Project Nebula: AI Recency Boosts CTR 18% in 2026

In the dynamic area of digital advertising, understanding the impact of AI recency and frequency modeling is paramount for effective attribution. Our recent campaign, targeting small to medium-sized businesses (SMBs) for a new cloud-based project management solution, offered significant insights into how AI-driven strategies can refine ad delivery and budget allocation. How can we truly quantify the value of an impression that precedes a conversion by weeks versus one that occurs minutes before?

Campaign Teardown: Project Nebula Launch

Our objective for the “Project Nebula” campaign was straightforward: drive sign-ups for a new SaaS product designed to simplify project workflows for SMBs. We allocated a total budget of $350,000 over a 10-week duration, from January 8 to March 18, 2026. The primary performance indicators were CPL and Return on Ad Spend (ROAS). This wasn’t just about getting clicks. It was about securing qualified leads that translated into paying subscribers. We knew traditional last-click attribution wouldn’t tell the whole story, especially with a product requiring a considered purchase.

Strategy: Blended AI Attribution and Dynamic Delivery

The core of our strategy revolved around an AI-powered attribution model that went beyond standard last-touch or linear approaches. We implemented a custom algorithm within our programmatic advertising platform, designed to assign fractional credit to all touchpoints leading to a conversion, with a heavy weighting towards recency and frequency. The model considered not only the time between an ad exposure and conversion but also the number of times a user had seen our ads across different channels.

A significant component was the use of dynamic frequency capping. Instead of a blanket “3 impressions per user per day” rule, our AI analyzed user engagement signals and predicted conversion probability. Users showing high engagement (e.g., watching a significant portion of a video ad, clicking through to a landing page) might see more ads, while those exhibiting fatigue (e.g., scrolling past quickly, no interaction) would be capped more aggressively. This allowed for more efficient impression delivery, moving away from a one-size-fits-all approach that often wastes budget on uninterested segments.

Creative Approach: Education Meets Urgency

We developed a multi-stage creative strategy. For initial awareness, we deployed short, animated video ads (15-30 seconds) highlighting common pain points for SMBs in project management. These ran across display and social media channels. Mid-funnel creatives included carousel ads showing specific features and benefits, along with case study snippets. For the conversion stage, we used static image ads with clear calls to action and limited-time introductory offers, emphasizing a sense of urgency. All creatives were A/B tested extensively, with the AI model constantly feeding back performance data to optimize variations.

Targeting: Granular Segments and Lookalikes

Our initial targeting focused on custom audience segments built from our CRM data, including existing free-tier users and lapsed customers. We then expanded to lookalike audiences based on these high-value segments, specifically targeting business owners and project managers within industries like marketing agencies, software development, and consulting. Geographic targeting was initially broad across the US, with plans to narrow based on performance data. We specifically excluded users who had converted within the past 90 days to avoid ad waste. Our platform allowed for real-time adjustments to these segments, a capability we leaned on heavily.

What Worked: Precision and Adaptability

The AI-driven dynamic frequency capping proved particularly effective. By reducing exposure for less engaged users, we saw a noticeable decrease in Cost Per Click (CPC) for high-intent segments. Our overall Click-Through Rate (CTR) averaged 1.2% across all ad formats, a 15% improvement over our previous non-AI-driven campaigns. The AI model’s ability to identify optimal impression frequency for individual users meant we weren’t overspending on those unlikely to convert. This is a critical point. Simply throwing more impressions at a user isn’t always the answer, and in fact, can be detrimental.

The recency weighting in our attribution model also provided valuable insights. It revealed that while a final click was important, often a series of earlier, less direct interactions (like viewing a video ad or seeing a display banner) played a significant role in building brand awareness and trust, in the end contributing to the conversion. For instance, the model showed that users exposed to our educational video content within the first two weeks of the campaign had a 20% higher conversion rate than those who only saw bottom-of-funnel ads. This led us to reallocate 10% of our budget towards top-of-funnel video advertising in the second half of the campaign.

The campaign generated 1.5 million impressions and resulted in 7,778 conversions (defined as a completed sign-up for the free trial). The overall Cost Per Conversion (CPC) was $45, which was well within our target range. The ROAS, calculated based on projected lifetime value (LTV) of trial users, was an encouraging 1.8x, indicating strong initial product adoption.

What Didn’t Work: Over-reliance on Broad Targeting in Early Stages

Initially, our broad geographic targeting resulted in some wasted impressions in regions where our product had less relevance or existing market saturation. While the AI eventually course-corrected, the first three weeks saw a slightly higher CPL than anticipated. We quickly refined our geographic targeting to focus on key metropolitan areas like Atlanta, Dallas, and Denver, where our initial market research indicated a stronger need for project management solutions. This adjustment, made in week four, brought the CPL down by 8% almost immediately. It’s a reminder that even the most sophisticated AI needs good initial data to learn from.

Another area for improvement was the initial creative rotation. We had a strong set of creatives, but the AI identified that certain combinations of ad copy and visuals led to higher ad fatigue rates within specific audience segments faster than others. For example, highly technical feature-focused ads, when shown too frequently to general awareness audiences, led to diminished CTRs. We adjusted by introducing more variety and refreshing creatives every two weeks instead of every three, which improved engagement metrics by 5% in those segments.

Optimization Steps Taken

1. Dynamic Budget Shifting: Based on the AI’s real-time performance analysis, we continuously shifted budget allocation between channels (display, social, search) and audience segments. For instance, when a particular lookalike audience started showing diminishing returns on social, the AI would reallocate budget to a better-performing display network segment.
2. Creative Refresh Cycles: We moved from a monthly creative refresh to a bi-weekly cycle, injecting new variations based on AI-identified fatigue signals. This ensured our messaging remained fresh and engaging.
3. Negative Keyword Expansion: For search campaigns, the AI identified several non-converting search terms that were consuming budget. We added these to our negative keyword lists, reducing irrelevant impressions and improving the quality of incoming traffic.
4. Landing Page Optimization: A/B testing of landing page elements (e.g., headline variations, call-to-action button colors) was heavily informed by AI analysis of user behavior on the pages. Pages with higher scroll depth and lower bounce rates for specific ad variants were prioritized.
5. Recency-Weighted Bidding: Our bidding strategy was adjusted to place higher bids on users who had recently interacted with our ads, recognizing the increased likelihood of conversion within a shorter timeframe. This is where AI recency truly showed its power, allowing us to capture high-intent users more effectively.

The campaign demonstrated that a deep understanding of AI recency and frequency modeling is no longer a luxury but a necessity for sophisticated digital advertisers. By letting AI guide our attribution and delivery, we achieved a level of precision that traditional methods simply cannot match.

In the end, the success of Project Nebula underlines that integrating AI into every layer of campaign management, from initial targeting to ongoing optimization, yields superior results and a clearer understanding of the customer journey.