Personalization is easy to describe and difficult to trust.
A feed notices behavior, predicts what might be relevant, and changes what appears next. When it works, browsing feels lighter. When it fails, the same mistake can repeat for days: one accidental click becomes a supposed obsession, a gift search becomes a permanent identity, or “not interested” seems to have no effect.
The difference is not only recommendation quality. It is control.
People should not have to reverse-engineer the feed
A 2019 Pew Research Center report on Americans and privacy found widespread concern about how personal data is used and a common feeling of having little control over it. The survey is U.S.-specific and does not describe every user, but its central concern remains relevant: personalization without understandable control creates uncertainty.
A shopping feed can respond by making its inputs simpler and more legible.
Useful personalization should feel steerable, not inevitable.
“More like this” is only half the conversation
Many shopping experiences are good at noticing what people like. The harder question is what to do with dislikes and short-term interests.
Amatok currently offers a visual feed shaped by products people are discovering. It records a small set of actions, but those actions do not reorder the feed yet.
- Like means this feels relevant and requires sign-in.
- Save means keep it available for later; saving can start on-device before you sign in, then merge once an account is connected.
- Minimized views offer a small hint about what held your attention.
- Skip is a small hint that something was not right for you.
- Explicit not interested is a planned control.
- Recent activity may help make future recommendations more useful.
The current feed records these actions without changing its order around them.
Preferences have a time dimension
A future personalized experience may need to distinguish three kinds of intent:
- long-term positive preferences: desk lamps or other styles that repeatedly appear over time.
- explicit negative preferences (if introduced): clear choices about things the shopper does not want to see.
- recent intent: needs that shift with a move, changing hobby, or gift planning context.
These are distinctions to test, not current behavior.
Today, these actions do not all change what appears next. The standard we are aiming for is straightforward: when they meaningfully affect recommendations, people should be able to understand and influence how much they matter.
Privacy is also about restraint
Amatok’s current goal is to keep a small set of discovery inputs useful.
That implies product decisions such as:
- keeping controls close to the feed;
- avoiding unnecessary profile questions before discovery starts;
- evaluating explicit negative feedback as a possible future control;
- separating recommendation relevance from potential affiliate value;
- evaluating whether visible correction and reset options are needed as the experience evolves.
How this differs from passive recommendation
A passive feed says, “We inferred this, so we will show more.”
A steerable feed says, “Here is a candidate. Your response helps determine what happens next.”
That direction can help shoppers spend less time fighting irrelevant recommendations and more time noticing products that fit the current situation.
If you want clearer control paths to appear over time—especially explicit negative feedback and correction options—join Amatok early access.
Sources and further reading
- Pew Research Center, “Americans and Privacy: Concerned, Confused and Feeling Lack of Control Over Their Personal Information”, November 15, 2019.