391: Apple Destroyed Facebook Ads. Now What? With Nick Shackleford

Podcast: Download (Duration: 56:04 — 64.4MB)

Measuring Facebook ads after iOS 14 requires triangulating three sources rather than trusting the ads dashboard: MER (marketing efficiency ratio, total ad spend divided by total revenue), a post-purchase survey asking how customers heard about you, and a week-over-week correlation between Facebook’s reported purchases and your actual Shopify numbers. Nick Shackelford has spent over $85 million on Facebook and manages between 160 and 275 accounts.

In this episode I brought Nick Shackelford back on the show to cover the state of Facebook advertising once iOS 14 had several months to settle. He runs Structured Agency, Constant Creative, and the Geekout events.

Below is the full picture: why the conversions API did not fix anything, the correlation sheet method, why interest stacking replaced lookalikes, the on-platform audience shift, and the multiple ad account tactic.

Key takeaways

• Ad spend dropped 30% from January to June across the 116 brands Nick had visibility into.
• CPMs rose 48% over the same period. Facebook still posted its most profitable quarter ever.
• The conversions API did not fix optimization. Opted-out user data cannot be used regardless of how it gets back to Facebook.
• Measure with MER plus a post-purchase survey. Neither feeds the ads manager, and together they tell you what is real.
• Lookalikes were dropped in favor of stacked interest audiences. Lookalikes work for two days and then die without explanation.
• On-platform audiences beat site-visitor audiences. Video viewers and engagers never left the app, so Facebook keeps that data.
• Make videos longer, not shorter. Longer watch time builds better remarketing pools.
• New advertisers are in a better position. One channel means no attribution confusion.

What actually happened to Facebook ad spend and CPMs

Across the 116 brands Nick had access to, ad spend fell roughly 30% from January to June while average CPMs rose 48% over the same window.

His read on the CPM number is that it is more trustworthy than the revenue reporting, because it measures on-platform impressions rather than off-platform conversions.

The part he flags as strange: Facebook posted its most profitable quarter ever while a large share of advertisers pulled budget. He has no data proving anything, and in eight years on the platform he has never seen that combination.

Where the pulled budget went: some to SMS, some to influencers and content production, some to traditional channels including billboards, and a meaningful amount simply held in reserve.

Why the conversions API did not fix attribution

The conversions API improves what Facebook can report, and it does not improve what Facebook can optimize against. Data from opted-out users cannot be used for targeting or lookalike building regardless of how it reaches the platform.

Facebook positioned CAPI integration as the most important fix available. Nick’s assessment after implementing it widely is that it has not changed how his team optimizes.

Offline conversion uploads through a Zapier integration are a marginal improvement over CAPI alone, because you post back as much as you can get. The full picture stays out of reach.

The same limit applies to third-party attribution tools. Hyros overlays its metrics onto the ads manager interface, which feels useful, and Facebook cannot act on numbers it is not permitted to use.

How to measure Facebook ad performance with MER

MER is total advertising dollars spent divided by total net revenue returned, and it is the top-level number to run the business on.

What goes into it: ad spend, shipping costs, and product costs. What stays out: employee salaries, overhead, vehicles, and everything else operational.

The reason this matters is the halo effect. Spending $100 on Facebook with zero attributed conversions can still produce branded search lifts, Amazon growth, and organic traffic that no platform will credit.

Nick’s working expectation for a healthy account: Facebook accounts for at least 15% to 20% of overall sales, with organic, Snapchat, and Google producing the rest in some combination.

How to use a post-purchase survey for attribution

Add a single question immediately after checkout: how did you hear about us? Offer Facebook, Instagram, Snapchat, and Google as options.

Pair it with a first-time customer acquisition tool. Nick uses Lifetimely, which is inexpensive and reports first-time versus repeat cohorts, or Triple Whale, which goes deeper for more money.

The legal reason surveys work when tracking does not: a customer volunteering information is different from a platform tracking someone who opted out. Analytics tools like Lifetimely sit in the same category as Google Analytics, business reporting rather than marketing data feedback.

The gap to be aware of: surveys only capture buyers. Anyone who hit a product page and left is invisible to this method.

The correlation sheet method for spotting bad reporting weeks

Build a simple sheet comparing Facebook’s reported spend, purchases, and cost per purchase against what Shopify actually recorded, tracked week over week.

Over a few weeks you establish your normal gap. Most accounts settle at Facebook under-reporting purchases and conversion value by 20% to 30%.

Once you know your baseline, an anomalous week identifies itself. If your usual gap is 25% and this week shows 50% or 60%, the drop is a reporting artifact rather than a performance collapse.

Facebook under-reports and effectively never over-reports, so the direction of the error is at least predictable.

Why interest stacking replaced lookalike audiences

Structured moved off lookalikes in late October 2021 and switched to stacked interest audiences that Facebook provides.

Before the change they ran heavily stacked lookalikes, combining 1%, 2%, 3%, and 4% audiences into single ad sets. The failure pattern was consistent: a lookalike performs for a day or two, then dies on day three or four with no explanation.

Interest stacking groups congruent interests into one ad set. For a mom-targeted product that means mommy, mommy of teens, stay at home mom, and motherhood together rather than isolated.

The tradeoff is learning. Separated audiences told brands exactly which segment converted, and stacking gives up that granularity in exchange for the stability that matters more right now.

Why on-platform audiences outperform site visitors now

Audiences built from on-platform behavior are outperforming site-visitor audiences: video watch time, engagers, followers, likers, and savers.

The mechanism is straightforward. Someone who watched your video never left Facebook, so no data had to travel off-platform and back, and none of it needs scrubbing for opt-out status.

This produces a counterintuitive creative recommendation: make videos longer. Push past the standard 15 or 30 seconds toward 45 seconds or a full minute, and extend the final call-to-action card by a few seconds.

The first three seconds still have to hook. Every additional second watched after that is data Facebook is allowed to keep.

How to segment video view audiences for remarketing

Segment 25%, 50%, and 75% view audiences separately, and use the split to decide what the second touch should say rather than just who sees it.

The Miracle Sheets example shows why. That brand sells to bachelors, mothers, and college students, three distinct consumers needing three distinct messages.

If your first video front-loads several value propositions and someone dropped at 25%, they never reached the arguments in the back half. Showing them a standard 75%-viewer follow-up assumes context they do not have.

The practical questions that follow: does this person need the same video again, does the second video need to open differently, or does the first creative need its call to action repeated earlier?

The multiple ad account tactic for scaling

When an account plateaus, open a fresh ad account under the same business manager and relaunch your best-performing ads into it.

Structure: one campaign, one ad set, all your proven creative, a high budget of $5,000 to $10,000, and a cost cap that prevents runaway spend. Keeping it under the same business manager preserves post IDs and their accumulated social proof.

Nick has three brands running this across different niches, with as many as 25 accounts open for a single brand. Nobody knows why it works.

Fresh accounts sometimes launch with CPMs so high they cannot be made profitable, and the very next account launches cheap. His approach is to kill the bad ones and keep going rather than diagnose it.

Why a brand new advertiser is better positioned than an established one

Starting today with no history is an advantage, because a single channel is unambiguous. Every sale came from the one platform you are running.

Established brands at $1M, $5M, or $10M are spread across Facebook, Google, Snapchat, and TikTok, which means every measurement is a triangulation problem.

The requirement for a new advertiser is getting UTM codes right at the campaign, ad set, and ad level. You give up first-click data and keep a clear view of which content works.

There is also a psychological edge. Established advertisers are anchored to 2018 and 2019 performance, which makes today feel like collapse. A new advertiser has no such baseline.

What to invest in instead of platform tactics

Content and customer definition are where the investment goes now. Putting up a product on a colored background and running traffic at it stopped working.

The demand is still there. Ecommerce census data and Q3 reports show consumers spending, and people do not stop celebrating birthdays or Christmas. The problem is trust in the platforms rather than absence of buyers.

Nick’s business-level framing: build something people come back to buy from. Consumable products, limited-edition drops, and genuine exclusivity are what is working, and changing a competitor’s color or cap is not.

Owned channels close the loop. Email and SMS are how you keep the customer once acquisition has become expensive and hard to measure.

Frequently asked questions

What is MER in marketing?

Marketing efficiency ratio: total advertising dollars spent divided by total net revenue returned. It includes shipping and product costs and excludes salaries, overhead, and vehicles. It is the top-level number to judge a business on when platform-level attribution is unreliable.

Does the conversions API fix iOS 14 attribution?

No. CAPI improves what Facebook can report and does nothing for what it can optimize against, because data from opted-out users cannot legally be used for targeting. Nick’s team implemented it widely and it did not change how they optimize accounts.

Why did Facebook CPMs go up after iOS 14?

Across 116 accounts, CPMs rose about 48% from January to June 2021 while overall ad spend fell roughly 30%. Impression-based metrics like CPM are more reliable than conversion reporting because they measure on-platform activity that never depended on off-platform tracking.

How do you know if Facebook is under-reporting your sales?

Track Facebook’s reported purchases against your actual store numbers week over week to establish a normal gap, typically 20% to 30% under-reported. When a week shows a 50% or 60% gap instead, the anomaly is reporting rather than performance.

Should you still use lookalike audiences?

Stacked interest audiences are proving more stable. Lookalikes tend to perform for a day or two and then die off with no explanation. Group congruent interests into a single ad set, accepting that you lose the ability to attribute results to individual segments.

What Facebook audiences work best now?

On-platform audiences: video watch time, engagers, followers, likers, and savers. Those users never left Facebook, so the data never had to travel off-platform and does not require opt-out scrubbing. This is also why longer videos now outperform very short ones for audience building.

How do you attribute sales without tracking?

A post-purchase survey asking how the customer heard about you, combined with a first-time customer acquisition tool like Lifetimely or Triple Whale. Volunteered survey data is permitted where tracking is not. The limitation is that it only captures buyers, missing everyone who browsed and left.