Understanding Stampix’ users through analysis of user-generated images

To support Stampix’ journey towards a 1-click photobook experience Stampix and ML6 leveraged machine learning using AWS SageMaker. To kick off, all photos are stored on S3 and then ingested on Sagemaker to anonymise all faces and personal information. Afterwards, ML6 analysed Stampix’ user database, classifying users’ interests and recommending photobooks as a collection of their photos. This interest classification and users’ mapping onto these, allows Stampix to address user preferences more precisely in order to optimise marketing campaigns.

Stampix is a photo printing service that has a clear mission to make people happy through photos. The company focuses on providing a fun and user-friendly experience, ensuring that each individual can easily navigate the process and experience the joy of receiving their printed photos right at their doorstep.

In addition to serving individual consumers, Stampix also collaborates with businesses to help them enhance customer satisfaction and foster loyalty. By seamlessly integrating Stampix's photo products into their loyalty programs or marketing initiatives, businesses are able to make their customers feel valued and establish strong emotional connections.

‍

To be able to deliver increased value for both consumers and businesses, it’s critical that Stampix understands who its users are as well as what they need, expect, and value from its services. These insights can then be used towards the business’ goal of creating a 1-click photobook experience to its users. Stampix first started to collaborate with ML6 to explore how to leverage machine learning techniques to achieve this goal.

In order to propel Stampix on their journey towards a 1-click photobook experience, ML6 and Stampix collaborated to establish guided personalised photobook creation based on pictures that were anonymised through SageMaker. Building this first version in co-creation with Stampix, allows them to learn the underlying concepts and ensures they have all the tools & know-how to continue this work. 

The approach is based on the idea that a computer can learn to recognise and group similar pictures together, without any human help. Once it has learned to do this, it can then bundle similar users together based on these groups of pictures. Each of these groups of people describes a user persona. Storylines are then associated with sets of key sentences, example images and target dates, and are mapped to those user personas. This allows Stampix to automatically suggest storylines that are most likely to be relevant to a given user persona. The user can then select the storylines that they are most interested in. They can then automatically suggest the most relevant resources to help the user achieve each of their selected storylines. In this way, Stampix is able to provide a 1-click photobook experience that is personalised for each user.

‍

This approach can be summarised in the following three steps:

1. Creating smart groups of images by understanding the content of the images.
2. Creating smart user personas based on the image groups they interact with.
3. Compose photobooks with a storyline based on identified user personas using sets of key sentences, sets of example images, and target dates.

‍

‍

Creating smart groups of images

‍

First we create smart groups of images by grouping them together based on what content is recognized in each of the images. For this step we use pre-trained deep learning models to extract the embeddings for each of the user’s photos. 

Next, we annotate the image clusters by embedding popular image captions taken from an existing large dataset containing millions of image captions to the same embedding space as the anonymised image embeddings. We can then calculate the closest caption embedding to each of the clusters’ centres which results in a set of predicted captions for each image cluster.

‍

Creating smart user personas

‍

To create user personas, we analyse users' interactions with the image clusters, grouping similar users together based on their preferences and interests, and identifying the dominant topics of each user persona by looking at the captions associated with the image clusters. More specifically, we create user embeddings based on each user’s interactions with the image clusters. These user embeddings are then once again clustered to group similar users together to form personas. 

‍

‍

These clusters can then identify the personas by looking at how much each image cluster attributes to the average user embedding of each cluster. This way the most dominant topics of each user persona can be identified by the textual content of the captions associated with the image clusters most contributing to that user persona.

‍

Composing photobooks using storylines

‍

By understanding the personas of their users, Stampix can make more informed decisions about which storylines and resources to suggest to each user. For example, a user persona that is interested in travel and nature photography may be more likely to be interested in a photobook with a storyline about a nature hike or a trip to a national park. On the other hand, a user persona that is interested in family and children photography may be more likely to be interested in a photobook with a storyline about a family reunion.

‍

These insights can then be used to generate photobooks using sets of key sentences, sets of example images and target dates describing a storyline aligned with the detected photobook personas. 

‍

This approach allows Stampix not only to automatically generate personalised photobooks for each user without the need for a human editor, but also to understand the interests and preferences of their users on a much more foundational level. This improves the user experience, generates more value for their customers and serves as a groundwork for many future iterations.

‍

We would like to thank Stampix for the opportunity to work with them on this project. ML6 will continue to provide expert support where needed.

‍