Your Data Already Knows. You Just Can't Hear It.
Somewhere in your sales history is a signal that 8 items are quietly losing margin every week. That a loyal customer hasn't been back in 40 days. That you have ฿200,000 sitting in stock nobody is buying.
You might have a dashboard. Maybe even a good one. But a dashboard waits for you to ask the right question. And you can't ask about a problem you don't know exists. By the time you notice a dropping line on a chart, cross-reference it with another report, and figure out the cause — a month has passed, and the money is already gone.
The Insight Engine doesn't wait for questions. It checks 16 aspects of your business on every cycle — inventory health, pricing patterns, customer behavior, operational trends — and when it finds something, it doesn't just flag a number. It tells you: here's what's happening, here are the specific items affected, here's how much it's likely costing you, and here's what you can do about it right now.
What It Actually Watches
Not vague "AI insights". Specific, auditable business checks — each one built to catch a real problem that costs real money.
Every business has blind spots. The Insight Engine is designed to illuminate them across four areas that drive profitability:
Inventory health
Which items are sitting unsold? Where is capital frozen? What's about to run out? Which new products aren't picking up? Six checks that catch stock problems weeks before they show up on a P&L.
Pricing & margin
Are your margins shrinking without you noticing? Do some items only sell with a discount? Is your profit concentrated in just a few SKUs? Four checks that guard the gap between revenue and cost.
Customer signals
Which regulars are coming less often? Which month's new customers are actually sticking? What do people naturally buy together? Three checks that help you keep good customers and sell more to them.
Operations
Which items have wildly unpredictable demand? Did revenue suddenly jump or drop? Are some products just taking up shelf space? Three checks that spot the things you'd only find by staring at spreadsheets for hours.
Each finding comes with the specific items involved, the numbers behind it, an estimate of the financial impact, and a concrete suggestion. Not "consider reviewing your inventory" — but "These 12 items in Edibles haven't sold a unit in 47 days. That's roughly ฿85,000 in tied-up capital. Consider a 20% markdown or returning to supplier."
"What If I Actually Do That?"
Every recommendation has a "What If" button. Click it, and the system runs a simulation on your real data before you spend a single baht.
The system says: lower the price on these items. Your first thought is: but what happens to my margin? That's exactly the right question, and that's exactly what What-If scenarios answer. They use your actual sales velocity, current stock levels, and estimated price sensitivity to project the outcome — not a generic formula, but a calculation tailored to how your business actually behaves.
Price & cost
What happens if I cut the price by 15%? Will the volume increase cover the margin drop? What if I renegotiate supplier cost instead?
Stock & ordering
What if I reorder this item earlier? How much will I save in stockout costs? What if I shift stock from one category to another?
Customer campaigns
What if I run a win-back offer for lapsed customers? How many are likely to return? What's the expected revenue from a retention campaign?
Merchandising
What if I bundle these two products? What if I run a promo less frequently — will demand hold? What if I give more visibility to this item?
Before you even see the simulation, the system checks whether the action is practical. Not enough stock for a promotion? It tells you. The recommendation conflicts with another insight for the same items? It flags that too. You see the projected upside and the potential problems in one place, before you decide.
The Part No Other Tool Does
Most analytics stops at "here's a chart". Ours goes further: it checks whether its own advice was right.
Here's why the waiting time matters. A price cut shows results in about two weeks — customers react fast. A campaign to bring back lapsed customers needs 45 days — people don't come back overnight. A change to your product mix takes two months to really show. The system knows this and uses different observation windows for different types of actions, so it doesn't judge a slow-burn strategy by sprint metrics.
When the observation window closes, the system compares what actually happened to what it predicted. Did profit go in the right direction? By how much? And crucially: was there something else going on during that time — a big sale, a supply disruption, a holiday — that could have skewed the result? If so, the system marks that measurement as unreliable and doesn't use it for learning. This is what keeps the system honest over time.
Trust Is Earned, Not Assumed
Every recommendation carries a confidence label — and the system has to prove its track record before that label goes up.
| Level | Status | What it means for you |
|---|---|---|
| R0 | New signal | We found something, but it's a first observation. Take it as a heads-up, not a directive. |
| R1 | Pattern detected | This check has spotted a consistent pattern. Useful for early warnings. No simulation available yet. |
| R2 | Actionable | You can run What-If scenarios on this. The math works, but the system hasn't been tested against your real outcomes yet. |
| R3 | Proven | This type of recommendation has been right at least 70% of the time over the last 3 months. You can rely on it. |
| R4 | Advisor | Highest confidence. At least 75% accuracy, consistently low error rate, confirmed over 3+ months. As reliable as a seasoned analyst. |
Getting promoted is hard — a recommendation type has to prove itself across three consecutive periods before the system upgrades its confidence. Getting demoted is instant — one bad period and the confidence drops. This is intentional. We'd rather show you a cautious recommendation that turns out to be right than a bold one that turns out to be wrong.
What Changes for Your Business
This isn't a tool you install and check once. It's a system that gets more valuable the longer you use it.
Visibility you didn't have before
- See dead stock, shrinking margins, and customer churn signals you weren't tracking
- Get specific, actionable suggestions — not generic advice like "optimize your inventory"
- Know exactly which items to reprice, reorder, or promote — with financial estimates
- Spend less time in spreadsheets and more time making decisions
From reacting to preventing
- Catch problems before they become expensive — not after the quarterly review
- Start seeing patterns: which types of actions work for your business and which don't
- Measure what actually happened after each decision, not just what you hoped would happen
- Your team develops a habit of data-informed decisions instead of gut feeling
A business that learns from itself
- Recommendations become more accurate because they're built on your proven history
- Each cycle teaches the system something — the value compounds like interest
- New employees can make informed decisions from day one, guided by the system's accumulated knowledge
- You stop losing money to problems you didn't see, because the system sees them for you
Most analytics tools give you a snapshot of yesterday. The Insight Engine gives you a direction for tomorrow — and gets better at pointing the way each time you follow it. This isn't about replacing your judgment. It's about making sure your judgment is informed by everything your data already knows.
You're Always in Control
The system suggests, your team decides. Having a holiday sale and don't want margin warnings flooding the screen? Turn off that check with one click and turn it back on after. Every setting — how sensitive the checks are, how long to wait before measuring results, which checks are active — is configurable per company, and changes take effect immediately.
There are no hidden algorithms. Every recommendation explains why it was generated, what data it's based on, and what assumptions it made. If the system says "reprice this item", you can see the exact sales numbers, margin trend, and demand estimate behind that suggestion. The goal is not "trust the machine" — the goal is "the machine earns your trust, one correct prediction at a time."