Methodology

Where every public number comes from.

Pricing, impact ranges, deploy windows, human-equivalent anchors. None are hand-waved. Each one is either an authored record or a computed row, here's the derivation in plain English.

1. Pricing ranges

Each role has a pricing unit (e.g. "ticket handled", "invoice processed", "lead qualified") and a unit cost range in EUR. Calculations are:

monthly_cost_low = unit_cost_low × monthly_volume
monthly_cost_high = unit_cost_high × monthly_volume
payback_months = launch_fee / max(monthly_cost_low, 1)

The unit cost range is authored per role against the role's operating economics — not pulled from a generic model. Ranges widen as the role crosses volume tiers.

2. Fully-loaded human equivalent

Every role is anchored to a real job title. The anchor isn't base salary, it's the fully-loaded annual cost of hiring that person, which we define as:

fully_loaded = base_salary × (1 + benefits_rate + tooling_rate +
               management_overhead_rate + training_rate)
+ onboarding_ramp_amortized

Typical load factor: 1.35–1.65× base salary. Range chosen per-role against mid-market European benchmarks. Published in each agent's Fully-loaded human equivalent field and used to compute savings ratios on the calculator.

3. Impact ranges

Each role publishes an impact range like "25-40% faster first-response time". Sources:

The role's outcome cluster (Improve Service, Cut Cost, Move Faster, Increase Revenue, Reduce Risk) sets the baseline profile.
Per-role overrides via agent_dashboard_impacts nudge the profile when the role's actual strength diverges from its cluster (e.g. AR Specialist is in "Move Faster" but overweights revenue lift via DSO reduction).
Ranges are published as projections until lighthouse deployments produce measured numbers. Every snippet tagged evidence_level = projection on /proof and on agent detail pages. Upgraded to modeled or measured as the data lands.

4. Deploy windows

Ranges like "14-28 days" come from the role's deployment complexity tier (low / medium / high), adjusted for:

Integration count required at launch
SOP maturity needed on the client side
Compliance sensitivity of the role's domain

Time-to-first-value is always earlier than the full launch, typically 2-4 weeks into the deploy window, when the first capability runs on real traffic.

5. Scenario calculator

Every agent page has a scenarios section with 4 authored scenarios per role, each anchored to an industry, company size, and monthly volume. For each scenario we compute:

agent_cost = unit_cost × scenario_monthly_volume
human_fte_needed = scenario_monthly_volume / typical_monthly_throughput
human_cost = human_fte_needed × fully_loaded_annual_cost / 12
savings_pct = (human_cost - agent_cost) / human_cost × 100

Computed rows live in agent_scenarios_computed. A nightly Vercel Cron regenerates these against the current pricing bands so outputs stay in sync if we update a unit cost or a salary anchor.

6. Version + freshness

Current methodology version: v1.0. Every pricing unit carries the methodology version under which it was authored, so historical pricing decisions remain auditable when we bump methodology.

Pricing changes write a row to pricing_history; computed scenarios and metric series regenerate nightly (03:00 UTC and 03:30 UTC via Vercel Cron).

Live catalog numbers

Pulled from Supabase on this page build:

26
hireable roles
107
capabilities
63
integration platforms
104
authored scenarios

Want the underlying pricing models, calibration tables, or a copy of a specific role's unit-cost derivation? Ask us.