Artificial intelligence for marketing
This article is a guide for using artificial intelligence for customer acquisition.
The times of one-size fits all marketing are over. Your company needs to understand which people it should target in your acquisition campaigns based on their value to your company and their probability to buy. With your data and custom built machine learning models we create artificial intelligence solutions that deliver new data points into your CRM which lets your marketing team optimize the targeting of your acquisition campaigns. Ultimately, this will lead to an increase of your marketing ROI.
We (data scientists, data engineers, machine learning engineers and AI marketing experts) have the skills, the computing power and the experience to bring your marketing to the next level.
Questions that I will answer in this post:
- How can I attract more of the customers who generate the highest return?
- What is the maximum amount I should spend per individual customer in new customer acquisition to maximize profit?
- How likely is it that anonymous visitor on our website to buy our product or service?
- Which of my website visitors should I target with a remarketing campaign and which not?
How can I attract more of the customers who generate the highest return?
We can build an artificial intelligence for you that uses your customer data to predict the Customer Lifetime Value (CLV) of each individual customer you have. From this algorithm we will learn which variables or characteristics of your customers make them valuable for you. Your marketing team can use this information in your customer acquisition campaigns to target those people with the same characteristics as your most valuable customers. While raising the return on investment (ROI) of your acquisition campaigns, this will increase the amount of customers with a high CLV in your customer base.
Use Case: Artificial intelligence for customer acquisition
A day in the life (before)
Scene or situation:
Susanne B is a performance marketing manager. She has received the task to optimize the targeting for a new customer acquisition campaign to improve the marketing ROI compared to past campaigns.
Desired outcome:
Find the relevant variables that can be used for targeting to attract customers with a high lifetime value while controlling the costs to acquire these customers.
Attempted approach:
Susanne has been gathering data for days, going to all her most reliable sources. She collected conversion data from past campaigns out of different paid media tools and requested data on the customers that spend most from the data team. After two weeks she receives a comprehensive report. She now digs through the different diagrams and tables and finds some variables that seem to correlate with a high spending. She uses these variables as targeting criteria in the next paid media campaign.
Interfering factors:
The problem with the whole approach is twofold. First, both Susanne and the data team used historical data, but every customer is on a different point in the customer lifecycle. The fact that a customer hasn’t spent much with the company yet doesn’t necessarily mean this customer has a low customer lifetime value. If the customer is in an early stage of the customer lifecycle or stays long with the company, most of the transactions will occur in the future. The second problem is that Susanne has only optimized the campaign for the variables that relate to high spending, but not only your highest spending customers are valuable to your company. If you can acquire a customer with a lower customer lifetime value while spending considerably less in customer acquisition, this customer could generate more profit for you than the customer with a high customer lifetime value.
Economic consequences:
The new campaign generates new customers that spend a lot, but as the targeting has become very narrow, focusing all the budget only on those people which fit the targeting criteria of the highest spending customers, the customer acquisition costs have increased dramatically for this campaign. Even though more high spending customers have been acquired, the marketing ROI of this campaign hasn’t increased much compared to the average ROI of other campaigns.
A day in the life (after)
New approach:
As the company’s data team’s resources are scarce and nobody in the data team has experience with lifetime value modeling, Susanne and her Head of Marketing have partnered with aaimo to calculate a reliable customer lifetime value for each customer. Now, Susanne can use the predicted customer lifetime value to split the customer base into value segments. She creates one ad campaign for each value segment, using the distinct customer segments characteristics from the company’s own database for targeting. By utilizing the customer lifetime value of each segment, she can calculate and maximize the KPI Return On Advertising Spend (ROAS). At the end of the campaign she reports the ROAS to her Head of Marketing. Every subsequent campaign’s performance can be compared using the ROAS.
Enabling factors: aaimo creates a new data field in the company’s CRM that Susanne can use to easily segment the customer base by customer lifetime value. She doesn’t need to request any analysis from the data team as this new data field allows her to intuitively understand the value the company gets from acquiring this specific customer. The new data field in the CRM allows an exceptional feat to Susanne: Now she can easily upload the customer data from the value segments to the different ad tools to create lookalike audiences. The ad tools use lookalike modeling to automatically create the optimal targeting setting for each value segment. The time Susanne would have needed to create the targeting for her campaign is now available for other tasks.
Economic rewards:
Susanne’s Head of Marketing is thrilled with the results of the campaigns. The ROAS of the campaign for each value segment is a high multiplier as the optimization algorithm of the ad tool never spends more to acquire a new customer than the customer is worth to the company. Furthermore, the Head of Marketing can use the ROAS to calculate the marketing ROI of the performance marketing campaigns. Based on the now exactly quantifiable marketing ROI, the Head of Marketing decides to distribute a higher budget towards performance marketing for customer acquisition. The Head of Marketing is delighted with the control provided by the new metric to the marketing managers for handling campaigns and reporting.
What is the maximum amount I should spend per individual customer in new customer acquisition to maximize profit?
Once we have calculated the CLV of your individual customers, we can group them into value segments. Each segment will have distinct characteristics which segregates it from other segments. Your marketing team can then use these characteristics and do one acquisition campaign per segment. As we know the CLV of each segment, you can use this value as a maximum bid on your conversions. This information will give your performance marketing team a better understanding of how much they can spend per acquired customer and thereby increase your marketing ROI drastically. Now you can optimize the targeting of your acquisition campaigns to attract new customers based on their predicted future value to your company. Pretty cool right? Plus, we provide this machine learning solution as a full service, which means the predicted CLV is delivered straight into your CRM where your marketing team can utilize it right away.
How likely is that anonymous visitor on our website to buy our product or service?
We can use artificial intelligence on your website and conversion data to predict which visitors will buy from you. With this information you can individualize the customer experience on your website. For example you could send a push notification with a discount in the browser to all users with a low probability to buy. In that way you can trigger some of them to make a purchase although not many or none of them would have made a purchase otherwise. Customers with a high probability to purchase could get a different message like a different discount or a bundling offer. You could furthermore experiment with showing bundlings of different products or a high priced product with high value to the customers with a high probability to purchase, to make a sale with a larger monetary value for you.
Use Case: Artificial intelligence for marketing
A day in the life (before)
Scene or situation:
David T is the CMO of an eCommerce company selling products to customers on their website. David wants to improve the marketing ROI of the company’s overall marketing spend. The website is the single point of sales of the company and all the paid, owned and earned media efforts drive traffic to their website. He thinks that improving the conversion rate of the website by individualizing the customer experience is their best leverage to increase the performance of their whole marketing strategy.
Desired outcome: Identify customer behavior on the website that can be used to trigger individualized experiences, leading to more sign-ups, sales or a higher order value. This will require a real-time scoring of the website visitor based on the observed behavior and a marketing automation system that can change content of the website dynamically or send push-notifications or show pop-ups to the website users.
Attempted approach:
David tasks his marketing manager Peter to create a conversion report of the last 6 months. He specifically briefs Peter to look for similarities between the converting people in the observable data. As 90% of the website traffic are anonymous users, the only observable data available is from Google Analytics (e.g. visited pages, time on specific pages, acquisition channel, session duration, browser language, country and devices). Peter creates a report showing the pages that have been visited most from converting users. David tasks Peter to display a push notification with a discount to each user that visits the pages specified by the report.
Interfering factors:
Single data points don’t predict behavior in a precise way. Some of the people visiting those pages may be influenced by a discount, but most are not. By using this heuristic, David’s company missed the opportunity to show a relevant experience to most of its website visitors. With the fading out of 3rd party cookies, historical data about anonymous visitors won’t be available anymore, reducing the amount of already sparse data even further.
Economic consequences:
By delivering experiences through push notifications, pop-ups or dynamic content on the website that are not relevant or not at the right timing, David’s company wastes the opportunity to influence customer behavior towards a desired outcome. As all the acquisition efforts through paid, owned and earned media are linked to the conversion rate of the website, the best lever to improve the efficiency of all the acquisition channels is not used, resulting in high opportunity costs.
A day in the life (after)
New approach:
David has decided to partner with aaimo to train a machine learning algorithm to predict the probability to buy and the order value for each anonymous website visitor. David’s marketing team uses the output from the artificial intelligence to individualize the website experience for each user. The marketing team creates an individual experience for visitors that have a high chance of purchasing a high order value, by showing them higher priced products. David’s marketing team implements another experience for users with a high probability to buy but with a low predicted order value, showing them a bundling offer. A third experience is designed for users that are predicted to leave without converting. Those users get a push notification with a discount to encourage them to make a purchase or to sign up to the newsletter.
As the newly available scoring system gives a label to each anonymous visitor (e.g. high potential purchasing order value), visitors with a high probability to purchase, that have not purchased yet, can be saved into a remarketing audience. This audience can be shared with Google Ads, Facebook Ads or other ad platforms for campaigning to stay top of mind with those customers for when they are ready for their next purchase.
Enabling factors:
Our artificial intelligence uses all available data to find hidden behavior and data patterns that can only be discovered by machine learning algorithms. As our artificial intelligences learn and improve with new data, a change in customer behavior is incorporated automatically so the experiences built by David’s marketing team always target the right users. As aaimo takes care of the creation, implementation, monitoring and constant retraining of the artificial intelligence, there is no need to hire a new data scientist and data engineer to enable this use case for your company.
Economic rewards:
Three months after the implementation of the machine learning based scoring system David can report more sales, a higher average order value and more sign-ups to the executive board. As the performance of all the acquisition channels is influenced by the optimized website experience, David’s performance marketing team reports lower CPA, a higher ROAS and an improvement of the marketing ROI. David is content with the segmentation of the anonymous website visitors into segments of high and low probability to buy and different predicted order values. He commissions aaimo once more to create an artificial intelligence that predicts the probability to buy for each product category they are offering on their website, so in the future his marketing team can further refine the website experience for their website visitors.
Which of my website visitors should I target with a remarketing campaign?
Remarketing only delivers a strong marketing ROI if it is used on the right targets. We can create an artificial intelligence that uses the behavioral data from your website visitors to predict which of those visitors are good targets for remarketing campaigns. You can then use this information in your backend to score all website visitors by their behavioral patterns and add all visitors above a certain score threshold to a remarketing audience. This audience can be sent to your favorite ad tools (e.g. Google Ads, Facebook Ads, Linkedin Ads) to create ad campaigns to remarket to those visitors with the highest probability of reacting to your remarketing efforts.
We help companies to leverage their data potential for marketing by building and maintaining artificial intelligence. Our solutions enable new opportunities for automation and provide you with important information for targeting decisions.
