A Complete guide to data-driven customer acquisition
Over the past ten years, I have been fortunate to work with over 100 entrepreneurs. Regardless of their geographic location, product, financial situation, or industry experience, they all have one thing in common: customer acquisition keeps them up at night. To fuel customer acquisition, companies spend relentless efforts to build their marketing engine. Great marketing brings in traffic, improves conversion rate, and reduces costs.
One way to optimize customer acquisition is through precise marketing, where companies only invest in their target customers. Traditionally, these efforts rely heavily on industry and marketing experience. However, forming a team with both kinds of expertise is challenging. As you can imagine, industry experts focus on their markets and products and are often not marketing experts. On the other hand, marketing talents that happen to understand a specific market are rare and often expensive. The situation is even more difficult for startups defining new product categories.
This is where data scientists come in to help with marketing. Companies can acquire marketing experiences unique to their vertical by regularly launching experiments, monitoring performance in real-time, and quickly iterating based on market feedback. Moreover, data science takes care of high-dimensional marketing data better than manual efforts and brings richer insights to serve business decisions. In this article, I will discuss low-lift data analytics approaches proven to improve marketing decisions and how to apply them step by step.
Choose target customers
As the old saying goes, making a choice is more influential than spending efforts. When companies have limited resources, which is always the case, spreading too thin across all initiatives leads to loosing an edge in the fierce market competition. Investing in the right customers provides more possibilities for success.
As discussed in Double Down on the Most Valuable Customers, customer value can be calculated as follows:
A customer’s long-term value = Revenue from the customer’s repeated purchase — its acquisition cost
For example, a luxury jewelry brand’s customers don’t make frequent purchases. In this case, it can set the customer value calculation timeframe to one year. Based on its sales and marketing attribution systems, the brand can calculate one-year revenue and acquisition costs for each customer (or visitor that doesn’t convert). With that, the brand quickly gets each customer’s value and decides who is worth its marketing budget.
In most companies, customer value distributions look like the graph below. Most of the customers bring low to negative value to the companies, and therefore should be ignored or discouraged. The split between the left and right in the graph below generally follows the 80/20 rule.
Identify campaigns that work for target customers
The nature of business is an experiment, which means that most of the activities in a business are doomed to not yield positive results. In a startup, as much as 95% of the work is ineffective. Companies must quickly isolate what’s working among the noises to succeed. By closely monitoring marketing performance in real-time, companies have the best chance to distinguish the good signal vs. the bad signal immediately after they emerge and take action upon them.
Based on the customer value analysis, the jewelry brand decides to roll out campaigns toward the high-value cohort. The campaigns run across ad platforms like Google, Facebook, TikTok, etc. The brand implements everything based on hypotheses — it considers customers like something, but it doesn’t know for sure. Only when the brand collects and evaluates campaign performance data can it tell which campaigns are worth investment vs. not.
While implications from the graph above seem obvious, most companies miss the signals. Because of that, when companies start to pay attention, they will gain tremendous growth opportunities by simply cutting back budgets on marketing initiatives that are not working.
Uncover why a marketing initiative works vs. not
Companies get an outsized return in the short term by doubling down on the marketing initiative that works. By replicating these successful experiences in new initiatives, companies have the edge over their competitors.
However, when the external environment changes, copying and pasting previous experience will likely not generate expected returns. Therefore, companies need to uncover why and why not an initiative is effective. So that when market and time change, companies can better determine which part of the old tricks still apply vs. not.
Manually estimating the why and why not behind a marketing initiative can be challenging. There are too many dimensions to analyze: customers’ demographics, touchpoint interactions, visiting behaviors, purchasing behaviors, and more.
Customer segmentation is a valuable technique that simplifies the above analysis. With the granular information provided by customer segmentation, companies can identify the most attractive campaigns, shopping experiences, and products for each customer cohort and review why they are attractive.
The jewelry brand tests its campaign strategies during Valentine’s Day. It gets initial success. When the next big holiday, Mother’s Day, approaches, the brand wants to understand whether Valentine’s Day experience applies. Therefore, it runs customer segmentation to analyze what factors contribute to Valentine’s Day’s success. From the analysis, the brand discovers that only a subset of Valentine’s Day’s target customers overlaps with Mother’s Day ones. And it can likely deploy the campaign strategy that works for the subset in Mother’s Day promotion.
Test and learn
After discovering what’s working through experiments, businesses can look to apply the success recipe when new customer acquisition opportunities emerge. The rich information from customer segmentation helps companies get one step ahead in business iterations.
Before implementing the previous experience at scale, companies must go through a test-and-learn process for a sample and validate whether the experience is applicable for the new opportunity. Some common approaches for test-and-learn include A/B testing and visualization.
Prior to the big promotions for Mother’s Day, the jewelry brand launches a small-scale campaign with all the lessons from Valentine’s Day. Most lessons are still valid, but creatives and messaging can have some tweaks. The brand modifies its campaign and tests at a larger scale. It goes through the same process a few times daily until it achieves satisfactory results and rolls out campaigns at full scale.
Personalization
Contrary to common sense, personalization is a one-to-many marketing mechanism for most businesses. Once companies find out how to attract each customer segment in terms of marketing channels, campaigns, creatives, messaging, etc., they are ready to roll out personalization in their customer acquisition initiatives. For example, a company can pre-define its website, browsing experience, and product recommendation for each customer segment. When a visitor fitting a specific segment comes onto a website, the company can automatically present the right shopping experience to them.
After running different promotions for a year, the jewelry brand has collected enough data for various customers. The brand understands what and how each customer cohort likes to shop. Therefore, the brand designs five different customer experiences to fit the most target customers. Presented the desired shopping experience, target customers convert faster and buy more than before.
Data science behind the scene
Two sets of data science technologies are critical in enabling the above analysis. The first includes data engineering technologies that can connect the dots along the user journey. It’s not uncommon to hear that the data pipeline infrastructure is the biggest blocker for data science projects. Thanks to data industry development, there are more and more connector tools that streamline the process of getting data from multiple sources.
The second key piece is identity technology. Due to privacy changes in iOS, it’s becoming more and more challenging to associate different touchpoints along the user journey without unique user identifiers. Luckily, marketing technologies independent of personal identifiers like marketing mix modeling provide an alternative.
I discuss how to use data science to level up your business and optimize your marketing in my articles. If you want to discuss customer acquisition or other data science topics, please follow me on LinkedIn or contact me at newsletter@ivyliu.io. Until next time.
