Factory Gives Droids Analytics That Track Every Model and User
Factory rolled out a redesigned Analytics dashboard that breaks down agent spend by model and user, exposing where every credit of autonomous coding actually goes.
- Factory shipped redesigned Analytics exposing consumption by model and user across every session.
- New Consumption view tracks credits over time with per-model and per-user splits.
- Cost Concentration chart confirms whether power users and automations map to expected workloads.
- Factory Router claims 63% cost reduction versus frontier-model rates in production.
- Adoption view reports DAU, WAU, MAU, sessions, messages, tool calls, and stickiness.
- Available today on Business and Enterprise plans with an Analytics API for BI pipelines.
Factory adds model- and user-level analytics for coding agents
Factory has released a redesigned Analytics suite for Droids, its AI software-engineering agents. The suite attributes consumption to individual models and users across an organization’s sessions while reporting adoption, workload, and estimated routing savings.
- Availability: Live now for Business and Enterprise customers
- Access: Owners and Managers can open Analytics from the profile menu
- Coverage: Credits, model mix, users, sessions, messages, tool calls, and active-user metrics
- Integration: Dashboard access and an API for external reporting
A single coding-agent session can generate multiple model requests, tool calls, and retries, so an organization-wide monthly total reveals little about the cause of a spending spike. User and model attribution lets engineering and finance teams connect consumption to workloads, investigate anomalies, and set rollout policies with more precise data.
Credits, down to the user and model
Factory displays consumption in credits, with the Usage Over Time chart separating total usage from each model’s contribution. A Model Mix chart shows the share assigned to each model, and user-level reporting identifies who initiated the work. Translating credits into billed currency depends on each customer’s contract terms.
A Cost Concentration chart plots cumulative consumption across the organization. Factory says a small group of power users and automations generates most usage in many deployments, according to its release notes. The chart can expose an expected high-volume workflow or an automation consuming more credits than planned.
Inside the 63% router claim
Factory Router selects a model for each request automatically, allowing developers to submit work without choosing a price and performance tier every time. Analytics estimates what each organization would have spent at frontier-model rates and reports the savings attributed to routing.
Factory claims the Router reduces aggregate production cost by 63% compared with frontier-model rates while preserving frontier-level performance in its internal evaluations. The public release omits the evaluation suite, model distribution, traffic window, baseline providers, and pricing assumptions. Actual savings will depend on request mix, model prices, and routing policy.
Enterprise Controls complement the Router by giving administrators centralized model defaults and policies. Those controls can limit inconsistent model selection while the new reporting shows how the resulting policy affects consumption.
Adoption meets consumption
The Adoption view measures whether teams are using Droids regularly and how much work flows through them. Paired with Consumption, it gives budget owners several views of a rollout:
| Signal | Measures | Decision supported |
|---|---|---|
| Reach | Daily, weekly, and monthly active users | Whether access is translating into adoption |
| Workload | Sessions, messages, and tool calls | How heavily teams use the agents |
| Retention | Stickiness | Whether users return after initial trials |
| Cost allocation | Credits by user and model | Which people and model choices drive consumption |
Productivity analysis requires separate outcome measures such as accepted changes, pull-request cycle time, rework, and production incidents. Factory’s metrics establish adoption and cost allocation, providing the operational data needed to compare those outcomes with agent usage.
The Analytics API exposes user- and model-level usage data for internal BI and finance systems. Each dashboard card also includes a tooltip explaining how Factory calculates the metric, which helps teams reconcile reports built from different measures.
The enterprise pressure behind the release
Factory was founded in 2023 by Matan Grinberg and Eno Reyes, according to a company profile. The company raised a $150 million Series C led by Khosla Ventures at a reported $1.5 billion valuation, and the profile lists daily Droid use at NVIDIA, Adobe, EY, Palo Alto Networks, and Adyen.
Agent workloads tie costs to requests, model selection, tool use, and retries, making seat counts a weak proxy for spending. Model- and user-level attribution gives AI coding programs the cost-allocation layer familiar from cloud FinOps, while adoption data shows whether that spending reaches the intended teams. A complete return-on-investment assessment will still need contract pricing and software-delivery outcomes alongside Factory’s telemetry.
Comments
