> ## Documentation Index
> Fetch the complete documentation index at: https://frimty.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Insights
> Automatic anomaly detection — no rules to configure
## Overview
An **Insight** is the one automation you don't configure. You don't pick an event, write a condition, or guess a threshold. You choose *what kind of problem* to watch for, and Alius compares **every product against its own history**, staying quiet unless something is genuinely wrong.
That last part is the point. A trigger fires whenever its event happens. An Insight is designed to say nothing on most days.
Insights compare each product to **its own** past performance — not to a store average
or to your other products. A product selling 3 units a day that drops to zero matters
just as much as one selling 300, if it earns a meaningful share of your revenue.
Insights are a **Pro / Max** feature and live under the **Automations** section alongside Triggers, Reports, and Goals.
***
## What you can watch
A product that reliably sold quietly stops — and Alius confirms it isn't just
out of stock or unpublished before telling you.
A product starts selling far faster than usual, with **Restock** and
**Alert supplier** buttons so you can act before it sells out.
A product starts getting returned far more often than its own history — a
quality, sizing, or listing-accuracy problem surfacing early.
Each Insight watches **one** of these. To cover all three, create three Insights — which is exactly what the Pro allowance of 3 is sized for.
***
## Why Sales Drop is the one to start with
When a single product stops selling, your store-level revenue barely moves. Nothing looks wrong on any dashboard, so the loss goes unnoticed for weeks — and revenue lost to a dead listing is revenue you can't recover later.
The hard part isn't spotting a decline; it's spotting a decline that *means something*. Alius rules out the boring explanations before it alerts you:
* The product is **out of stock** → no alert (that's a supply problem you already know about)
* The product is **unpublished or archived** → no alert (you took it down on purpose)
* The product's stock data looks **stale** → no alert (Alius would rather stay quiet than guess)
So when a Sales Drop alert arrives, it means demand actually fell while the product was available to buy.
***
## Sensitivity
Sensitivity is the only setting. It controls how large a change must be before you hear about it.
| Sensitivity | Sales Drop fires at | Sales Surge fires at | Return Rate Spike fires at |
| ------------ | ------------------- | -------------------- | -------------------------- |
| **High** | a 50% decline | 2× the usual rate | 2× the usual rate |
| **Balanced** | a 60% decline | 2.5× the usual rate | 2.5× the usual rate |
| **Low** | a 71% decline | 3.5× the usual rate | 3.5× the usual rate |
**Balanced is recommended.** High surfaces more, including some ordinary fluctuation. Low only speaks up when something is unmistakable.
Everything else — the comparison windows, the minimum sales volume, the revenue threshold — is fixed deliberately. The promise of an Insight is that Alius finds the pattern, so there is nothing else for you to tune.
***
## How Alius keeps Insights quiet
Noise is what kills alerting tools. Four rules keep Insights worth reading:
At most **3** velocity alerts per store per day, ranked by revenue impact. The
products that matter most reach you first.
A product must account for at least **1% of your store's revenue** to qualify.
A \$3 accessory dropping from 20 units to 2 is dramatic and irrelevant.
If many products move at once — a sale ending, a season turning, an ad budget
cut — Alius says nothing. That's one story, not twenty problems.
An ongoing problem alerts **once**. You'll only hear about it again if it gets
materially worse, or if it resolves and comes back.
Insights are evaluated once per day at **04:30 UTC**.
***
## Where the sales history comes from
When you create a Sales Drop or Sales Surge insight, Alius imports roughly the
**last 5 weeks of your per-product sales history** from Shopify's own analytics,
so detection starts right away rather than after a month of waiting.
A product still needs enough selling history of its own to establish a reliable
baseline. Products you only recently started selling stay quiet until they do —
that's deliberate, since judging a product on a few days of data would flag
ordinary fluctuation as an anomaly.
Two details worth knowing:
* Imported history is measured **net of returns and discounts**, while data Alius
records from then on is measured gross. The difference is small, and it fades
out entirely within about five weeks as live data replaces the imported rows.
* Importing affects Sales Drop and Sales Surge only. **Return Rate Spike** builds
its baseline from live order and return data, so it needs a few weeks of your
store's activity before it can fire.
Use **Fire (test)** at any point to see exactly what an alert will look like.
***
## Action buttons on Sales Surge
A surge is only useful if you can act on it before the product sells out. On **Slack**, **Telegram**, and **Discord (Bot)**, Sales Surge alerts carry two buttons:
| Button | What it does |
| ------------------ | ---------------------------------------------------------------------------------------- |
| **Restock +N** | Adds stock to every inventory-tracked variant of the product, sized to the current surge |
| **Alert supplier** | Emails your supplier a prebuilt reorder request for that product |
Both are **on by default** and can be turned off individually. Enabling **Alert supplier** requires a supplier email address.
On hubs that don't support buttons (Email, Discord via incoming webhook, HTTP webhook), the alert still arrives in full with no broken controls.
**Restock** here works on the whole product, adding stock to each tracked variant.
The similar button on the [Predictive Stockout report](/docs/features/reports) targets one
specific variant instead — Insights work at the product level.
***
## Creating an Insight
1. Click **Create Automation** → choose the **Insights** tab.
2. Pick an anomaly type: **Sales Drop**, **Sales Surge**, or **Return Rate Spike**. There's no "Blank" option — the anomaly type *is* the template.
3. Set the **Title**.
4. Choose a **Sensitivity** (High / Balanced / Low — Balanced recommended).
5. For **Sales Surge**, review the **action buttons** and add a supplier email if Alert supplier is enabled.
6. Select a **Hub** and its channels/recipients.
7. Save.
There is no schedule to pick and no conditions to build — detection runs automatically once a day.
Use **Fire (test)** to preview a real alert. If no product has enough history yet, the test uses sample numbers and tells you so. A test fire never affects real alerts.
***
## Where Insights show up
Insights have no separate findings page. Everything Alius detects arrives as a notification on your hub and is recorded in [Notification History](/docs/features/notification-history) like every other automation.
***
## Limits
| Plan | Active Insights |
| ---- | ----------------- |
| Free | 0 (not available) |
| Pro | 3 |
| Max | Unlimited |
All three anomaly types draw from the same allowance, so a Pro store can run one of each. Only **active** Insights count toward the limit. See [Pricing](/docs/guides/pricing).