---
title: "ML Product Segmentation for Google Shopping | Product Metrics"
description: "Product Metrics ML segments your products every day: six segments, each written to Merchant Center as a custom label for Shopping and Performance Max."
canonical: "https://www.productmetrics.io/product-segmentation"
pageType: feature
language: en
publisher: "Product Metrics"
dateModified: 2026-09-30
image: "https://www.productmetrics.io/og/product-segmentation.png"
---

> Content index: https://www.productmetrics.io/llms.txt

# Give every product the right advertising priority. Six segments, refreshed daily.

Product Metrics ML segments your products and writes each segment to Merchant Center as a custom label, so Shopping and Performance Max spend where products earn.

1. **Read products.** Reading 12,400 products from Google Ads, Merchant Center and shop orders.
2. **Weigh signals.** Product Metrics ML weighs clicks and return for every product.
3. **Segment into six.** Segmented on volume and return. Volume is ad clicks, return is POAS against a 2 target.
4. **Pegasus 42 on ROAS.** On ROAS, Nike Pegasus 42 clears the 300% bar at 3.18 and looks like a Star.
5. **Pegasus 42 on POAS.** On POAS, its 44% margin tells the real story: 1.4 against a 2 target makes it a Question Mark.

Nike Pegasus 42: ROAS 3.18 vs 300% target → Star; POAS 1.4 vs 2 target → Question Mark. Same product. Margin tells the real story.

_Illustrative data from 12,400 example Nike products. Product Metrics is not affiliated with Nike, adidas or PUMA._

- 1M+ products analysed
- €750K+ monthly Shopping spend under management
- 6 segments, refreshed daily

## Two questions segment every product. How many clicks does it get, and what do they return against your ROAS or POAS target?

_Counts and examples are illustrative. Product Metrics is not affiliated with Nike, adidas or PUMA._

### Stars

806 products

High volume, high return. Proven winners that draw clicks and beat your target.

Nike example: Vomero 18, Vaporfly 4, Air Force 1 '07

- **Volume (ad clicks).** High
- **Return (POAS vs target).** Above target
- **Action.** Scale budget
- **Campaign advice.** Own PMax campaign or asset group with a lower target, so budget scales.
- **Label.** custom_label_0: stars

### Question Marks

1,254 products

High volume, medium return. Plenty of clicks, return just under target.

Nike example: Pegasus 42, Zoom Fly 6, Tech Windrunner

- **Volume (ad clicks).** High
- **Return (POAS vs target).** Just under target
- **Action.** Test price & feed
- **Campaign advice.** Keep at target; check price position and product journeys before scaling.
- **Label.** custom_label_0: question_marks

### Cash Cows

622 products

Low volume, high return. Efficient products that lack reach.

Nike example: Vomero Plus, Structure 26, Dunk Low

- **Volume (ad clicks).** Low
- **Return (POAS vs target).** Above target
- **Action.** Protect & grow reach
- **Campaign advice.** Buy reach: a looser target and a bigger budget share.
- **Label.** custom_label_0: cash_cows

### Dogs

1,543 products

High volume, low return. They soak up clicks without paying back.

Nike example: Pegasus Plus 2, Air Max 90, Club fleece crew

- **Volume (ad clicks).** High
- **Return (POAS vs target).** Below target
- **Action.** Lower priority / fix page
- **Campaign advice.** Lower its priority; fix price, title or feed before spending more.
- **Label.** custom_label_0: dogs

### Drainers

4,914 products

Low volume, no return. Spend with nothing to show for it.

Nike example: ACG Pegasus Trail, Air Max Plus

- **Volume (ad clicks).** Low
- **Return (POAS vs target).** No return
- **Action.** Exclude or pause
- **Campaign advice.** Isolate with a strict target, or exclude until something changes.
- **Label.** custom_label_0: drainers

### Ghosts

3,261 products

Inactive: no impressions, no clicks. Google is not showing them.

Nike example: Pegasus 42 GORE-TEX, Dri-FIT Club cap

- **Volume (ad clicks).** No impressions
- **Return (POAS vs target).** Nothing to measure
- **Action.** Reactivate or clean feed
- **Campaign advice.** Check Merchant Center status and feed quality; relaunch in a test campaign.
- **Label.** custom_label_0: ghosts

## Nine signals in. One segment updated, every day.

- **Google Ads.** clicks, CTR, rank
- **Shop orders.** revenue, orders, CVR
- **Margin.** cost of goods → POAS
- **Price & stock.** price, stock cover

1. **Collect signals.** Clicks, cost and orders come from Google Ads and your shop. Margin turns revenue into profit.
2. **Product Metrics ML.** Product Metrics ML places Pegasus 42 on volume and return against all your other products.
3. **Assign segment.** Lots of clicks, POAS under target: a Question Mark. Demand is there, profit is not yet.
4. **Write the label.** The segment is written to custom_label_0. Merchant Center merges it on the next fetch.
5. **Campaign picks it up.** Performance Max now holds Pegasus 42 in the Question Marks asset group, at your target.

Nike Pegasus 42 in this run, with margin data:

| Field | Value |
| --- | --- |
| Return | POAS 1.4, target 2 |
| Segment | Question Marks |
| custom_label_0 | question_marks |
| Campaign | PMax Question Marks |

_Illustrative data. Follow Pegasus 42 to its Performance Max asset group, then switch margin data off. Product Metrics is not affiliated with Nike, adidas or PUMA._

### Thresholds learned from your account

No industry rule of thumb. Your click distribution sets high volume; your ROAS or POAS target sets the return bar.

### Enough evidence before a Star

A product needs a minimum number of clicks before it can be a Star. One lucky sale does not move budget, and products don’t jump between segments week to week.

### Re-segmented every day

On a rolling 30, 60 or 90-day window, against all your other products. When clicks and returns change, the segment and label follow.

## The model decides, and you keep the final say. Product Score rates every product 0–100 on your weights; Product Rules act on it.

Product score: metric weights, score 0–100 per product:

| Metric | Weight |
| --- | --- |
| Price Competitiveness | 30 |
| Profit Margin | 25 |
| Inventory Level | 20 |
| Return Rate | 15 |
| Delivery Speed | 10 |

Score preview:

| Product | Score |
| --- | --- |
| Nike Air Force 1 '07 | 88 |
| Nike Vomero 18 | 83 |
| Nike Pegasus 42 | 58 |
| Nike Air Max 90 | 44 |

Rule #1 Promote high scorers: When a condition matches, Then set a value:

| Part | Option |
| --- | --- |
| When | product_score greater than 70 |
| When | product_type contains "Running" |
| When | stock less than 5 |
| Then | Product Segment set to stars |
| Then | Category Priority set to High |
| Then | Category Priority set to Low |

Writes custom_label_0 = stars in the supplemental feed

1,070 products match, 690 labels change

Live preview for rule #1 Promote high scorers:

| Product | custom_label_0 |
| --- | --- |
| Nike Air Force 1 '07 | stars → stars (unchanged) |
| Nike Vomero Plus | cash_cows → stars |
| Nike Dunk Low | cash_cows → stars |

_Illustrative data. Pick a When condition and a Then value. Product Metrics is not affiliated with Nike, adidas or PUMA._

## Every segment becomes a label in Merchant Center, updated daily. Your campaigns hold the right products, with your bid strategy per segment.

_Illustrative data. Example with Nike products; Product Metrics is not affiliated with Nike, adidas or PUMA._

### A campaign structure that follows the labels

Split by custom_label_0 once. Products then move between campaigns and asset groups as their segment changes. Your bid strategies stay yours.

| Segment | Campaign | Bid strategy |
| --- | --- | --- |
| Stars | Performance Max, own campaign, asset group per category | Target ROAS or POAS below goal to buy volume |
| Question Marks | Performance Max, own campaign | Target at goal; review price and journeys first |
| Cash Cows | Performance Max or Standard Shopping | Maximise conversion value with a looser target |
| Dogs | Standard Shopping | Target above goal; lower priority |
| Drainers | Standard Shopping, low priority | Strict target, or excluded |
| Ghosts | Not serving | Fix feed or stock, then relaunch |

### Labels move products

A new segment means a new label, and the product moves campaign on the next Merchant Center fetch.

Merchant Center fetch 06:30, merged

| Product | custom_label_0 | Campaign |
| --- | --- | --- |
| Nike Vomero 18 | stars | PMax Stars |
| Nike Zoom Fly 6 | question_marks → stars | Moved to PMax Stars |
| Nike Pegasus 42 | question_marks | PMax Question Marks |
| Nike Air Max Plus | drainers | Standard Shopping |

Zoom Fly 6 crossed its POAS target, 1.52 to 2.12, and its label changed to stars.

### See which products have room to grow, before you raise a budget

Product Potential forecasts the extra clicks and revenue from a higher impression share, and the cost at today’s CPC.

Product Potential, last 30 days:

| Segment | Impr. share | Pot. clicks | Pot. revenue | Pot. cost |
| --- | --- | --- | --- | --- |
| Stars | 81.2% → 88.0% | +1,550 | +€15.2K | +€3.1K |
| Question Marks | 76.4% → 80.0% | +520 | +€2.0K | +€1.0K |
| Cash Cows | 58.3% → 70.0% | +1,020 | +€5.5K | +€1.8K |
| Total | 3 segments | +3,090 | +€22.7K | +€5.9K |

Start with the products that earn their next euro.

## Your first segmentation, in three steps. Log in, choose ROAS or POAS, add one supplemental feed.

Built for 500 to 50,000 products on Performance Max and standard Shopping, with Shopify, WooCommerce, Magento, Lightspeed and Shopware 6.

### Connect Google Ads and Merchant Center

Log in with Google and connect your Google Ads account or MCC. We read your data and write only the labels you approve. No credit card.

### Choose ROAS or POAS

ROAS works right away, without margin data. Connect margins to segment on POAS.

### Add the supplemental feed once

Add the feed URL to Merchant Center once. Labels merge on every fetch; campaigns split on custom_label_0.

## What PPC teams ask first. Anything else? Ask Berend.

### How does Product Metrics decide which segment a product is in?

Machine learning places each product in one of six segments by volume and return. Volume is ad clicks, set against your own click distribution. Return is ROAS or POAS against your own target; the app shows which metric and target are in use.

### Will products jump between segments every week?

No. Products are re-segmented daily on a 30, 60 or 90-day window, but minimum-click thresholds mean a product only changes segment on enough evidence. One lucky sale does not move budget.

### Do I need margin data?

No. Segmentation works on ROAS right away. Connect margins to switch to POAS, which stops low-margin bestsellers posing as Stars: Pegasus 42 clears a 300% ROAS target at 3.18, but earns POAS 1.4 against a 2 target.

### What campaign structure do I build with the segments?

Split campaigns or asset groups on custom_label_0 once, for example as in the table above: Stars and Question Marks in their own Performance Max campaigns, Cash Cows with a looser target, Dogs and Drainers in Standard Shopping with strict targets. Products then move as their segment changes.

### What access do you need? Do you change my bids or budgets?

Connect your own Google account or MCC. Product Metrics reads your data and writes only the labels you approve, via a Merchant Center supplemental feed. Bids and budgets stay yours; labels decide where each product sits.

### Can I override what the model decides?

Yes. Product Rules run on top of the model: when a condition matches (brand, price, stock, product score, feed age, custom labels, even the weather), set the segment or priority. Product Score (0–100, your weights) can drive rules such as product_score > 70.

### Is Product Metrics an attribution tool?

No. Product Metrics is not an attribution tool. It uses your Google Ads and shop data to show which products perform and why, and turns that into segments and labels.

### What does it cost, and is there a trial?

A graduated share of your rolling 30-day Google Shopping spend, excl. VAT; the first €10K each month is free. The free analysis needs no card; the 30-day trial does, and cancelling during the trial costs nothing. Full Signal Tracking is included, not billed separately.

- [Book a demo](https://cal.com/berend-vrakking/demo)

## Related pages.

- [Break-even ROAS calculator](https://www.productmetrics.io/break-even-roas-calculator): Segments use ROAS or POAS. Find the return your products need to break even.
- [Inventory Insights](https://www.productmetrics.io/inventory-insights): Stock cover feeds Product Score and your Merchant Center custom labels.
- [Google Shopping ads](https://www.productmetrics.io/google-shopping-ads): How Google Shopping ads work, and how to manage and optimise them product by product.
- [POAS vs ROAS](https://www.productmetrics.io/blog/poas-vs-roas): Profit on ad spend: break-even is 1, and margins decide the rest.
- [How to improve ROAS by product](https://www.productmetrics.io/blog/how-to-improve-roas): Fix the products that drag the account down, not the account average.
- [What is Google Merchant Center?](https://www.productmetrics.io/blog/what-is-google-merchant-center): Custom labels, data sources and price benchmarks, from an advertiser’s side.

## See your own products in six segments, today. Log in with Google and connect Google Ads and Merchant Center. Your first €10K of monthly Shopping spend is free.

- [Analyse your products for free](https://app.productmetrics.io/): Free analysis, no credit card
- [Find your hidden profit](https://www.productmetrics.io/pricing#calculator): 30-second estimate · no login
- [Book a demo](https://cal.com/berend-vrakking/demo): a video call with Berend, then a demo account with example data
- [See pricing](https://www.productmetrics.io/pricing)

---

Published by Product Metrics (Berend Vrakking, founder).

HTML version: https://www.productmetrics.io/product-segmentation