Custom Labels: Your Performance Max Super Weapon

How AI models are trained

Whether you’ve selected a traffic light, motorcycle, stairs, fire hydrant or bridge, these labels are attached to the images.

This forms the basics of training AI. A bunch of images have been given to the different AI models with labels of what they are. AI can then recognise other images that have the same characteristics. 

It’s so simple, but genius really—getting billions of people across the globe to help train AIs without even realising it 💁‍♀️ Oh, and thank you for your service 👏

Custom label 0 enters the chat

In your shopping feed, custom labels work the same way. The different labels you attach to products, give Google’s AI the inside scoop on what’s what. It helps the AI pick up on patterns between the products. That could be learning about the audience that clicks on or buys similar products, or the average order value or return on ad spend for a set of products.

The more Google’s AI understands the relationships between your products and their performance, the better it’ll be at hitting your campaign goals.

The murky waters of “no label” territory

But if AI’s so smart, why can’t I just let it figure those relationships out on its own?

You’re kinda right, it could absolutely figure out those relationships over time, but that’s like sending your AI on a blindfolded shopping spree. And you don’t need three guesses as to what that leaves you with.

Just imagine the amount of wasted ad spend as the AI tries to figure things out 💸 Your low-performing products will get way more airtime than they should, while your top products are left gathering dust on the virtual shelf.

You’ll be desperately pleading with your campaigns—cue Sabrina Carpenter’s ‘Please, Please, Please’—not to embarrass you when reporting day comes. The whiplash from bad days to good days back to bad days again will have you rethinking all your life choices 🫠

Get custom labels working for you

Just like training AI on images, use custom labels to train Google’s AI on your products. Custom labels are Google AI’s cheat sheet, helping it make smarter decisions about your campaigns.

You’ve got five different slots you can use for each product. You don’t have to use them all, but the more info you can give Google to help it do its job, the better.

Custom labels can also give you more control to split out different campaigns. Whether it’s controlling spend to push more to your high-profit products, or making sure new products are getting visibility, when products are labelled well you can get more creative with your account.

You can already utilise other attributes for separating campaigns. And, Google will be able to pick up on other attributes for connections—think product type, brand, colour, size, gender, age etc. You want to use custom labels to build connections across these divides.

Label ideas to get you started

Not sure where to begin? Hopefully these spark something for you…

• Performance: bestsellers, high LTV, high margin, high AOV
• Category: back to school, gifting, room, style
• Product: hero, upsell, new in, add-on, collaboration
• Seasonality: summer, christmas, mother’s day, all-year, AW25
• Image style: product, lifestyle, model
• Audience: parents, students, business, golfers, home-owners
• Repeat rate: single purchase, subscription, monthly

The possibilities are endless. The key is choosing labels that align with your products and give Google’s AI the clues it needs to work its magic.

Custom labels are just the beginning

AI is changing the game. The days of endless bid adjustments have gone. Strategic thinking, creative testing, and utilising business insights are the future.

To stay ahead in paid search and shopping, we’ve got to team up with AI and use its power to our advantage.

So start getting familiar with your new co-worker, and as always, happy testing! ✌️