The days of third-party data are over. The cookieless future has settled and users worldwide demand their data be kept private. Forbes reports that companies are now faced with increased pressure to find new ways to track marketing behaviors. Still, many questions remain.
How can companies deliver personalized ads at scale, search for new customers and conquer new markets, launch effective marketing campaigns and optimize products and brand experience? In short, how can businesses connect with their customers?
I will share a customer case example that answers these questions. The essence and approach of this model and technology can be extrapolated and applied to different industries and technologies.
In this case, we used ID-POS data, which bypasses all the challenges of third-party data as an abundant and rich first-party data source. And we push the limits using high-performing feature engineering technology that can successfully manage Big Data data marts, provide visibility for business intelligence users and guide marketing campaigns.
Advertisement
Let’s dive straight into it and unriddle the mystery of creating modern, highly personalized customer experiences while increasing performance in the cookieless future.
Using ID-POS Data to Drive Modern Signal-Based Targeting
A February 2023 Deloitte report found that the right combination of digital transformation actions can generate up to $1.25 trillion in value. But the wrong combinations, and not aligning technology with business strategies, can erode value by over $1.5 trillion.
Businesses that strive to connect with customers share similar pain points. In the case of Lawson, operating a chain of franchise convenience stores with more than 15,000 retail stores in Japan, the company was facing increased complexities in its modernization.
Lawson’s nationwide operations demanded flexible and agile responses to meet customers’ changing demands. Like similar large retail enterprises, Lawson focused its efforts on developing value-based targeting and signal-based targeting. This translates into using purchase history and customer base membership information to drive marketing campaigns. As we all know, in marketing, the main goal is to know the customer.
“By understanding what customers value accurately, we can recommend products that match individual needs, improve our stores, and use customer data to support manufacturers’ product development and promotion activities,” said Mr. Kobayashi of Lawson.
However, despite their efforts, the company was becoming increasingly frustrated when introducing new products — even those of outstanding quality. Lawson also was finding many difficulties in managing its data marts, where the retailer stored customer information that strictly focused on its marketing campaigns. Identifying customer trends and behavior required analyzing data in volume and speed that were beyond the capabilities of their in-house teams.
Companies also need to go beyond typical customer segments such as gender, age or past purchase history if they want to meet customer expectations. The fifth edition of the Salesforce State of the Connected Customer of reveals that 73% of customers expect companies to understand their unique needs, and 88% of users say a company’s experience is as important as its product or services.
But signals-based targeting, when operating at national or international levels, has unique problems. ID-POS data generated in this business model becomes Big Data, making data quality levels difficult to obtain and valuable features hard to identify.
Predictive analytics and machine learning models that efficiently identify customer desires from a vast amount of raw historical purchase data need to be high-performing. Companies need to reimagine how they manage customer data marts, as these can no longer be manually maintained and managed. Therefore, automation today is critical, and visibility and instant accessibility are necessary. But the cost of outsourcing the management of data marts tends to be high.
“Data marts need to be updated continuously for data changes, new data additions, new product trends, etc. While we need to do targeted marketing on more products, outsourcing it to an external data scientist was challenging regarding flexibility and cost,” Mr. Kobayashi told us.
Automated Feature Engineering and Business Intelligence
In machine learning, features are simply a specific type of data that traditionally has been hand-picked by a data scientist for its ability to predict outcomes. But scanning manually through data marts to identify these features is very time-consuming. This is where automated feature engineering comes in. In the case of signal-based targeting, automated feature engineering technology can slash these times from days to just hours or minutes by automatically engineering feature discovery across multiple data tables. dotData’s Feature Engineering capabilities allowed Lawson to create a system that uses ID-POS data to run highly customized campaigns.
Just as important as the ability to automate feature discovery is the ability to rapidly access data insight. IBM lists a lack of in-house skills as a top challenge for companies creating data-driven targeted ads campaigns.
When developing machine learning technologies, it’s essential for the client and all the users to have access to the data and the ability to use it, no matter their technical skill levels. Lawson integrated our solution into a custom-developed user-friendly dashboard, giving department personnel access to the insights without increasing human resources and technical staff.
The ease of use made it possible for marketing people with no knowledge or experience in data analysis to create predictive models by simply entering a few required pieces of information and by selecting parameters.
Data Quality, Automation and Flexibility
The Lawson team analyzed and used ID-POS data to extract features based on customer value. Later that information was used to optimize performance and sales. Using this model, a designer created multiple data-driven coupon campaigns.
Once marketing teams can access customers´ desires, they can build customer groups. The technology is also capable of generating predictive scores. In action, we saw how the purchase rates increased by a factor of 400% in a specific coupon campaign due to improved targeting accuracy based on predicted scores. Additionally, purchase rates increased by 300% thanks to design optimization based on consumer desires — totaling a 1,200% enhancement in performance.
But this type of technology provides insight that can not only be used for marketing campaigns but also used internally to drive decision-making, develop and launch new products, improve store sales and increase the lifetime value (LTV) of members.
Automation is essential for businesses that want to manage data marts without incurring costly outsourcing programs or new hiring costs. When internal human resources are limited, automation and customization can be the solution, if properly applied. Quality of data, features and model performance are also critical. Agility and flexibility, in this case, allow retailers to rapidly pivot and discover new features that align with new strategies without having to develop the entire machine learning process from scratch with each new campaign.
“We started creating this signal-based targeting marketing strategy based on the hypothesis that if we can target customers based on what they value, we will be able to promote the product based on what they will find appealing and increase sales. And this case proved that value-based marketing is more effective than marketing based on age and gender,” Mr. Kobayashi told me.
Going forward, companies should look into the value of ID-POS and other sources of first-party data. Signal-based targeting strategies, enhanced by new technologies, can be used to develop an omnichannel brand and customer experience of the highest level.
“Instead of cookies, which are increasingly becoming less precise and more difficult to retrieve, we believe that purchases in retail stores will be important data points to help us understand customer desires,” Kobayashi shared with us.
From mobile apps to customer journeys, and new or existing physical and virtual stores, retail continues to benefit from the new era of digital acceleration. The more data a company has, the bigger the challenge and risks, but also the higher the reward.
Ryohei Fujimaki is the Founder and CEO of dotData. Before founding dotData he was the youngest research fellow ever in NEC Corporation’s 119-year history; the title was honored for only six individuals among 1000+ researchers. Fujimaki received his Ph.D. from the University of Tokyo in machine learning and artificial intelligence.”
