Media Buying Guides

After September brands using the ad pixel are DEAD. 

And so are the tracking apps that thought another dashboard would save it…

The September 2026 updates to the Meta and Google pixels made the sale harder to see from a browser than it has ever been — and every sale the pixel misses is a customer the ad AI never learns from.

BUT : if you run attribution like most brands, you can pay as much as 30–50% more to get a customer from your ads, because you are optimizing against half your data.

Not fixing this after the September updates can kill your Q4 scale against competitors who already did.

The good news : the fix is mechanical, it works on the ads you already run, and it replaces a slice of your revenue with one fixed cost. 

Below I’ll break down how it works. Zero changes to your ads.

The large boost (or loss) comes from 3 changes in your ad conversion data.

1. Change #1 : Fix your conversion data with server-side tracking
2. Change #2: Use your pixel for TARGETING by feeding it MORE data: every sale it touched, plus the predicted value of each buyer.
3. Change #3: Stop renting a one-size-fits-all tracking app. Run attribution you own, over data you can query yourself.

Do these and Meta’s and Google’s ad AIs will
-Find you customers at a lower cost.
-Target higher-value customers,
- Optimize faster.
+Your ad team will waste less and scale faster too.

Put those together and you stop paying for the customers the pixel never saw.

First: why listen to me?

I’m Mikee Shattuck. I’ve spent fifteen years buying media — at one point fifty million dollars a year across Google, Meta, YouTube and LinkedIn — and I built OpenAttrib because the tracking I could buy kept disagreeing with the bank account.

So I’ve been on both sides of this: the ads, and the numbers the ads get judged by.

Trust me when I say this.

Since the September 2026 updates, the ONE thing separating failing and winning ad campaigns: HOW they gather conversion data, and what they feed the platforms’ AI.

Here’s what changed, and what to do about it.

FIRST : You must use CORRECT data

Most brands don't even do this. That is why their ads stop scaling.

Since the Andromeda updates to Meta’s ad algorithm, AI controls all your targeting. So bad AI data = bad targeting.

The problem:

If you’re using the basic ad pixels for attribution, the data is
usually off 20–30%. 

The result is your team AND your AI working with bad data. This blows up cost out the gate and kills AI targeting.

In the AI era, if you’re serious about running ads you should be using some kind of server-side tracking. (I’ll show you the one I built near the end of this post.)

The boost in tracking accuracy instantly lifts an ad account’s ROAS.

Fixing this is the easiest boost to an ad account you’ll ever see.

Here’s the arithmetic. The pixel sees the clicks it can follow and the sales it can match, and nothing else. Every sale it misses is a customer its AI never learns from — which means the campaigns that looked unprofitable are often the ones quietly feeding it the worst data.

Correct that feed and the same budget lands differently, in days rather than quarters. No new creative. No new campaign to launch. Nothing about your ads changes.

THEN once your data is accurate apply the 2 changes below for another near instant bump in ad ROI.

Big Change #2: Stop using your ad platform’s pixel for “tracking”

Easy 10–15% boost in ad ROAS

This sounds a tad… crazy. We want to send the pixel the “wrong” data.

Your targeting is only as good as the data you feed your pixel.

And ad pixels are not accurate. They miss about 20–30% of short-term sales and 90% of long-term sales. Track with them and your media buyers fly blind.

But accuracy isn’t even the point.

The big change is this: more data = better targeting. The data doesn’t have to be accurate. It has to be MORE.

Two steps get you there. Step 1: the pixel gets every sale it touched, even the ones it shouldn’t get credit for. Step 2: it learns who your best customers are, from predictive data.

This data is insanely good for targeting. It is not for tracking or ad decisions.

Step 1: Give the pixel the sale… even when you shouldn’t 😂

This is the KEY to top tier AI targeting on Meta/Google.

More data on your buyers, callers and leads = better targeting.

So you give Meta and Google ANY sale they touched, even if they shouldn’t get credit for it.

That takes heavy server-side tracking: collect every click in the sales journey, find the Meta and Google clicks, and send the sale back to them.

Here's an example that makes or breaks AI ad targeting

Imagine a 15-click customer journey with one random Meta click at click 3. Meta will likely never get credit for that sale. (And that click shouldn’t drive your buying decisions either.)

Doesn’t matter. We want Meta to have that sale.

So your server-side tracking has two jobs. For YOUR buying decisions, it credits the LAST click. For META, it still reports the sale to Meta’s pixel, so Meta gets the buyer data.

The more data you send the ad platform, the better it finds customers.

The diagram below breaks this down. 

That’s step 1. Now the platform sees every sale it touched.

NOW Step 2 tells it which of those sales are worth chasing:

Step 2: Use predictive data to show the platform which buyers are worth MORE

Two 20 orders look identical to the pixel. One of them is a 500 customer. We NEED to optimize for the 500 customer, not the 20 one.

A 10 buyer who refunds is not the same as a buyer who spends 200 with you over a year.

But unless you tell the ad platform who is who, it optimizes for the lower-value buyer every time.

The platforms can’t see this on their own. Their tracking windows are too short to know who spends the most with you over time. So again, the platform optimizes for your lowest-value buyers.

Stop doing this.

By giving ad platforms predictive data we can get higher value bigger spending customers. Here’s how.

Your existing sales data already shows what your best buyers look like. Add their behavior (location, device type, what they do on your site). Now you can predict the lifetime value of each new customer.

From their first sale and behavior, you can predict a 500 lifetime spender vs. a 20 refunder. The same goes for calls and leads.

You then send the 500 signal into Google and Meta, and their AI optimizes for the 500 buyer.

These two steps make your targeting far more efficient on both platforms.

 Here it is in motion:

That’s your first 10–15%. OpenAttrib does both halves of it on every conversion — the missed sale, and the value you say the buyer is worth — because it collects the click on your own subdomain and takes the sale from your own Stripe or WooCommerce webhook. The product tour walks the pipeline. Now the second change, and the second 10–15%.

Big Change #3: Stop using one-size-fits-all server-side tracking

Another 10–15% boost in ad ROAS

Bare minimum: use server-side tracking to track your ads and feed your pixels.

Ad platform tracking wildly over- and under-reports. It misses 20–30% of the sales it drives short term and 90%+ long term. It also claims sales it didn’t drive.

If your team cuts and scales on this data in 2026, they’re working from bad data. That kills returns by default.

But this still doesn’t cut it

I built a tracking company, and I don’t want you living in its dashboard.

This will sound strange from the person selling you attribution.

But here it is: the dashboard is the least important part of any tracking tool, and a generic one is a liability. Your funnel is not my funnel, and your best customer is not your competitor’s.

In the age of AI coding, renting a fixed dashboard and a fixed attribution model is just… expensive.

What I actually use instead

What I use (and what you should use) is attribution pointed at your own questions.

Let me explain.

The tracking has to be server-side and first-party, or the data is wrong before you start. OpenAttrib records the click on your own subdomain, records the sale from your own webhook, and pushes the conversion back to Meta, Google and TikTok from your server with the value you say it is worth.

Then the data is yours to interrogate, because it lands in a Postgres you control — ours hosted, yours on your own box.

This gives me:

• 100% custom reporting, metrics and dashboards
• Your own attribution models — first click, last click, linear, position, or a seven-day time decay, with your own SQL on top
• The ability to add any feature I need, and keep it
• Access to my data from any angle I want
• The ability to automate and control my ads from LLMs with this data
• The ability to change exactly what goes into each ad platform, per campaign

Every business has its own metrics and its own customer journeys.

So every business should have its own custom attribution suite. Period. In the age of AI code, there is simply no reason to run a one-size-fits-all dashboard or attribution system.

The payoff: you cut and scale on numbers built for YOUR funnel, not a generic one. That’s the second 10–15%.

We’ve done 95% of the work

I know that sounds like a lot of work. It isn’t, because OpenAttrib ships as a working app: the tracker, the pipeline, the five attribution models, the conversion APIs and the reporting, before you write a line.

On a self-hosted instance you get the full source, so “add a feature” is a prompt rather than a support ticket — and nobody can take it away from you.

Start with ours and add the specific things you need.

The Easy Way To  Do All Of This

Look, I built OpenAttrib, so of course I’d love you to use it: the server-side tracking, the pixel optimization, the fixed cost, and your data in a database you control.

But make no mistake. Whether you use us or not, in 2026 you must:

1. Use server-side tracking, or your cutting and scaling is off.
2. Feed your pixel ALL the sales it touched AND predictive data.
3. Run a custom attribution system, not a one-size-fits-all one.

That is the whole difference between scaling in 2026 and stalling — not a new creative, not a new offer, but the platform finally seeing what actually happened.

By all means, you can jerry-rig other solutions to get there.

But if you’d like a shortcut: OpenAttrib does all of it from one server-side tracker, with a fixed cost rather than a cut of the revenue it measures. Book Demo and we will walk it on your own accounts.

That’s the whole commercial argument. The tools that bill you on tracked revenue get more expensive exactly when your ads start working. This one doesn’t.

Book Demo (tell us the setup — nothing is charged and no card details are asked for)

→ See the whole pipeline: the click, the sale and the five models, end to end.

→ See it for your kind of shop: ecommerce, SaaS, call tracking and agencies, each on its own page.

That’s all!
Mikee