Only 5.32% of Competitor Events Are High-Significance: What 504,500 Signals Show
The noise postestablished that most price changes are automated background activity, not real decisions. That finding only covered one event type. Beaconmon's AI significance classifier, a Growth-tier feature, applies the same idea to everything the platform tracks: stock changes, product launches and removals, site changes, and promos, not just price.
We pulled the classification for 504,500 competitor events from production, everything tagged low, medium, or high by the classifier. Only 5.32% get flagged high-significance. The other 94.68% looks exactly like what the noise post described for pricing, just extended across the whole event surface.
Key Takeaways
- →504,500 competitor events analyzed across every type Beaconmon tracks: price, stock, product, site, and promo.
- →94.68% of events (477,650) are classified medium, background activity that does not need a real-time response.
- →Only 5.32% of events (26,820) are classified high, the ones actually worth a closer look.
- →The low tier is barely used, 30 events out of 504,500. In current data, significance functions as a two-tier signal: worth attention versus everything else.
Finding 1: 1 in 19 events is worth your attention
Across 504,500 competitor events, the classifier split activity into three tiers:
| Significance tier | Event count | Share of all events |
|---|---|---|
| Medium (background) | 477,650 | 94.68% |
| High (worth attention) | 26,820 | 5.32% |
| Low | 30 | 0.006% |
Only 5.32% of the events in our dataset are classified high-significance. Framed as a ratio, that is about 1 in 19 events. If your monitoring setup surfaces every competitor event with the same urgency, you are asking a merchant to sift through 19 events to find the one that matters.
This is a different signal than the one in the price-change noise post, which filters by the size of a price move. Significance scoring classifies across every event type, including ones that do not have a natural magnitude to filter on, like a product being added or a competitor going out of stock.
5.32% of all competitor activity is classified high-significance. An alert channel that treats every event as equally urgent is showing a merchant 19x more volume than the signal actually requires.
Finding 2: this holds across event types, not just price
The 504,500 events in this dataset are not all price changes. Beaconmon tracks nine distinct event types, and the significance classifier runs across all of them:
| Event type | Count | Share of all events |
|---|---|---|
| price_changed | 472,712 | 93.70% |
| product_added | 10,827 | 2.15% |
| out_of_stock | 5,213 | 1.03% |
| back_in_stock | 3,990 | 0.79% |
| product_removed | 3,464 | 0.69% |
| sale_started | 3,021 | 0.60% |
| site_changed | 2,666 | 0.53% |
| sale_ended | 1,651 | 0.33% |
| promo_detected | 956 | 0.19% |
Price changes dominate the raw volume, which is why the earlier price-data post could make its case on price events alone. But the significance layer is not a price-only feature. A competitor going out of stock, a new product appearing in a catalog, a homepage rewrite, a promo banner going live: all of it runs through the same classifier, and the same 5.32% high-significance rate holds when you look at the full 504,500-event set rather than price changes in isolation.
That matters because a merchant who only filters price noise is still exposed to full volume on everything else. Stock changes, product churn, and site changes need the same triage that price changes do, and the classifier applies it uniformly.
The 5.32% high-significance rate is not a price-change artifact. It holds across price, stock, product, site, and promo events alike, so filtering only on price magnitude still leaves the rest of the event surface unfiltered.
Finding 3: why 'low' is barely used in practice
The classifier has three tiers, but in this dataset one of them is almost empty. Only 30 of 504,500 events, 0.006%, are tagged low. Everything else lands in medium or high.
This is worth stating plainly rather than glossing over: it is not a claim that nothing in ecommerce competitor activity is genuinely low-value. It is an observed characteristic of how the current classifier resolves ambiguity. Given an event, it tends to decide between "background activity" and "worth a closer look" rather than actively reaching for a third, lower bucket. In practice, that makes significance scoring a two-tier signal today: high versus everything else.
For a merchant configuring alerts, this simplifies the decision. You are not choosing between three thresholds. You are deciding whether to see the 5.32% now, or see all of it later.
The low tier accounts for 0.006% of events in this dataset. Build your alert rules around the medium versus high split, not a three-way threshold the classifier is not currently using.
What to do with a 95%-noise event stream
Alert only on high-significance events
Configure real-time alerts to fire on high-significance events only. That is 5.32% of total volume, the events the classifier flagged as worth a closer look across price, stock, product, site, and promo activity. Everything else does not need to interrupt a workflow the moment it happens.
Route everything else to a digest, not a real-time channel
Medium-significance events are still worth seeing, just not the moment they happen. Route them to the Intelligence Digest instead of a live alert channel. A merchant gets a consolidated view of the 94.68% without the interrupt cost of treating it as urgent.
Treat significance as a triage layer, not a suppression layer
Nothing gets thrown away. Every event, high or medium, is still recorded and visible in history. Significance scoring changes what demands immediate attention, not what gets tracked. That distinction matters if you ever need to go back and reconstruct a competitor's activity over a longer window.
Significance scoring and Intelligence Digest are both Growth-plan features. If you are on Starter, the Digest still groups every change into a weekly summary, the AI classification layer that sorts high from medium is what Growth adds on top.
Frequently asked questions
What percentage of competitor activity is actually significant?
In a dataset of 504,500 competitor events, spanning price changes, stock changes, product launches and removals, site changes, and promos, only 5.32% (26,820 events) were classified high-significance by the AI classifier. The remaining 94.68% is background activity. About 1 in 19 events is worth a closer look.
Does AI significance scoring apply to more than just price changes?
Yes. The classifier runs across every event type Beaconmon tracks: price changes, stock changes, product additions and removals, site and homepage changes, and promo detections. The 5.32% high-significance rate holds across that whole surface, not just price.
Should I only get alerted on high-significance events?
That is the recommended setup. Configure real-time alerts for high-significance events only, and route medium-significance activity to a daily digest instead of a live channel. Nothing gets discarded, medium events stay visible in history, but your alert channel stops competing for attention with background noise.
What is the 'low' significance tier and why is it barely used?
Low exists in the classifier but shows up almost nowhere in practice, only 30 of 504,500 events (0.006%) are tagged low. In practice the classifier resolves almost everything into medium (background) or high (worth attention) rather than actively downgrading events to low. Treat it as a two-tier signal for now: high versus everything else.
Haimanot built Beaconmon after watching Shopify merchants lose sales to competitors they never saw coming. He writes about competitive intelligence, ecommerce pricing strategy, and how merchants can turn competitor data into decisions that protect margin.