Increase your ecommerce ROI by predicting users’ next shopping stages

There’s a special type of satisfaction we get when things fall into place. We have a sense of calm, of things being done well, and, if the stakes are high enough, our hope in humanity may be restored.

Think about how it feels to:

• Find what you need on your first Google search
• Spot the jacket you’ve been eyeing on sale (and in your size!)
• Receive a delivery earlier than expected

As customers, we value these types of experiences. We judge and evaluate the brands and companies we buy from based on our satisfaction with the transaction. But when we switch roles,  from consumers to product builders, we may get caught up in our burning priorities and unwittingly forget the importance of creating a great customer experience. However, our customers never forget and our KPIs clearly reflect their satisfaction levels.

Understanding your customers’ needs can be an ace up your sleeve that helps you meet their expectations (and those of your manager).

To make things even more exciting, you get to work with technology that only the biggest companies in the world could afford until recently. You can leverage machine learning and AI-powered optimizations to take your marketing performance to new heights.

Make subtle improvements for a big impact

According to Jon Bird, CEO of a leading marketing communications agency and Forbes contributor, ecommerce predictions for the next phase of marketing technology converge around a common idea: shopping “will be more digital, but will feel more human and technology will be less visible — but far more empowering”.

Using AI doesn’t make a difference for your customers unless their experience is better, faster, more relevant, and more gratifying. Being able to predict their next shopping stage can help you make it easier for them to convert.

You can make it easier for online shoppers to convert by giving them:

• Customized offers
• Personalized discounts
• More-relevant recommendations for matching products
• Targeted notifications

What’s more, you can get better ROI for your marketing budget by segmenting and targeting users who are more likely to purchase through your remarketing campaigns. Here’s how AI is making this entire process easier, faster, and more effective.

Engage segments of users you never even knew wanted to buy

The idea of building customized experiences for online shoppers isn’t new. More often than not, this type of personalization translates into segmentation. This involves identifying common characteristics among customers and then targeting them based on predefined user segments.

Traditionally, marketing specialists have relied on creating a fictional customer profile to define an ideal user who would be most likely to convert. They’ve based their assumptions on criteria drawn from research and analytics, such as:

• Age
• Gender
• Location
• Device used

This approach has become standard; companies  tend to use similar tactics. The standard marketing toolbox includes:

• Online ads
• Email notification to prompt purchase finalization for items they’ve added to their cart or marked as favorites
• Discount rates
• Rule-based calls to action

Besides high competition for customers’ attention, this approach has some other, less obvious disadvantages.

For one thing, it completely misses outliers: users who are equally committed to buying but don’t check the same boxes. These “ghost users” can become a source of revenue for your business if you can find a way to effectively reach them.

Customer segmentation has other downsides as well: those who fit the same profile (age, gender, location) will receive the same deal or call to action irrespective of where they are in the buying journey: product page, add to cart, checkout, etc.

Putting all users in the same bucket, regardless of their intent, represents a shortcoming that can lead to lost opportunities, wasted money, and sometimes even  frustrated customers who give up their loyalty or users who reduce their chances of becoming loyal customers. 

Thankfully, AI can help with that and our proposed approach to segmentation is more nuanced than a standard marketing approach. Instead of making assumptions about customers’ personal characteristics, we rely on machine learning to analyze their browsing activity and history. We can still compare users to each other to identify patterns among them, but we now also calculate their probability of converting, at an individual level. This results in a new criterion for targeting users: converters vs. non-converters.

Here’s how this might work.

Spot users who are most likely to convert

If you’ve been working in ecommerce for a while, you’re likely familiar with what Google Analytics offers. You can generate activity reports or segment users to better understand who is using your website and why. In addition, in the Enhanced Ecommerce module, you can find an overview of your sales funnel. The funnel, presented in the Shopping Behaviour section of your GA dashboard, gives you a high-level view. However, it doesn’t provide any qualitative information about why some users convert and some don’t.

You can certainly dig deeper in each report. For example, the Audience > User Explorer report is particularly useful for analyzing the behavior of individual users. GA identifies each user by browser cookie or user ID, which is available if your ecommerce application has user accounts. But no matter how closely you examine your GA data, the mystery remains: How can marketers identify users who are close to converting?

Use customer contexts to predict what they’re going to do next

Our assumption was that users’ activity and history are the most relevant indicators of their conversion probability.

The learning algorithm predicts the outcome of a user’s browsing session based on information from previous sessions by plugging directly into the data source which can be Google Analytics free or 360, BigQuery or any other 3rd party analytics platform.

Going beyond splitting users into converters and non-converters, AI/ML can help you achieve much more specific targeting. This gives you the ability to engage users based on the shopping stage they’re most likely about to enter.

You can automatically calculate the probability of the customer’s: 

• Continuing to browse products
• Adding a product to the cart
• Checking out
• Finalizing the transaction

By comparing these probabilities, the ML model can help you identify which shopping stage the customer is likely to enter. As a result, you can correctly place the user in the corresponding segment (for instance, those who are  likely to add a product to their shopping cart).

What happens behind the scenes

In the background, we can divide the data extracted from the Google Analytics API into three categories:

• User/browser: Information about the user (identified by cookie ID); includes details such as browser and mobile device
• Sessions: Information such as shopping stage, session duration, number of transactions, revenue, and days between sessions
• Hits: Details (e.g., price, name, category) about products browsed, added to the cart, processed through checkout, or purchased , as well as about products and other click events from Google Analytics.

Users have a variable number of sessions and sessions have a variable number of hits.

Our model can have a recurrent architecture: it uses the data in a temporal sequence. The session and hit data is organized from oldest to newest and fed into the AI model one session at a time. After processing all the sessions for a particular user, the model calculates the probability that each shopping stage will occur in a future session. 

So far, we have conducted lengthy experiments with two attribution models:

• Linear attribution modeling, in which a transaction is equally important for all sessions, regardless of the time that has passed between the first and last sessions
• Time decaying attribution modeling, which places more value on the most recent sessions by applying a half-life weight based on the age of the session

A real-world AI experiment to predict the next shopping stage

To show you how this practical AI application works in real life, we created an experiment based on data from a furniture ecommerce website. Our data came from 24,188 online shoppers. Of these, 98 performed a transaction during their last session. Total revenue was $151,674, with an average transaction value of $1,547.

In a real-world situation, each targeted user would have a customer acquisition cost. For this experiment, we assumed it to be $50. We didn’t take into account the gross margin for the calculations below.

The point of the experiment was to evaluate the efficiency of different targeting methods:

• Two of them based on the machine learning model (with both attribution models) described above
• One random
• One using a statistical method

For each scenario, we calculated the percentage of true positives (correctly targeted users) and false positives (incorrectly targeted users).

Based on each user and their history, we calculated a transaction probability between 0 and 1. If the user was targeted and made a purchase, we had a true positive. If the user was targeted but didn’t finalize the transaction, we had a false positive.

Here is what the results looked like:

Method	Targeted users who made a purchase (true positive)	Targeted users who didn’t make a purchase (false positive)	Total projected revenue	Targeting costs at $50/user	Revenue – costs
Random	90.84	11,949.12	$102,456.14	$601,998	-$499,541.86
Statistical	9.99	525.12	$8,508.40	$26,755.5	-$18,247.10
ML, time decay attribution	20.02	135.72	$21,842.74	$7,787	$14,055.74
ML, linear attribution	34.04	98.93	$39,524.48	$6,648.5	$32,875.98

Using a random scheme, we got the most true positives (correctly targeted users), but the high cost meant that profit was not on the books. By comparison, the linear model, while targeting a smaller group of users, maintained a low percentage of false positives (0.41%). From these projections, it’s clear that linear attribution was the most profitable targeting method.

Using shopping-stage predictions to drive ecommerce performance

Once you have your next-shopping-stage predictions, there’s a lot you can do to optimize your ecommerce website and marketing flows. You can become a more relevant and helpful shopping destination for your customers by:

• Improving search results by segmenting audiences based on intent
• Filtering recommendations based on intents
• Displaying pop-ups with personalized calls to actions
• Sending personalized email alerts
• Remarketing to users who are more likely to convert 

Using an AI-as-a-service platform means you can save time that you’d have otherwise spent poring over Google Analytics reports. Rather than be stuck in reporting meetings, your team can get aligned faster and put your insights to good use.

What’s more, predictions can be provided via an API that you can easily integrate with your reporting dashboard. Additionally, having an infrastructure that fully automates the model training and predictions is the scalable solution at hand. Here’s how integration works in an ecommerce scenario:

• User goes to your ecommerce website
• Website identifies user by browser cookie and calls API
• Probability of completing a transaction for the user is returned
• Based on the transaction probability, the website displays a pop-up with a personalized message that could include a call to action, discount code, or creative trigger

There is no doubt that AI/ML is a powerful driver for creating better customer experiences. To get there, you just need an open mind and willingness to experiment with new ways of looking at your existing data.