Why CPM is the most important variable in Facebook Ads, and why you might be getting it all wrong.

This post was co-authored by Matt Griffin, Growth Designer

Over the past few years, Facebook advertising has become a critical source of lead generation for B2B SaaS and consumer e-commerce startups alike. Our company was no different. It took us roughly six months to figure out this strategy but once our creative audience targeting, and call-to-action flows finally clicked, we were able to scale this channel handsomely for well over a year.

Then suddenly, our campaign performance started to deteriorate, taking down our lead flow with it. Naturally, we dove deep to diagnose the root cause of the issue. As expected, the decrease in performance wasn’t isolated to a single variable. Multiple ratios conspired against us and their compound effect was killing us. But there was one variable that stood out. Over the previous weeks, our cost per impression (CPM) had nearly doubled, effectively cutting our funnel- throughput in half, so there was no point focusing on anything else until we fixed this issue.

It appeared that the increase in CPM was related to recent changes to the Facebook algorithm that favored more user-generated content, and cut down on the number of ads shown to Facebook users. In other words, the inventory of ad placements had been significantly reduced and what we were experiencing was the law of demand and supply in action.

Over the two months that followed, we tested multiple different strategies to bring the CPM of our campaigns back down to previous levels. Most of these strategies taught us something new, and eventually we managed to find the right approach to set us up for another year of good performance. In this post, we’ll share our thought process as we worked through the problem from the marketing strategy standpoint, as well as the design perspective.

Our first strategy to bring our CPM back inline was simple. Let’s shift our geo-targeting to less competitive countries where CPM will naturally be lower, but make sure demographics are still generally consistent with our target audience. At first, this strategy seemed to work wonders, our CPM and cost per lead were lower than ever. But unfortunately, the enthusiasm didn’t last long. The prospects in these countries turned out to be very different from our general North American customers, and our sales closing rates tanked.

Our next thought was to slice and dice our regular targeting to find narrow pockets of high performance. Hyper-targeting can work in some cases but it did not help our situation. Our problem was low ad inventory, so narrowing down our targeting wasn’t solving the real issue.

So, we decided to get a bit more creative. Generally speaking, the CPM reported in dashboards is the average of multiple successful bids. The larger the size of the bidding war, the higher the price required to win it. Keeping targeting as specific as possible can help you stay clear of heavyweight competitors with deep pockets.

This insight informed our next test. Rather than targeting at the state or provincial level, we isolated the best performing cities and towns across North America and spun separate campaigns for each one in hopes of keeping our bidding wars very small. Unfortunately, this didn’t work either. Our CPM did improve but, overall, our campaigns didn’t spend very well so we were back to square one.

It seemed as though we were looking at it all wrong. Perhaps improving campaign targeting wasn’t the answer, but rather the type of marketing campaigns we were running. All of our campaigns were designed as direct response middle to bottom of the funnel. We had no brand building or top-of-the-funnel campaigns.

That strategy makes sense for early-stage companies but it becomes limited as companies scale. This question informed our next round of testing. Perhaps it was time for us to extend our campaigns up-funnel. This would allow us to focus on less expensive leads, that could later pair with a separate set of nurturing and conversion-oriented campaigns.

We leaned on our best performing content and tested a couple of gated-content lead-generation campaigns. The results were encouraging. Our costs per lead was a fraction of regular leads and volume was high. If we could then get enough of these prospects to speak to sales, we’d be back in business. Not quite though. There was another variable: velocity. Nurturing campaigns didn’t change the fact that time had to be right for prospects, these campaigns just help us stay top of mind. Even though the unit economics were promising, it would have taken months to get to the necessary critical mass for this strategy to power our sales engine. Still, it was not a complete fail. The solution could work, just not within the time horizon we needed it to. So, we dialed down the budget but kept the campaign to continue learning this strategy.

I said earlier that we had no brand awareness campaigns, but that wasn’t entirely true. We didn’t have any campaigns solely dedicated to brand awareness, but in fact, our direct response campaigns were also building brand awareness through hundreds of thousands of carefully targeted impressions. Some of those impressions led to requests to speak to sales via our regular paid lead-gen funnel, but many others found their way to us “organically”. As it turns out, these paid-assisted organic leads were amongst our best performing leads, so for our next strategy, we decided to try to deliberately control this spillover effect.

We placed a pixel in our organic landing page so we could track the cost per assisted lead more closely, and later on be able to optimize our campaigns against it. As anticipated, we found a very predictable correlation between our spend and paid-assisted organic leads. What surprised us, though, was that there was a lagging effect of a few weeks. We assumed that this lag was due to timing and that brand recall was therefore playing an important role, so we decided to experiment with branding as a way to improve brand recall and through it, the paid-spillover effect.

Prior to this point, we had tested dozens of ad creative variations and found that three of them outperformed the rest. They were our set of control ads. Their style was rather utilitarian and had been judged exclusively on their ability to convert. This new round of testing was different though since we were changing the definition of success. Our new goal was to affect brand recall, not just immediate conversion. To accomplish this, we started by trying two different creative styles. Let’s zoom into this creative process, and examine how we attempted to solve the problem from the design perspective.

Creative Approach 1A: Branding Intensive

As part of the ongoing effort to redesign our branding, we had been developing a visual language that hoped to give the brand more depth and personality. This required creating new design components and an asset library that could be used across the team for all of the new marketing deliverables such as emails, articles, e-books, stationery, and clothing, as well as ads. Our first approach was to leverage this style to create ads that would catch the eye and be more memorable.

Creative Approach 1B: Persona Informed

Our second approach was persona oriented and designed to resonate with agency folks. We intended to be empathetic by showing that our company understood the struggles of their work-life and that our product could help them streamline their workflows. We chose stock photography that captured joy and team building and messaging that showed we understood their struggles and could help to make their lives easier.

Soon after releasing the new set of ads, we realized that they were underperforming. Cost per click and cost per acquisition were higher for these new ads than for our control set, and there was no evidence that the paid-to-organic ratio had improved. But although the new creative didn’t perform well for direct response, the brand intensive ads did perform well for sponsored content campaigns. This set of ads promoting our lead magnets was showing better cost per lead, which strengthen our new up-funnel campaigns.

But there was a silver lining. Although our first round of creative tests did not affect our paid-to-organic spillover effect, we noticed that some of these ads got particularly high relevance scores and these same ads also had a far better cost per impression.

This led us to our next strategy. We set out to create ads designed to receive a high relevance score, in hopes of being rewarded with cheaper impressions. Our first guess was that engagement played an important role in that, so we tested several different creative styles aimed at catching people’s attention.

Creative Approach 2: Shock and Awe

This new creative approach combined a clear and concise product messaging with unusual stock photography. On a social feed, you are given mere seconds to grab a prospect’s attention and reel them in for a closer look at what you offer. We believed that by looking out of the ordinary, the odds of capturing their attention would be greater. With more attention being paid to the creative, the odds of viewer interaction such as clicks, comments, and conversions could be higher, thus driving up our relevance scores.

As we hoped, the shocking photography generated a higher level of viewer interaction. That said, a lot of the engagements focused more on the obscurity of the photography than our product messaging. Still, our relevance score on direct response increased dramatically and suddenly things started to click again. We started being able to consistently create ads that received a good relevance score and our cost per impression started to improve, but our success was tempered by leads with low intent. Although the eye-catching and striking creative ads had performed better than many of the recent ad campaigns, many of the leads resulting from it had very low to middle funnel performance, which effectively negated our up-funnel victories. We were getting close though. Next, we needed to continue grabbing our audience’s attention but keep their focus on the copy.

Creative Approach 3: Copy-first Design

This time, the creative ads would be structured with minimal visual language and contain content messaging that exists at the top of the ad’s information hierarchy. Our formula for this new series of ads was to create an eye-catching gradient with two or more colors as the background. We placed short and direct messages into the center of the canvas that was typeset in bold font families. The colors that make up each gradient portrayed feelings such as calmness, excitement, and trust. It was important to ensure that the font color contrasted heavily with the gradient background so that it stood out. We wanted to ensure that the product messaging existed at the top of the ad’s information hierarchy so that it would be the first element interpreted by the viewer.

This new creative direction turned out to be the breakthrough we have been looking for. After a couple of weeks, our campaign performance was back to normal levels. Due to the minimalist style of the creative ads, we were able to stop testing design styles and focus on testing the effectiveness of our product messaging in the ads. After testing out different messages and gradient colors, we found a set of ad designs that received near perfect relevance scores. But our sales pipeline was still irregular, with a mix of high-quality and low-quality leads. We needed to figure out which copy was performing well down the funnel as well. Thanks to our integration of UTM tracking into our CRM data, we were able to determine this with just a few pivot tables. We found that two of our creative-copywriting combinations were responsible for most of the recent leads, so we killed every other ad combination and focused our entire budget on these two ads. A week later, our funnel was healthy and our sales pipeline was filling up again.

We had finally fixed our Facebook lead generation channel! In the process, we learned a lot about which type of creative ads work well in social ads and which don’t. We learned that clever targeting has a ceiling, and that to break through that ceiling you need good creative. We learned that up-funnel strategies can create a powerful flywheel, but they take time to build. We learned that the halo effect can be more than a nice bonus, it can be the most important benefit of advertising spend. We learned that better engagement and relevance scores do improve cost per impression, but not all impressions are created equal. We learned that bringing focus to your messaging is crucial for lead quality, particularly by leveraging self-selection. Most importantly though, we learned that in Facebook advertising clear beats clever — just get their attention and then get to the point.

Product Operations @Shopify — ex Head of Growth @HeyOrca