The old way: Scattered puzzle pieces

Until recently, Meta's ad system worked a bit like this:

Imagine your smartphone knows that:

• You clicked on some sports shoe ads last month
• You visited a few fitness pages recently
• You liked posts about running

The system would try to connect these dots to guess what ads might interest you. It worked okay, but had some big problems:

1. It couldn't tell if you looked at running shoes before or after watching marathon videos
2. It missed connections between things you did close together in time
3. It needed humans to decide which patterns were worth looking for

‍

The new way: Understanding your digital journey

Meta’s new system actually tracks the sequence of stuff you do online, keeping track of what happened and when it happened..

Think of it like this:

• Knowing someone visited a travel website, a hotel site, and looked at luggage (just random facts)

vs

• Knowing they searched flights to Spain, then immediately checked hotels in Barcelona, then looked at carry-on luggage (now that tells a story!)

One is just data. The other actually means something.

How it works: Similar technology as ChatGPT

Meta's new approach uses similar technology to what powers ChatGPT:

• ChatGPT learns patterns from words to predict what comes next in a conversation
• Meta's system learns from user activities to predict what ads you might like next

Both use smart AI systems that are good at finding patterns in ordered lists. One looks at text, while the other looks at user behavior.

For example, just as ChatGPT can understand that if you ask about "Italian recipes" and then mention "pasta," you're likely continuing the same conversation, Meta's system can recognize that if you browse running shoes and then look at fitness watches, these actions are connected in a meaningful way that shows your current interests.

Why good data matters

For these systems to work well, good data is really important:

1. Catching everything: If the system misses that someone clicked on an ad before buying something, the AI won't see the connection. Server-side tracking can catch these events even when browser tracking fails.
2. Getting the timing right: It matters if someone bought running shoes right after seeing an ad or three weeks later. Knowing when things happen helps understand why they happen.
3. Adding details: Basic "someone clicked something" data isn't enough. Adding information about their device, screen size, where they came from, and what else they did makes the data much more useful.
4. Keeping data clean: Even the smartest AI gets confused by messy data. Cleaning up incoming data, removing duplicates, and making sure everything is in the same format helps a lot.
5. Connecting devices: Knowing that it was the same person looking at shoes on their phone and then buying them on their laptop helps see the full picture of what that person wants.
6. Quality over quantity: A small amount of good data beats millions of bad data points. This is true for all data - from first seeing an ad (upper funnel) to making a purchase (lower funnel). Most companies only focus on getting good data about purchases, but the entire customer journey needs quality data at every step. Remember: bad data in = bad results out.

The results: Better ads for everyone

When Meta switched to this new approach, they saw good results:

• People saw ads that matched what they were actually interested in
• Higher value for advertisers and 2-4% more conversions on select segments
• Their systems ran more smoothly with less computing power

What this means for your business

If big companies like Meta are putting money into sequence learning, that tells us where marketing is heading. Here's what it means for businesses:

• Good data matters more than lots of data
• Understanding the flow of customer actions reveals what they really want
• Even the best AI can't work well with poor data

Companies that focus on data quality have seen real benefits:

• Lower costs to get new customers by understanding which steps lead to purchases
• More free users upgrading to paid by seeing which features make people want to pay
• More people finishing checkouts by fixing the right parts of the buying process

Having good data isn't just a technical issue—it directly affects your business results.

TrackBee: The solution for sequence-ready data

TrackBee helps your business get the same kind of data that Meta now uses:

1. We collect everything: We catch all customer actions, even when browsers block normal tracking.
2. We add important details: We create complete customer profiles that show patterns others miss.
3. We remove duplicates: We clean up messy data so you see what really happened.
4. We connect across devices: We can tell when the same person uses different devices or browsers.

These features give you the good data you need for systems like Meta's new approach.

Try TrackBee for free today!

Become a TrackBee partner

‍