Period: 2026-04-02 (cost tracking rollout) → 2026-05-05 Scope: projects created post-rollout with ≥50 candidates that ran through scoring (real, working projects, not test stubs) Adjustments applied:
- CrustData realtime cost reduced by 70% (recent provider-side fix)
- Imputed CrustData cost for LinkedIn-sourced projects (so the price reference is provider-agnostic)
Goal: give a defensible price reference for client conversations — what does a project of size X actually cost us.
| Variant | Total | Mean | Median | p75 | p90 |
|---|---|---|---|---|---|
| RAW (as recorded) | $20,566 | $226 | $64 | $167 | $323 |
| RT-discount (−70%) | $18,667 | $205 | $54 | $125 | $289 |
| + CrustData imputation (final) | $20,551 | $226 | $88 | $172 | $323 |
The discount and the imputation roughly cancel on the aggregate — but the per-project distribution gets fairer: small LinkedIn-sourced projects move up; heavy CrustData-sourced projects move down.
| Field | p25 | median | p75 | p90 | max |
|---|---|---|---|---|---|
| companies | 93 | 133 | 251 | 745 | 1,641 |
| candidates (sourced total) | 616 | 2,009 | 7,383 | 16,904 | 181,809 |
| scored candidates | 263 | 533 | 2,791 | 8,368 | 25,959 |
| intelligent AI agents | 0 | 0 | 1 | 3 | 6 |
Sourcing split: 31 via CrustData, 60 via LinkedIn. ~28% of real projects use intelligent AI agents.
| Provider | n | Spend (RT-adj) | CrustData share | Imputed addition |
|---|---|---|---|---|
| crustdata | 31 | $11,217 | 8.5% (already counted) | — |
| 60 | $7,450 | 0% | +$1,884 imputed |
Imputation rate (derived from crustdata-sourced projects, RT-discount applied):
- $0.111 per company (used as primary basis — sourcing is the dominant CrustData cost)
- $0.0115 per scored candidate (alternate basis)
Median LinkedIn project: $36 → $85 after imputation. Average added per LinkedIn project: $31.
| Scored candidates | n | median | mean | p75 | p90 |
|---|---|---|---|---|---|
| <500 | 44 | $27 | $52 | $100 | $106 |
| 500-2k | 21 | $82 | $105 | $148 | $255 |
| 2k-5k | 11 | $205 | $797 | $369 | $812 |
| 5k-15k | 10 | $215 | $315 | $323 | $579 |
| 15k-30k | 5 | $605 | $828 | $1,327 | $1,772 |
| Companies | n | median | mean | p75 | p90 |
|---|---|---|---|---|---|
| 0-100 | 43 | $28 | $115 | $104 | $206 |
| 101-500 | 33 | $52 | $130 | $111 | $284 |
| 501-1500 | 11 | $255 | $876 | $579 | $1,089 |
| 1500+ | 4 | $215 | $421 | $605 | $812 |
| Intelligent agents | n | median | mean | p75 | p90 |
|---|---|---|---|---|---|
| 0 | 52 | $88 | $92 | $106 | $206 |
| 1-2 | 25 | $102 | $188 | $289 | $579 |
| 3-4 | 13 | $23 | $834 | $1,089 | $1,772 |
| 5+ | 1 | $200 | — | — | — |
The 3-4 agents bucket has bimodal behaviour: a cluster of cheap-to-run experiments coexists with the heavy mapping projects (#1547 alone = $6,389 imputed). Median ≠ mean is the signal.
| Intelligent agents | n | median | mean | p75 |
|---|---|---|---|---|
| 0 | 52 | $0.045 | $0.17 | $0.25 |
| 1-2 | 25 | $0.041 | $0.07 | $0.07 |
| 3+ | 14 | $0.124 | $0.30 | $0.17 |
So intelligent agents change the unit economics ~3× when 3 or more are configured (each agent runs against every candidate).
| Profile | n in data | RT-adj median | Final (imputed) median | Closest project |
|---|---|---|---|---|
| ~1k companies + 2 AI agents + 10-20k scored | 1 | $453 | $579 | #1473 Miro CTO |
| 1.5k-3k companies, 0-1 agents (sourcing-heavy) | 4 | $52 | $215 | Cenotian / Rivan |
| ~500 companies, 0 agents | 3 | $115 | $172 | lemon markets, Molyon |
| <200 companies, 0 agents | 27 | $70 | $85 | majority of small projects |
| <200 companies, 1-2 agents, 1-10k scored | 4 | $144 | $164 | Treatwell, Whatnot AE |
estimated_cost ≈ $0.05 × scored_candidates (baseline AI scoring)
+ $0.11 × companies (CrustData sourcing, post -70% discount)
+ $0.10 × scored × intelligent_agents (only if 3+ agents)
| Project shape | Estimated client cost |
|---|---|
| Small: <100 companies, <500 scored, no agents | $30 — $100 |
| Mid, no AI: 100-500 companies, 1k-3k scored | $80 — $250 |
| Mid + AI: 100-300 companies, 1-2 agents, 3k-10k scored | $200 — $500 |
| Large + AI: 500-1500 companies, 2-3 agents, 10k-20k scored | $500 — $1,500 |
| Mapping/heavy: 1k+ companies, 3+ agents, 20k+ scored, RT-heavy | $1,500 — $6,500 |
- Per-event cost differs by sourcing provider. Scoring on CrustData-sourced projects averages $0.023/event; on LinkedIn-sourced projects, $0.008/event. The 3× gap suggests either model differences or different rescore patterns — worth a separate dive.
- The 3-4 agents bucket has wide variance. Median is misleading there; use mean ($834) as a planning anchor for ambitious projects.
- #1547 (CRO/CSO DACH Mapping) still drives most of the high-end tail — $6,389 imputed, 31% of total adjusted spend. Treat it as the realistic ceiling for "mapping-style" engagements.
- Imputation uses an aggregate per-company rate. A project that sources 1,500 companies but never enriches via RT is overestimated by this; one with heavy RT use is underestimated. For pricing purposes the bias direction is conservative on the typical project.
Source: project_cost_events (production DB), filtered to projects.created_at >= 2026-04-02 and projects with ≥50 candidates having score_value IS NOT NULL.