Interfaces need good defaults

Open an analytics page and the first thing it asks you to do is choose a date range. Until then, the chart is empty.

That can be a reasonable decision if the range is essential to the task. Often, though, the product already has enough context to show something useful. It's asking the user to make a decision the team left unresolved.

I look at these first states when an interface feels like more work than it should.

Choose a starting point you can explain

“Last 30 days” is a plausible starting range for an activity report. It isn't automatically right for payroll, a seasonal business, or a report used to reconcile yesterday's transactions.

A default needs a reason grounded in the task. So does the sort order. Recent activity helps someone resume work; due date helps someone decide what needs attention. Showing every item can hide the few that matter now.

The selected state should be visible and easy to change. A useful default loses much of its value if people don't realize a filter is active. In a report, show the date range near the result and include it in an export.

Remember deliberate choices

If I choose a project and return ten minutes later, I'd usually like the product to remember it. If I select a one-off filter to investigate an error, carrying that filter into next week's session may be confusing.

Those are different kinds of state. It helps to decide which choices belong to the current view, which belong to the account, and which should be shareable in a URL.

Remembering an explicit preference is also easier to explain than guessing from behavior. An export format I selected is a good candidate. Silently moving navigation items because I used them more often is much harder to trust.

Where a saved state can hide information, give people a clear way to reset it.

Empty results need a different response from a new account

A new media library has no uploads. A populated library filtered to the wrong date also has no visible uploads. The screen shouldn't give both cases the same message.

For a new library, an upload action may be enough. For filtered results, show which filters are active and offer a way to clear them. Don't encourage someone to upload a duplicate because the existing file is hidden.

The action can carry most of the explanation. There's no need for a paragraph welcoming someone to an empty list.

A blank prompt still asks for product knowledge

“Ask anything” assumes people know what an AI feature can access and do.

A more useful starting point might offer a task tied to the current document or selection: summarize this report, compare these proposals, draft a reply to this message. Show the context that will be included, especially when it contains private material.

Suggestions should remain suggestions. Choosing a starter prompt is different from sending customer data to a service or publishing generated content. Those actions need an explicit decision.

Defaults can also do harm

Convenience isn't a sufficient reason to preselect a paid add-on, consent to tracking, or make a private document public. A safe starting state matters more than saving a click.

I review consequential defaults separately from convenience settings. What happens if someone accepts this without noticing? Can they undo it? Will the choice expose data or create a charge?

For the rest, a useful test is to open the product as a new user and then as a returning one. Before changing the layout, check whether each person starts with the information and controls their next task needs.