Today’s economy requires us to be more diligent in advertising spending. Fortunately, the viable paths to profitable marketing can be found within data.
The economy today is drastically different than it was a few years ago. Nowadays, we are all required to operate with less resources. When it comes to marketing, that means we need greater precision in all of our initiatives. Fortunately, the viable paths to profitable marketing can be found within data.
Since we began helping ecommerce brands optimize their marketing with data-driven approaches, we have repeatedly identified remarkable optimization opportunities within the first month. We saw the same optimization patterns over and over again across ecommerce verticals. In this article, I will share what we have learned and discuss how you can find marketing optimization opportunities in your organization through the lens of data science.
As discussed earlier, brands can gain tremendous growth opportunities by simply cutting back budgets on initiatives that are not working. In our customers’ cases, we saw many such opportunities go unnoticed, especially when a brand invests heavily in marketing and has a large number of ad campaigns running.
An example of an ecommerce brand’s ad campaign performance is shown below.
If you are experiencing unprofitable advertising and cannot determine the cause by analyzing high-level ad performance data, try examining individual ad campaigns to evaluate if underperforming campaigns are impacting your profits.
As a rule of thumb, if an ad campaign has been running for a long time but is still far from breaking even, it is likely wise to reduce investment in that campaign.
The flip side of the low-hanging fruit is that high-performing ads are often ignored. In many of our customers’ cases, we discovered profitable ads that deserved more investments. These promising signals were previously buried in the noises created by underperforming campaigns.
To highlight high-performing ads and never miss an opportunity, you first need to consolidate your campaigns and ads.
For instance, if you have ten ad campaigns and two of them are consistently profitable, three are mediocre, and five are far from profitable, you should focus on the two profitable campaigns, reduce spending on the five unprofitable campaigns, and investigate the three mediocre campaigns to see if there are any high-performing ads.
With these two approaches, our customers who were bleeding cash on paid marketing came much closer to breaking even on advertising.
Most ecommerce startups have very specific target audiences — otherwise, the space would be crowded with competition from large players. However, when it comes to advertising, there is an industry myth that brands can rely on ad platforms’ broad targeting approaches to achieve optimal results.
It is clear this strategy is not applicable to all, as many ecommerce startups have told us that they have been struggling for a long time to get broad targeting to work. Let’s explore why broad targeting may work for large brands or brands that started advertising a few years ago, but doesn’t work for your startup.
Ad platforms use artificial intelligence, machine learning, and data science to improve their algorithms. However, these algorithms rely heavily on large amounts of training data. On one end, there are inputs such as consumer demographics, behaviors, and interests. On the other end, there are conversion events like purchases, which serve as outputs. Feeding the ad platforms with more conversion events can help them better identify the ideal customer profiles for you.
Before privacy changes in iOS 14 and cookies, ad platforms could track a lot more consumer data, such as browsing and purchasing behaviors across platforms. This provided more inputs and outputs for their algorithms to function well, even with little data.
However, these privacy changes have significantly impacted ad platforms’ tracking capabilities and have resulted in reduced accuracy of their targeting. In real-world case studies, we have found that ad platforms that heavily rely on consumer behavior tracking are more affected by the changes than other platforms. As a result, these platforms’ budget allocations across market segments may not align with brands’ business cases.
Misalignments are more prominent for startups due to limited conversion data. It is not uncommon for ad platforms to allocate over 50% of ad spend to market segments that are not the target audience for our customers.
Due to these industry changes, it is critical to inform ad platforms of your target audience to achieve satisfactory advertising results. By specifying your target audience, you can narrow the search and testing scope of ad platforms’ algorithms, shortening their learning phase and improving advertising performance.
In practice, these specifications may include your customers’ location, age, gender, income level, interests, favorite products, and other relevant factors. You can evaluate individual market segment breakdowns to determine if any segments outperform others.
Our customers who specify their target audience in ads quickly achieve profitable results. Even if they choose to broaden their target later on, their advertising performance remains at a high level because ad platforms have been trained for their business cases.
If you are using broad targeting and cannot achieve satisfactory results, try analyzing your ad performance by market segments. You will likely find insights from those breakdowns.
Although there are numerous paths to optimize marketing, not all of them are immediately obvious. If over 50% of your ad campaigns are underperforming, you are unlikely to find actionable insights from your market segment breakdowns.
In this case, you should first eliminate the marketing initiatives that are clearly underperforming and prioritize the high-performing ones according to step 1, scaling back on unprofitable initiatives, and step 2, doubling down on high-performing ads. Once you reach a point where half of your ad campaigns are performing well, you can start analyzing market segment breakdowns to evaluate viable paths toward profitable advertising.
It’s important to note that advertising performance data is often skewed due to outlier events. For instance, a high-value purchase from a specific customer can make the performance of that market segment appear favorable. However, such events may not be sustainable. Therefore, before making any changes based on good or bad signals in a market segment, you should check the time series of that segment to ensure those signals are consistent.
The definition of consistency depends on the size of ad spend and the timing of the year. For example, if you are spending a small amount, make sure a trend is consistent for one or two weeks because performance can be volatile from day to day. Conversely, if you have high ad spend, a few days of ad performance is enough to tell the story.
Additionally, you should be cautious about promotional periods when evaluating trends. Many ads perform well during promotions, but cannot yield much return during regular times. Analyzing ad performance during non-promotional periods will provide an accurate picture of each market segment.
If you are experiencing low returns from your advertising and feel like you have tested everything possible, don’t get frustrated. Instead, try to evaluate things from a data science perspective. We have worked with numerous brands that are in the same situation, and we were able to quickly identify viable paths to profitable advertising for them through data.
In my next articles, I will take a deep dive into each approach and share more takeaways from real-world case studies. Please stay tuned!