Northbeam is an attribution tool for marketers. We generate independent, first-party advertising performance data, giving advertisers a clear, unbiased view of their ad performance. Our pixels, AI, and machine learning technology allows Northbeam to attribute credit for purchases across omnichannel customer journeys, meaning our advertisers can escape the “walled garden” problem of relying exclusively on in-platform data.

In this work, Northbeam has tracked more than $30 billion in ad spend from more than a thousand advertisers over the past few years. Our advertisers skew heavily toward what Meta defines as “Disruptors”: data-oriented, ecommerce-focused, and rapidly scaling brands like Ridge, HexClad, Grüns, and Comfrt – all Northbeam customers.

As a result, Northbeam is uniquely positioned to report on macro trends that performance and growth marketers are experiencing. I analyze this data every week in The Media Buyer, Northbeam’s free newsletter, now at 120,000 subscribers.

For this report, I’ve gone deeper, analyzing year-over-year H1 trends in the Northbeam data. What you are seeing is aggregated performance data across thousands of advertisers who were running Northbeam-measured ads in both H1 2025 and H1 2026. I have compared aggregate performance between the two periods across metrics like total ad spend, clickthrough rate, CPM and more. The attribution model is one-day clicks-only, and accounting mode is accrual unless specified otherwise. This gives us the most conservative and literal interpretation of ad performance.

My recommendation: explore this data and identify opportunities to compare your own ad account’s performance to the numbers you see here. If you are outperforming the benchmarks you see, congratulations – you can consider your ad account as outperforming. If you are underperforming against the benchmarks you see: consider why this is happening.

This data is representative of what Northbeam customers are achieving. Joining them could help you unlock increased performance of your ads, regardless if you’re above or below these benchmarks.

The analysis suggests that a new era of performance marketing is rising. Historical patterns of performance are beginning to shift, emerging ad platforms beckon, and efficiency is eroding to existential levels.

Simply put: Performance marketing in H1 2026 is more expensive and less efficient than it was in H1 2025. Performance marketing influencers on X and Linkedin have lamented this trend for months. Northbeam’s analysis gives data and support to their arguments.

First, we will explore high level metrics like spend, customer acquisition cost (CAC), and return on ad spend (ROAS), comparing H1 2025 to H1 2026. We will break this down by a few different categorizations of the ad accounts we are tracking.

We have studied thousands of ad accounts at Northbeam. The most common request for comparative analysis is at the “industry” level: meaning most advertisers want to understand how they compare to other fashion or skincare brands, for example. While industry analysis is useful for seasonality analysis, we’ve identified that extreme variability in customer journey, average order values, and target audiences of specific products inside these industry categorizations often confound the results. For example, a jewelry brand selling $20 earrings has a dramatically different performance marketing strategy than one selling $3,000 wedding rings, but both count as “jewelry.”

Our data scientists identified three stronger ways to categorize and compare ad accounts: the average order value (AOV), revenue, and ad spend of each ad account. A $20 AOV earring business doing $10 million in revenue has more in common with an accessories brand at the same AOV and revenue than it does with another jewelry business at a higher price point.

As you can see, median spend increased 12.38% but median CAC also increased 10.34%. Even without deeper analysis, and even without evaluating inflation or the power of the US dollar, this is a clear signal that year over year, advertisers are finding less efficiency in total across all their spend.

The “percentile” measures allow you to see the distribution of changes in performance year over year across ad accounts in the data. Looking at CAC. we see that the 75th percentile of accounts saw a -6.95% change, whereas the 25th percentile saw a 32.57% change. This means CAC worsened for most ad accounts, the majority more than 10% per year.

Revenue and MER show a similar wide distribution of “winners” and “losers.” The median revenue increase was only 3.32% year over year while median MER decreased -1.11%. However, the revenue percentile ranges swing widely between -21% (25th percentile) and +30% (75th percentile). This suggests that success is not equally shared and growth is not guaranteed for these ad accounts.

Categorizing the data by annual company revenue demonstrates a gulf between small and large companies. Businesses doing more than $50 million in revenue a year generally outperform smaller ones that lack resources, measurement, and margin to compete in increasingly expensive ad auctions.

Let us assume that sub-$5m revenue businesses are still establishing themselves, where $5m-$10m revenue brands are actively scaling. Looking at the $5m-$10m band compared to the >$100m band, we see dramatic changes in year over year performance. Revenue improved year over year for the $100m+ band, while it got worse for the $5m-$10m band. The same can be said for median MER, new revenue, and CAC, which increased double for the $5m-$10m band compared to the larger one.

Larger advertisers have the resources, creative output, and brand equity to outperform smaller advertisers. The data suggests smaller advertisers have to increase their spend more to acquire customers at a more costly rate than well-established advertisers. The ad auctions, while defined in a way that suggests equal possibility as an “auction,” are increasingly challenging for smaller advertisers. Scale has become more expensive.

Accounts spending less than $200k a month saw much weaker performance year over year than their higher-spending counterparts. New Revenue increases across percentiles are poorer across accounts spending under $200k a month. You can see that outsized gains (in the 75th percentile) are much better for accounts spending more than $200k a month across all their channels.

This suggests that clinical, predictable scale is less likely this year than it was last year. Meteoric growth is limited to a small but significant subset of outlier ad accounts who have found arbitrage and advantage partially due to their high volume of spend.

Sorting by AOV shows how increased prices are squeezing out advertisers with smaller margins. Generally, higher AOV and “luxury” products or bundled products have better margins. We can see that ad accounts with AOVs above $150 are outperforming year over year compared to smaller ones. Note for example the median new revenue increase of >$500 AOV accounts (+15.58% YoY) compared to $50-$100 AOV accounts (-7.32% YoY).

Increased AOV and its higher margins provides a cushion that permits these ad accounts increased flexibility in product margins, the ability to react to fluid ad auction prices, and simply pay more to acquire customers. As with any auction, whoever can pay the most usually wins.

Northbeam has broken several stories on Meta Ads performance this year. The platform has drawn the ire of many a performance marketer, while at the same time being the most predictable and effective full-funnel scaling platform available.

To better understand what is happening with Meta Ads, I’ve pulled weekly data aggregating various metrics across all advertisers we have on Meta. The charts paint a picture of an ad platform in flux.

The first step is to analyze reach. A good efficiency metric to look at is CPM, which many iconic billion-dollar brands study as a rude proxy for how cheaply they’re acquiring reach. Also good is cost per click: looking at how efficiently you’re capturing that engagement signal, which then serves as a retargeting and remarketing enrichment.

CPM shows a healthy normal flatness for the last 12 months, right up until the new year. CPMs increased 10%-20% on average starting January 2026, a normal trend I’ve seen several years in a row now.

Cost Per Clicks are tanking as well, which would suggest improved efficiency and ad performance. Or at least an influx of ad placements that are getting tons of clicks.

However, both CPM and CPC have started trending upwards since the start of the year. January 2026 marks the beginning of a new trend in Meta performance data, which is congruent with anecdotal reports from growth marketers.

Once again, right up until January 2026, Meta conversion rates were dropping by as much as 50%, and clickthrough rates were skyrocketing. This is suggestive of increased low-intent traffic dragging down normal conversion rate numbers. Each line moving in near-synchronized directions suggests that increased traffic is not necessarily high-purchase intent.

However, once again we see both trends course-correcting in the last month and a half.

Conversion rate has dropped in lockstep with Cost Per Click. The data suggests that advertisers are finding it much easier to reach and engage with audiences, but it isn’t translating into equal increases in purchase intent.

Clickthrough rates are at 12-month highs. Audiences are clicking. They’re viewing landing pages. So why aren’t they converting? Is the drop in conversion rate a function of increased volumes of traffic — more visitors, lower conversion rate?

It is possible. So let’s look at New Customer Acquisition Cost (CAC) and Blended ROAS (Return on Ad Spend, both new and returning customers) to evaluate.

Across the last 12 months, CAC has actually been dropping. So although conversion rate is dropping, this appears to be a function of increased traffic coming through ads. More people are clicking, meaning a lower percentage of those clicks end up as purchases.

Dropping CAC is a great sign, and it hit a 12-month low in February 2026. Then, it immediately began to rise and erased more than eight months of gains.

ROAS shows the same trend: blended ROAS improved in a linear fashion all the way up to BFCM weekend.

These trends suggest a clear pattern: Meta click-through effectiveness has increased while CPMs remained flat, demonstrating efficiency at capturing site visitors in increasing volumes. Up until early 2026, this was also reflected in improving Blended ROAS and dropping New Customer CACs.

In short: right up until Q1 2026, Meta performance was looking great. More traffic, better efficiency, at the cost of conversion rates.

But early in 2026, something shifted across these metrics.

ROAS plateaued right after BFCM 2025. But it hasn’t started trending up again like it did the previous year — if anything the trend looks to be flattening downwards.

Same with CAC: After hitting that low in February, New Customer CACs skyrocketed, erasing the efficiency gains earned through the entire previous year.

CPCs bottomed out and have bounced upwards in February 2026, breaking the downward trend. Looking at the chart, you can see a similar but less variable trend in the previous year. CPMs bounced in price, but over the 12-month horizon, they’re still cheaper than previous highs. CTRs still keep climbing as well, which is driving continual decreases in conversion rate.

These charts demonstrate that Meta’s ad algorithm is evolving. In aggregate across all Northbeam advertisers, we see related if not causal changes in performance between metrics.

We are stepping into a new era of performance marketing. The post-COVID wave of cheap acquisition is beginning to dry up under numerous pressures: weakening consumer confidence, increased customer acquisition costs, and even tariffs, to name a few.

Many are quick to blame Meta for these changes, but I think it’s actually a perfect storm of changing user behavior, the advent of AI, and our macroeconomic environment. To be blunt: blame is not a strategy. We must adapt.

This Northbeam data demonstrates that, as is usual for marketers, the tectonic plates of channels, funnels, and formats are shifting under our feet.

Marketing is a channel-nomadic practice. A new medium appears, rises in popularity with consumers, marketing conquers the medium, many products are sold, the medium begins to wane in popularity, consumers move on, marketers follow them, the cycle repeats again.

It is unclear what is next for performance marketers. What is clear is that the expected performance benchmarks across our existing channels are evolving. Responsible marketers will course-adjust in real time.

🥫 Snapchat will stop recommending fully AI-generated video in Spotlight. Meta's Vibes app went the exact opposite direction: who will be right in the end?

📊 Only 53% of marketers say AI delivers meaningful ROI, per new TransUnion research. Yet 89% are raising AI budgets — what are we doing y’all??

🫪 Early whisperings suggest ChatGPT ads don’t work very well. Which is bad, considering OpenAI’s entire financial model depends on them basically becoming the world’s biggest ad exchange.

Love this newsletter? Please forward it to your friends, coworkers, or Haley Stevens. Subscribe here to get this data in your inbox every week.