The Feed Default Is a Product Decision

Two ends of the feed

In my May 2, 2023 First 1000 issue, TikTok’s stranger feed and Twitter’s For You direction pointed to a product question: feed design has polar choices, and the middle can become an uncommitted version of both.

TikTok made the stranger feed feel normal. In that snapshot, Twitter had also moved toward a For You page as the default way to interact with the service.

TikTok then tested the opposite direction. The first test was a topic-specific feed. The control arm won, so the topic feed was not launched. The second test was a Friends feed containing posts from people who mutually follow one another. At publication, the Friends test was still running.

The graph changes the activation problem

A stranger feed can create value before the user has friends. That makes it attractive for a new social product. The product can show something immediately, even when the user’s graph is empty.

The tradeoff is trust. The user needs to learn why a post appeared, why this creator matters, and whether the feed understands the reason they came. Discovery can feel magical when it works and arbitrary when it does not.

A friends-only feed can feel personal and legible, but only after the user has people to follow and those people have posted. A founder choosing this default is choosing a graph-building problem before a ranking problem.

A small community with strong shared context may get more value from a few real relationships than from an infinite catalogue of strangers. Decide which constraint the product wants to solve first.

The topic feed is a third product

Topic feeds look like a compromise. They promise discovery without the full randomness of an algorithmic home feed. The topic-feed test is a useful warning because the control arm won, and the topic-specific feed never launched.

A topic filter needs a job beyond making the feed feel organized. It might help a user complete a task, follow an event, or browse a bounded interest. If it only removes variety, the general feed may already be doing the job better.

Before adding a topic feed, ask what decision it improves. Does the user want to find a person, a type of content, a local event, or a way to control the ranking system? “Topics” by itself is a label, not a product strategy.

Choose the default with one testable promise

Write the promise above the feed before you build the tabs: discovery, relationship, or control.

Then make the first session test that promise. For discovery, measure whether a new user finds and saves something they did not know to search for. For relationships, measure whether the user finds a person worth returning to and whether the graph produces new activity. For a topic feed, measure whether the filter helps the user complete a declared task rather than merely changing the visual layout.

My recommendation is not to copy the latest feed surface without naming the value it is supposed to create. TikTok’s stranger feed and friends-only experiment point in opposite directions because the products are solving opposite problems.

A feed-default checklist

• Write the promise above the feed.
• Name the graph or ranking constraint you are choosing.
• Give a new user a first-session test of that promise.
• Track the product-specific action, not a generic scroll count.
• Pick the default before building the tabs.

Sources and original research

• Original First 1000 issue: Friends vs Stranger Feeds, May 2, 2023

Historical facts are attributed to these sources. An observed product change does not establish conversion lift.