Stop Optimizing on View-Through: Why We Make Every Meta Decision on Click Data

Picture a busy restaurant with a doorman out front. Every single person who walks in, he taps them on the shoulder on the way past. Then at the end of the night he tells the owner he personally brought in 80% of the diners. Technically true, he was standing near the door when they entered. But most of those people had already booked a table, already decided, already driving over before he ever existed.

That doorman is view-through attribution on Meta. And if you pay him a bonus based on his own count, you'll keep him long after he's stopped earning his keep.

This is the single biggest reason I see good accounts make bad budget calls. The reported numbers look healthy, the agency points at the ROAS, and underneath it all the money is quietly flowing to campaigns that are mostly taking credit for sales that were going to happen anyway.

So here's my take, plainly: we feed Meta as much data as we can, but we make every spending decision on the click data. Those are two different jobs, and confusing them is what trips people up. Let me pull them apart.

What view-through actually is, and why it flatters you

A view-through conversion is when someone sees your ad, does not click it, and then buys within a window (one day, usually). Meta counts that as a win for the ad.

Sometimes that's fair. Sometimes the ad genuinely planted the idea.

But here's the thing Meta is very, very good at. Slipping in at the last second. Someone is already at the checkout on your site, decision made, card half out. They flick over to Instagram for ten seconds while a page loads, scroll past your ad without stopping, flick back, and complete the purchase they were always going to complete. Meta logs a view-through conversion and takes the credit.

That's the doorman tapping a shoulder.

A click conversion is a different animal. For that to count, the person had to actually click. They tapped the ad, landed on your site, did something. It's not perfect proof, but it is at least real evidence the ad did something other than be nearby. They engaged. That's the line I care about.

So when I'm deciding where the next dollar goes, view-through credit is the noisiest signal in the account, and click credit is the cleanest one I've got. I weight my decisions accordingly.

The two jobs attribution is doing

Here's the bit that confuses people, and it's worth slowing down for, because the nuance is the whole point.

Your attribution setting is doing two completely separate jobs at once.

Job one: teaching the algorithm. Meta uses conversion data to find more buyers. The more signal you give it, the better it optimises. This is why we run the 7-day-click, 1-day-view setting as our optimisation window rather than a strict 1-day-click. Letting it learn from both the clickers and the viewers gives it a richer picture of who converts. In my experience a tight click-only window also gets jumpy day to day. One day it's a 5x, next day it's a 0.4, then a 2, then back down. The wider window smooths that out and the algorithm makes steadier decisions. More data in, more stable delivery out.

Job two: judging the results. This is you, sitting down to decide which campaign gets more budget and which gets cut. A totally different question. Here you are not feeding a machine, you are grading one.

And my rule is simple. Teach it on views and clicks. Judge it on clicks only.

Feeding Meta the view data helps it work. Judging Meta on the view data fools you. Same number, two opposite uses. Most people pick one setting and use that single figure for both jobs, and that's the mistake.

A worked example of the same account, two ways

Let me make this concrete with some round numbers. Made up, but the shape is exactly what I see.

Say you've got four prospecting campaigns and you're looking at total reported ROAS:

• Campaign A: 4.1x
• Campaign B: 3.4x
• Campaign C: 2.6x
• Campaign D: 2.2x

The obvious move, the one most people make on a Monday morning, is to pour more into A, hold B, and start squeezing C and D. They're the laggards, right?

Now I add one column. I split out the 7-day-click ROAS, stripping the view-through credit back out so I can see what each campaign earned off actual clicks.

• Campaign A: 4.1x total, but only 1.5x on click
• Campaign B: 3.4x total, 3.0x on click
• Campaign C: 2.6x total, 2.4x on click
• Campaign D: 2.2x total, 2.1x on click

Now look at A. Most of its "performance" is view-through credit. On real clicks it's the weakest campaign in the account. It's the doorman. It's been standing near the checkout taking the bow.

And D, the one you were about to cut, is holding nearly all its value on click. It's doing real prospecting work, going out and finding people who hadn't decided yet.

So the call flips. I'd pull 10-20% out of A in steps over a few days and feed it into D and B, the campaigns earning their keep on clicks. Not all at once, because yanking budget around in big jumps throws delivery back into a mess. Small increments, then watch.

What usually happens when you do this is the campaign you're now feeding gets slightly worse for a day or two while it digests the new budget. That's normal. But across the whole account, the blended cost per purchase drops, because you've moved money off the campaign that was mostly claiming sales and onto the ones actually making them. On a healthy spend that's the difference between a profitable month and a flat one.

Same data. The total-ROAS view and the click view told me opposite stories. One of them was the doorman talking.

You can run this on any metric

The cleanest way to do this isn't even ROAS. It's cost per purchase, on the 7-day-click column. ROAS bounces around with AOV and refunds. Click-based cost per purchase is blunt and hard to lie to. If a campaign's click cost per purchase is way out of line with the rest of the account, that's your laggard, regardless of what its blended number says.

And the same logic runs all the way down. At the ad level, before I kill anything, I check whether it's converting on a click basis or just floating on view-through. I've seen ads that looked fine on total ROAS get cut the moment you see they've earned almost nothing off actual clicks. The blended number was hiding a dud.

Where this leaves you

None of this means view-through is evil or that you should switch your optimisation to click-only and starve the algorithm. That's the overcorrection, and it usually tanks delivery. Keep feeding Meta the fuller picture so it can do its job.

It just means you stop letting the doorman write his own performance review.

So the practical version: keep your optimisation window wide, feed the machine everything. But when you sit down to move money, add the 7-day-click column and make the call off that. Teach on everything, judge on clicks. The two-minute habit of adding that one column is what stops you scaling the campaign that's quietly fooling you.

If you've got a strong suspicion your reported ROAS is flattering you but you're not sure which campaigns are the doormen, that's one of the first things a Signal/Noise Audit pulls apart. We line up the blended numbers against the click data and show you exactly where the credit is real and where it's just standing near the door. No hard sell, just a clearer read on where your money is actually working.