.com,basf.com`
- `{{contrast_cohort_domains}}` *(optional)* — Comma-separated domains for the contrast set. Example: `thyssenkrupp.com,evonik.com`
- `{{category}}` *(optional)* — The product category you sell, used to anchor technographic comparisons. Example: `Cloud cost optimization`
## Purpose
Sharpen an ICP from {{closed_won_domains}} against {{contrast_cohort_domains}} (or a peer cohort if empty). Anchor the comparison on {{category}} when provided. Every distinguishing feature cites the signal it came from. Reference skills carry the cohort, citation, and modeled-vs-observed discipline.
## Process
1. **Cohort gather.** Run `company_firmographic` across {{closed_won_domains}} and the contrast set. Capture industry, revenue band, employee band, geo, parent-subsidiary state.
2. **Tech and spend overlay.** `company_technographic` (intensity-sorted) and `company_spend` across both cohorts, scoped to {{category}} where present. `company_operating_signals` for cloud posture + AI maturity + network modernization.
3. **Find the 5 firmographic features that distinguish.** Bands where won cohort concentrates above contrast: revenue, employees, geo, industry, parent-vs-standalone. Cite the row counts.
4. **Find the 3 technographic patterns that distinguish.** Adoption-rate or intensity-band deltas in the categories adjacent to {{category}}. Cite `company_technographic` rows.
5. **Synthesize the ICP definition.** One paragraph plus a five-row distinguishing-features table plus a three-row distinguishing-patterns table. If contrast cohort is sparse, qualify the claims explicitly and propose a stronger contrast set built via `search_companies`.
## Output Format
- ICP paragraph (3 to 5 sentences)
- Firmographic features table (5 rows: feature, won-cohort distribution, contrast distribution, signal source)
- Technographic patterns table (3 rows: pattern, won-cohort adoption, contrast adoption, signal source)
- Confidence note: how many accounts in each cohort, where the signal is thin
## Quality Checklist
- Every feature cites HG firmographic / technographic / spend / operating-signals row
- Spend numbers are framed as modeled estimates, never as verified financials
- Sparse-cohort caveats appear when either set has fewer than 10 accounts
- No fabricated industry deltas; if HG returns sparse data, say so