In the last few years, we’ve seen a seismic shift in how paid advertising works, and it’s something you’d only notice if you were paying extremely close attention. Performance marketers have since been racing to adapt to the evolving landscape of targeting, optimization, and measurement and honestly, it can feel a bit overwhelming just trying to keep up with all the changes. 

Not to worry, the team at MAVAN has spent years reverse engineering ad network algorithms and expanding our understanding. We’re constantly stress testing and refining our learnings to make sure we don’t let ourselves, or our clients, fall behind. 

What you’ll see below is an in-depth chronology of the evolution of performance marketing, complete with critical insights, real world examples, and practical advice. 

1: The Deterioration of Precise Targeting Inputs

Remember years ago when Meta (then Facebook) announced it was going to move away from interest-based targeting? That moment marked a fundamental shift in how we approach performance marketing. Eventually, other networks followed suit, further causing the deterioration of interest-based targeting and keywords, making it harder to pursue contextual targeting. 

Post Cambridge Analytica, privacy concerns became the talk of the town, leading to more restrictions like Apple’s iOS 14.5 update, and the introduction of cookieless tracking that impaired device-level attribution, reduced the performance of lookalike campaigns, and limited other things like retargeting. Ad networks are now working with considerably less granular data inputs on their users. Performance Marketers now need to apply a deeper understanding of what data the network is likely using for targeting signals and what it lacks.

2: The Rise of Automated Targeting & Creative

As manual targeting took a backseat, fully automated campaigns powered by machine learning and AI took center stage across major ad networks. Google Ads for example, switched to Smart Bidding, where the algorithm optimizes bids in real-time towards conversion values, paving the way for Google to force all mobile advertisers to adopt Universal App Campaigns that automated all targeting for you. Targeting has transformed: we must now define clear goals and data signals (like “purchase” events) to guide the algorithm towards our most probable audience.

At the same time that automated targeting was becoming the norm, automated creative generation also emerged as a transformative force in digital marketing. Channels like Google and Meta have revolutionized ad creation through ad units like Responsive Search Ads (RSA) and Dynamic Creative Optimization (DCO), allowing marketers to feed raw materials like copy, images, and videos, into an automated system that creates dynamically tailored ads. The latest evolution has even incorporated generative AI that enables ad platforms to develop customized creative content for specific audiences. Personalized ad creation at this level would be almost impossible to achieve manually, making this an unprecedented opportunity to engage customers on an even deeper level.

3: Broad Targeting and Lower Funnel Optimization Events Deliver ROAS at Scale

With automated targeting being the new norm, broad audience targeting is now king. This approach, when used with the right optimization events, can deliver Return on Ad Spend (ROAS) at scale as broad targeting over time will usually outperform narrow targeting when fed the right data signals. 

Picking the right optimization event isn’t always straightforward but it’s important to remember that it needs to occur frequently enough post-click (typically within 24 hours) to provide substantial data for the algorithm to learn quickly.

Here’s an example: Let’s say we have a campaign optimizing for an “Add to Cart” event. If users tend to add a product to their cart within a day, this event becomes a valuable indicator of purchase intent for your ad network. In turn, this helps the algorithm fine-tune its targeting faster than waiting for purchases to build up.

4: The Necessity of Feeding Ad Networks More Data

Back in the day, I would set up thousands of campaigns that split out every possible permutation of targeting and then used automated bidding logic to optimize based on relatively early signals of success. This no longer works… at all.

The current machine learning models employed by ad networks need exponentially more data fed into them before they can even begin to deliver precise and effective optimization. This requires a consolidation of campaigns and budgets in order to aggregate enough data to fuel optimization. It’s now uncommon to need more than a few campaigns running each with a handful of ad sets.

The logic behind this is two-fold. Firstly, an algorithm with access to more data can draw more accurate inferences and make smarter decisions. Secondly, a larger budget means more ads, more clicks, and more opportunities for the algorithm to learn from a larger data set. Therefore, marketers should focus on providing quality, high-volume data to help machine learning models effectively fine-tune campaign performance.

Since the algorithm needs deep funnel events in order to find your true audience, it needs to see a lot of data to find those conversions and build a targeting model off of them. For example, the minimum scope you need to hit in order for an ad set to leave the learning phase is 50 events over 7 days, according to Meta. If you have a campaign optimizing toward purchases and it costs $200 per purchase event at the start of a campaign, that means you need to spend a bare minimum of $10k on that one ad set in order to get enough learning to let the algorithm get into its optimization phase. 

This is probably the concept that people understand the least. It requires solid expectation setting within the company when budget planning. By setting a proper budget you can avoid running into costly optimization issues. 

5: The Continued Importance of Upper Funnel Metrics

Ad network algorithms look at metrics across your funnel, from first impression to final conversion event. Despite the heavy reliance on lower funnel optimization events (which happen at a late stage of the purchase funnel), algorithms don’t ignore upper funnel events. Click-through rate (CTR), relevance score, and spend scale/delivery continue to play a crucial role in the overall performance of your campaigns. 

These metrics provide early indicators of success, assisting the algorithm during the “learning phase” in understanding user engagement and the overall appeal of your ads while it gathers additional data to understand performance against lower funnel goals.

For instance, a high CTR often signifies that your ad is relevant and appealing to your audience. Ad platforms use CTR as a signal of quality and engagement. Similarly, a high relevance score (which includes ad quality score) implies that your ad is likely to engage users, which tells the algorithm that your ad might perform well, leading to better delivery and performance at lower costs. 

Conversely, a sudden drop in these metrics can alert you to potential issues, providing an opportunity to make necessary adjustments before the situation deteriorates. On Meta, you can look at engagement rate, conversion rate, and ad quality metrics. On Google, you can look at Quality Score and Impression Share Lost (rank).

As crazy as it may seem, looking at which ads get the most delivery & spend is now an important metric. Ads that don’t get scaled up by the algorithm mean they failed to pass the test for one of many possible variables that the algo is looking for, like poor ad quality score, which may limit your campaign even before you have any conversions. Look for ads with higher spend as an early indicator that an ad likely has a combination of positive performance signals that a platform is looking for in order to scale.

6: Creative is the New Targeting

In this new era of automated campaigns, creative content is more important than ever. It’s your primary tool for audience identification and engagement. Well-tailored creative will resonate with your audience, and just as important, non-target audience groups will ignore it. Learning which audiences will and will not convert from your creative teaches the algorithm to focus on higher converting audience segments.

Take this scenario, for example: a product appeals to millennials and Gen Z, but for different reasons. Instead of relying on a one-size-fits-all ad, designing distinct creatives that appeal to the unique preferences of each demographic can significantly improve a campaign’s conversion rates for each audience. 

Also keep in mind that tailoring your creative to match the placement format, such as using square creatives for Instagram feeds or vertical video for Stories, also unlocks additional opportunities for your ads to be shown. In other words, you want to craft creatives that are extremely compelling to a specific persona and ensure it follows best practices for the placement.

7: Measurement Shifts from Individual to Aggregated Data Sets

Recent privacy restrictions with Apple and Google deprecating device ID tracking have created a new challenge in attribution measurement. No longer do we have the luxury of relying solely on individual user-level attribution. Instead, we must now work with aggregated data sets as well as multiple conflicting sources of performance data, making it harder than ever to confidently measure performance data. 

Ad networks now show you “modeled performance data” using their own black box machine learning models, which never seem to line up with your internal data sources. Far from using last-click attribution methodology, Meta now uses privacy-safe attribution models and probabilistic methods that leverage “aggregated and anonymized reporting, probabilistic modeling, and contextual targeting.” In other words, Meta and other networks are inventing their own attribution methods to close the gap created by new tracking restrictions. These changes require a new approach to data interpretation.

Companies are better served developing their own internal source of truth. Those that have managed to successfully navigate this shift have largely done so by building their own proprietary measurement approaches, since no complete, out-of-the-box solution currently exists. 

Large advertisers have been better equipped to adapt to these challenges than small advertisers, since data science teams are capable of developing sophisticated internal attribution models and have the ability to leverage first party data sets. Approaches differ from company to company, but we see large advertisers developing internal attribution models using heuristics, leaning on pre-iOS 14 conversion models, pulling in Apple’s SKAN postbacks, and other post-hoc methods to triangulate to a source of truth.

However, this may not be possible for smaller companies. In these cases, a flexible and creative approach is essential to navigate the evolving landscape of data privacy and attribution. Some lean more heavily on blended organic and paid data to act as a guardrail for overall performance. While others are testing out incrementality measurement solutions. 

For further reading on measurement approaches, I recommend you check out: AppsFlyer and Adjust resource centers for iOS & SKAN.

In this new era of performance marketing, our roles have evolved from human-controlled to AI-influenced strategies. The magic formula for high-performing paid advertising now requires a strategic blend of campaign structure, data signals, and tailored creatives.

Key Takeaways

1. Consolidate campaigns and ad sets: Allow the algorithm to access more data, improve its learning, and optimize performance by limiting campaign splits.
2. Optimize towards frequent lower-funnel events: Choose events that typically happen within 24 hours of a click, such as “Add to Cart”. This helps the algorithm refine its targeting.
3. Focus on producing high CTR ads with high ad quality metrics: High CTR signifies ad relevance and audience engagement, influencing ad delivery and overall performance. Build your creative testing strategy around identifying high CTR/high conversion combos. Regularly test and prune ads to maintain high CTR and high ad quality metrics.
4. Allocate enough budget to meet data thresholds: In order to exit the learning phase on Meta, ensure each ad set can hit 50 conversion events in 7 days. Avoid the trap of not having enough data to make decisions.
5. Implement multiple attribution methods: Develop a flexible approach to navigate the shift towards aggregated data sets – explore using SKAN postbacks, blended internal revenue/signups, ad platform reported attribution, and post-signup surveys to establish a reliable source of truth for your campaign performance.

Remember, as we continue to navigate the changing landscape of performance marketing, MAVAN is here to support you every step of the way. If you found this guide helpful, please consider subscribing to our blog and sharing your thoughts on this article on LinkedIn.

If you’d like a free consultation to learn more on these topics, please reach out to letsgo@mavan.com to set up an exploratory call.