Argo’s recent post on the topic of personalization and customization is excellent: https://uxplanet.org/personalizations-identity-crisis-35473201ad44 as it deals with using commonsense more than an overhyped AI/ML engine.
Key takeaways are:
- Collage patterns with content carousels with justification: “new this week, because you watched, frequently bought together, similar items to compare”
- Suite of patterns based upon the themes above that are centered around exploration behaviors in a “because you use Heroku” … and enable “explore/view all” … and know what other similar organizations are doing …
- Augmented intelligence with intelligently curated content to surface “key collections” and “for you” and “featured”
- Self-Guided by asking how you’re doing “did this solve your problem?” or “find the solution that meets your need?” or “what do you want to do today?” or “let us help you find what you’re looking for”
- Community engagement by enabling the crowd to weigh in on faetures and ideas
- Multi-party engages not just an individual but a whole event (like a Dreamforce)
- Grow with the customer: Sequential learning via trailhead or the likes
- Orchestration (wrong word?) — let folks understand what data you’re using to customize to them
- Self-actualization is the act of letting them “share an award” for what they’ve gotten good at within the community they care about
Update:
- This is a good piece on when you do have a personalization engine running: https://uxdesign.cc/mobile-app-personalization-when-ui-meets-ai-to-elevate-cx-d39cb50cc557
- AWS has a service in production that’s interesting called Personalize: https://towardsdatascience.com/build-a-recommender-system-in-less-than-an-hour-using-amazon-personalize-68bee9931c60
- Useful how-to is here: https://realpython.com/build-recommendation-engine-collaborative-filtering/
- KNN Package for Dart is here: https://pub.dev/packages/ml_algo and TensorFlow (lite) for Flutter example is here: https://pub.dev/packages/tflite_flutter and at the simplest level there’s a k-means algo here: https://pub.dev/packages/kmeans and there’s Firebase ML of course: https://pub.dev/packages/firebase_ml_custom (example here)
More low-tech:
- Weighted average systems seem fine for simple calcs.
- Further example here.
- 2008 paper here.
- This Medium post is v good: https://towardsdatascience.com/intro-to-recommender-system-collaborative-filtering-64a238194a26
