How to Think About Appice
Most teams come to Appice with a campaign mindset. This guide explains the shift required to operate a real-time decisioning system effectively — and why the difference matters.
Not campaigns — decisions
A campaign is something you plan, schedule, and send to a segment of customers. A decision is something Appice makes the moment a signal arrives — for one customer, right now, based on everything it knows about them.
Decide in advance what to send, to whom, and when. Set it up, approve it, fire it. Wait for results. React to the data in the next cycle.
Define what matters. Connect signals. The system decides in real time — the right action for the right customer at the exact right moment, without a send schedule.
In practice: Instead of "create a churn campaign for customers who haven't transacted in 30 days", you configure: "when a high-value customer shows three consecutive days of reduced activity, evaluate their churn score — if above threshold, trigger retention action immediately."
The difference: the first fires on a schedule. The second fires the moment the signal crosses the threshold — which could be at any time, for any customer, 24/7.
Not journeys — real-time responses
A journey is a pre-planned sequence of steps a customer walks through over time. A real-time response is what Appice delivers based on what a customer actually does — not what you predicted they would do.
Map out a multi-step sequence. Day 1: welcome. Day 3: feature prompt. Day 7: re-engagement. The customer moves through your predefined flow.
The customer's behaviour defines the interaction. Appice listens for what actually happens and responds to each signal — dynamically, without a fixed sequence.
In practice: A new customer logs in, skips onboarding, and heads straight to a loan product page. Appice reads those signals — login + skip + product browse — and surfaces a personalised loan offer in session, in real time. No day-7 re-engagement journey needed. The signal already fired.
Appice does support multi-step decision flows (and you can configure these in the Decisioning Engine). But even multi-step flows are signal-driven — each next step waits for the customer to do something, not for a timer to expire.
Not batch — continuous
Batch processing runs on schedules — hourly, nightly, weekly. By the time the batch completes, the moment has already passed. Appice operates continuously: every signal, every customer, every moment.
Run overnight jobs. Refresh segments. Sync data warehouses. By morning, you have a picture of what happened — and you act on it during business hours.
Signals stream into Appice in real time. The decisioning loop runs without interruption. Actions fire in milliseconds. At 2 AM. At 2 PM. For one customer or a million simultaneously.
In practice: A fraud signal detected at 2:47 AM triggers an alert within 80ms. There is no overnight job. There is no "we'll catch it in the morning" scenario. The loop does not sleep.
Not features — a system
The final shift: Appice is not a collection of independent features (segmentation, campaigns, push, analytics). It is a single decisioning system where each component feeds the next.
When you configure Appice, think about the loop, not the feature. "What signal do I want to respond to? What decision should it trigger? What action should that produce? How will I measure the outcome?" That is the question to ask for everything you build.
The practical test
Before you build anything in Appice, ask three questions:
- What signal am I responding to? If the answer is "a schedule" — rethink. If the answer is "a customer behaviour or system event" — proceed.
- What decision should this signal trigger? Define the logic: who gets which action, under what conditions, with what constraints.
- How will I know if the decision was correct? Define the outcome you are measuring. This closes the loop.
Where to go next
- How Appice Works — the full system model (Sense → Decide → Act → Learn)
- Signal Ingestion — connect your signals
- Decisioning Engine — configure your first decision