Automations that branch.
An automation is a flow: something happens, Noli waits, checks a condition, and takes a different path depending on the answer. Thirteen lifecycle triggers, six actions, and branching on what the customer actually did.
How a flow is built
Every flow is made of four kinds of step, dragged onto a canvas:
| Step | What it does |
|---|---|
| Trigger | Starts the flow. Exactly one per automation. |
| Wait | Pauses for hours or days before the next step. |
| Condition | Splits the flow in two - a yes path and a no path. |
| Action | Sends something, or gives something. |
The classic shape is: trigger → send email → wait two days → did they open it? → if not, send a Viber message. That one flow is already more than most loyalty apps let you build, and it takes about a minute.
The thirteen triggers
These are the moments Noli can already detect. You do not wire up events - the loyalty engine and your store connection supply them.
| Trigger | Fires when |
|---|---|
| Welcome | A customer joins the loyalty program. |
| Order placed | An order is paid. The workhorse trigger. |
| Order fulfilled | Shipment goes out - the natural moment to ask for a review. |
| Abandoned cart | A checkout was started and never finished. |
| Win-back | A customer has gone quiet for a period you set. |
| Almost at the next tier | Close enough to the next tier that one more order gets them there. |
| Reached a new tier | Promoted - congratulate them and show the new perks. |
| Expiring points | Fires before points expire, so they get spent instead of lost. |
| Points milestone | A balance threshold you choose is crossed. |
| Birthday | The saved birth date comes round. Can gift points automatically. |
| Ask for the birthday | No birth date on file - collect one, optionally for points. |
| Time to restock | Purchase predictions say a consumable is running out. |
| Review received | A review lands - thank them, and reward the good ones. |
Two of these collect data rather than spend it. Ask for the birthday fills in a missing birth date - usually in exchange for points - which is what makes the Birthday trigger useful later. Time to restock needs Purchase Predictions switched on.
What a step can do
| Action | Notes |
|---|---|
| Send email | The branded template for that trigger. Override any part; merge tags work everywhere. |
| Send Viber message | Plain text through your bot. Customers without Viber are skipped silently. |
| Send SMS | For the people you cannot reach any other way. Billed per segment. |
| Poll with buttons | A question with tappable answers. The flow pauses until they tap. |
| Award points | Adds bonus points, recorded as a proper bonus transaction. |
| Grant reward | Hands over a specific reward from your catalogue. |
Outbound messages cost credits. Email, Viber, SMS and polls each debit your balance at the normal outbound rate - SMS per segment, the rest per message. A flow with a wide trigger and no conditions can get expensive quickly, so narrow it before you switch it on.
Conditions, and asking instead of guessing
You can branch on what you already know about the customer:
- Did they open the email? The basis of every fallback flow.
- Are they on Viber? Route the reachable ones to the cheap channel.
- Which tier are they in? Treat VIPs differently.
- How many orders? A first-timer is not a regular.
- Points balance, churn risk, time since last purchase.
Or you can just ask. A poll with buttons sends a question with tappable answers, pauses the flow until the customer taps or the wait expires, and then a poll answer is condition routes them down the right path. It turns a broadcast into a conversation, and the answer is data you keep.
Before you switch one on
- Check what you are missing first. Ask the Merchant Copilot which lifecycle moments have no coverage - it will name the gaps and draft the flow to close them.
- Look at the run history. Every automation records what fired, for whom, and what each step did. A flow that silently skips everyone usually has a condition that no one satisfies.
- Start narrow. One trigger, one send, one condition. Widen it once you have seen a week of real runs.