Get Started

Make On-Target Suggestions to Customers With Product Recommendations

Use Cases & Projects, Dataiku Product, Scaling AI Benoit Rojare

Even though we entered the supermarket with the sole objective of buying detergent for the next laundry, we came out of the store with softener, anti-discoloration wipes, and … Tic Tacs? That happens more often than we want to admit and is the result of a thought-through buying suggestion strategy. In a supermarket, the layout of the aisles, shelves, and different front display is the only way — a passive way — to recommend products to consumers. Only smaller shops with the personal attention of sales people can make a real difference in recommending products. 

supermarket

The era of online shopping brought a major change. By reconciling  the consumer activity across different touchpoints (website, social media, emailing, SMS, offline, etc.), personalization can finally take place. And personalization pays off: For 95% of companies that invest in it,, they get 3x the result of this investment within a year, according to a 2021 report from SAP. It is no surprise that recommender systems are now part of the landscape for retailers and CPG companies that want to strengthen the relationship with their customers by pushing the right content to the right customer at the right time and through the right channel. 

But how does such a system work and where do they fit into the customer journey?

Making Recommendations With Collaborative Filtering

To know what a consumer might be interested in buying other than what is already in her or his basket, there are different methods. All of these methods are basically filters trying to infer from similar past sales: similar because of the product, consumer, etc. For instance, a recommender system based on the past purchasing behavior of a certain type of consumer will base its filters on the attributes of this known group of customer: recency of the last purchase? average basket size? From these elements, the recommender system will come up with a list of products that this group of customers is likely to buy. This type of recommender system is called collaborative filtering — it is very efficient, but numerous other types of systems exist. 

Once the digital marketer ends up with a list of products to recommend, it will then be her or his choice to place these recommendations in the website in a blunt or subtle manner: It could be a certain section of search results or suggestions on the check out page. We can also find recommendations outside the websites, on email outreach or tracked ads that appear on other websites.

This question of efficiently creating a robust recommender system was asked countless times by Dataiku users, which is why the Business Solution team created a pre-packaged project to quickly start things off on the topic. Let’s look at what you might find in this solution exactly, but first a reminder of what a Business Solution actually is.

How Can Dataiku's Business Solutions Help You Reach Full Potential?

Business Solutions are Dataiku add-ons accelerating the way to achieve advanced or foundational industry-specific use cases within your organization. They are an operational shortcut to achieve real-world business value. Taking advantage of Dataiku’s core features, they are built to be fully customizable and entirely editable.

They come with:

  • A user-friendly interface that enables fine tuning to match specific business requirements
  • Ready-to-use dashboards that can be customized
  • Documentation and training materials

Dataiku industry specialists develop solutions for every vertical, among which:

As a result, business professionals experience a boost in AI productivity and can rationalize their resources.

How Does It Work in Practice?

The Product Recommendation solution provides a reusable project wireframe to accelerate the development of analytics tailored to your data and business structure.

Learn more abut Dataiku's Product Recommendations solution in the video above.

With this solution, marketing analysts, CRM specialists, digital media managers, or marketing directors can access:

  • A full end-to-end pipeline
  • Collaborative filtering plugin
  • Machine learning models
  • Ready-to-use dashboards to consume insights

From a user perspective, the solution is made of the following easy-to-use components:

1. Ingestion of the Data: Sales History and Items

2. Pre-Processing of the Sales History Data

Split customers into categories depending on the number of past purchases or interactions with the brand.

RFM score distribution historic client interactions

3. Collaborative Filtering

Use all sales data to learn relationships across the entire customer base and apply that knowledge to your segment of choice (e.g., the growth customers).

auto collaborative filtering

4. Model Training

Enrich the output dataset from part with customer and item data. Train different models and select a classifier to predict the target (we use a random forest in our case).

model training

5. Recommendation

For each “user + item” pair of the customers with growth potential, predict the target.

customer recommendations

Start implementing your Product Recommendation system right now, with these simple requirements:

  • Historic data on the past transactions
  • Dataiku version: 9.0 or later

You May Also Like

Mitigating Data Bias by Implementing Responsible AI Practices

Read More

Graph Neural Networks: Graph Classification (Part III)

Read More

Smart Manufacturing: How We’ll Get There and How We Won’t

Read More

Data Transformation: How to Transform Data More Efficiently

Read More