Terrabase runs decisions like pricing, ordering and next best action end to end: from the first signal to the measured result, with your people in charge.
A sales agent picks the right plan to pitch.
Megan is 3× over her data cap on a 31-month-old handset, and her contract ends in 3 weeks.
Plans, devices, usage, eligibility and commissions add up to a deal sheet no agent can hold in their head. Terrabase ranks it into three actions before the agent says hello, and re-ranks live when details change on the call.
Observed · a global telecom provider4 to 6× revenue per agentA store manager lines up stock, prep and crew.
Evening demand is expected to run 30% above a normal Saturday.
Sales history, ERP inventory, shelf life, local events, crew rosters: every store's tomorrow hides in a dozen systems. Terrabase resolves it into one plan per store: order, prepare, deploy crew, hour by hour. The manager approves in minutes, before the shift.
Observed · a Fortune 500 QSR30% less waste · 42% fewer stockoutsA pricing team picks the price that earns the most.
Their price went up 6% overnight. Ours is now clearly the cheaper choice on the shelf.
Competitor prices shift daily across every channel; availability, search rank and demand move with them. Terrabase reads all of it, recommends a governed price with the reason attached, and routes it for sign-off. Weeks of work lands in minutes.
Observed · a global consumer goods exporter+17% gross margin from a 4% moveEach of these runs for weeks, drawing on your systems of record, market signals, policies and what your operators know.
It is built for work that spans days to weeks, across systems and sign-offs. It keeps a decision graph that remembers every case, checks policy on every action and retrains models on your own outcomes.
The Queue is the list an operator opens. It holds the decisions waiting for approval today.
It works as a simple approval interface: approve a decision, change it, hold it or reject it, and add a note explaining why. Terrabase carries out what was approved, and the routine decisions keep running on their own.
The Book holds every rule the operation runs on. Each rule carries its version, its owner, where it came from and every decision it has touched.
When operators keep overriding the same call, Terrabase writes it up as a proposed rule change. A person reviews the proposal, and nothing changes until they approve it.
The Pulse shows every run as it happens: when it started, how long it took and how healthy the engine is.
It also tracks which engine version is live, when the models last retrained and whether every data feed is fresh. If a number drifts, the team sees it here before anyone has to ask.
A decision runtime is software for running the same business decisions again and again. For each decision, it gathers the evidence: the signals, the numbers, the forecast, your policies and guardrails, and who has the authority to act. It weighs the options, recommends an action, records what was actually done, follows the result and remembers what worked. When a store orders for a big weekend, the runtime remembers whether that order was right. Terrabase is built for recurring decisions like pricing, ordering and next best action.
A dashboard helps you understand the business. A decision runtime carries a decision from signal to action to outcome to learning. Terrabase compares the available actions and checks them against your policies and guardrails. It routes the decision to the right person or runs it inside the bounds you set, and records what was approved and executed. When the result lands, it is tied back to the decision that produced it, so you can look at any past decision and see exactly what happened after it.
Your team sets the authority for every decision: some run automatically inside approved bounds, some wait for review, and some actions are never allowed. An operator can approve a recommendation, change it, hold it or reject it. Terrabase records the action, the reason and the policy version that applied, so any decision can be reviewed later.
Yes. Every decision Terrabase recommends comes with a full replay: the signal that opened it, the options it weighed and the ones it refused, the rule that fired, the policies and guardrails that applied, and who approved it. You can ask any decision to explain itself and export the evidence for an audit.
Two ways. Models learn on their own: forecasts retrain on your outcomes as they come in, so predictions keep up with the market. Rules only change with your sign-off: a proposed change to a policy, a guardrail or a workflow is tested, versioned and approved before it touches a live decision. That is how a manager’s override this season becomes the default plan next season.
We start with one decision you already make on a schedule. With your operators and your data team, we map its inputs, write down its policies and agree on how success is measured. We test it against your past cases, then run it alongside your current process until you trust it with the live decision.
One you already make on a schedule.
We write down its inputs, its policies and what success means.
It runs beside your current process, and every outcome is measured.
You hand it the decision. You stay in charge.
In 30 minutes we map its data, its owner and its metric with you. Six weeks later it runs alongside your current process.