Originally Published 2016
I originally wrote this piece for the Quora design team’s recruiting blog. There’s a lot more I’d love to add to this because it is now 6 years old, but there’s still interesting content here.
Let’s say you’re an app that wants a compelling, sticky homepage to bring your users back to everyday. Your content could be anything — photos from friends, a catalog of movies or an ecommerce site with millions of items.
You want to do your content justice, but there’s too much to show at once. So what do you do? For example, what if Amazon just put items as they came in on top of their home page?
It's surprising how hard of a design challenge picking and reordering content can be. As the example illustrates, one size does not fit all. Twitter orders its feed by time, with a small ranked section. Facebook preserves recency, but some life stories and photos are more important than links. Netflix doesn't care about recency but buckets into genres.
So, which approach is the right one for you? We think about this a lot at Quora, so here's a few questions you can ask yourself to narrow down your options.
1. How important is recency / timeliness to your content?
Sometimes you can't afford to reorder content in a non-chronological order. Think about how this applies to your pool of content. If you're a Amazon or Netflix, time hardly matters, but if you're Twitter, this is critical.
It's important to think of this constraint first. The less you're bound to time, the more flexibility you have in future design decisions.
2. How much better is your best content than your worst content?
Think about all the content you would consider showing a user. This could be stories from a few people they follow or a large library of things they can shop.
Oftentimes, the best and worst content in the pool are not that different in how engaging they are. For example, Instagram is about visual stories. My photos are all about the same quality -- somewhat decent photos of food for example. For someone who follows me, my posts are homogenous in their quality.
When the pool of candidate content is homogenous, you don't need it reordered with a Machine Learning (ML) algorithm. But if the best (think wedding photos) is much better than the worst (reshared posts), then ML is a godsend.