Skip to main content
Domain Packs

The same engine —
tuned to your vertical.

Domain packs are the layer that tells the agent what matters in your industry. Swap them in and out without touching the engine — and create your own right from the dashboard.

How they compose

Three layers, one stack

Outputs

Findings your team can ship

Severity-ranked insights and prioritized recommendations that speak your industry's language — without any prompt engineering on your side.

Your domain knowledge

The swappable layer

Tells the agent which areas to investigate, which metrics to track, and which segments to compare — all expressed in the language of your industry, not code.

The shared engine

Discovery, validation, and reasoning

Shared across every pack. Exploration, self-healing, and verification — the part of the platform you never have to think about.

Inside a pack

What a pack actually describes

Analysis areas

The questions worth asking in your industry. What the agent should investigate — retention, monetization, activation, engagement, or whatever shape your domain takes.

Key metrics

What 'good' looks like in your world. Conversion rates, time-to-value, repeat usage, LTV — whatever signals you measure success against.

Audience segments

Who the agent should slice the data by. New vs returning, paying vs free, high-intent vs drive-by — every industry has its own cuts.

Success criteria

How to judge whether a finding matters. Severity thresholds, baseline comparisons, and the signals that separate a headline from a footnote.

No prompts to write, no files to edit. Everything lives in the dashboard, in plain language.

Create your own

Build a pack right from the dashboard

Describe your industry in a few sentences, pick the analysis areas you care about, set the metrics and segments the agent should track, and save. The next discovery run uses it automatically — no code, no deployment, no waiting.

See the creation flow
app.decisionbox.io/settings/domain-packs
Domain pack editor

Try it in two minutes

Clone the repo, run docker compose up, and point it at your warehouse.