Seeded Sample Data
Ship a new account with realistic content already in it so the product can be understood.
The problem it solves
Products that only make sense once populated cannot explain themselves to an empty account. The user has to imagine the value before they can see it.
Some products are legible when empty. A text editor is. A pipeline view, an analytics dashboard, a project board and a CRM are not. Their whole argument is in the arrangement of many items, and one item proves nothing. Seeding puts a working example in front of the user on the first screen.
The judgement call is what to seed. Get this wrong and you have created work, because now the user has to delete your mess before they can start theirs.
Three ways to seed, worst to best
- Lorem content
- Task 1, Task 2, Project Alpha. Teaches the shape of the UI and nothing else. Users delete it on sight and learn nothing.
- Plausible content
- A realistic sample for a generic team, with dependencies, an overdue item and a comment thread. Shows the product doing something, not just holding rows.
- Content from their answers
- Built from the two or three setup questions the user just answered, so the sample uses their industry, their team size and their words. Highest build cost, highest payoff.
Whichever you pick, three rules hold. Label it clearly as a sample. Make removing all of it one action, not a hunt. Never let it mix into real reporting, because a founder who demos the dashboard to their board and finds a fake deal in the pipeline will not forgive it.
Worked example
Portmark, a recruiting pipeline
Portmark seeded five example candidates spread across four stages, one of them stalled for nine days so the stale card warning was visible. Each sample record carried a tinted Sample badge and the workspace header held a single Clear sample data button. In a hypothetical trial cohort: A group of users bucketed by when they joined, tracked over time rather than blended together., the design intent was to make the stalled card the thing people clicked first, since it is the only item on screen that looks like a problem.
- Seed the state that shows the product working, including one item in a warning state.
- Use names and numbers that read as a plausible team, not as placeholder text.
- Exclude sample records from every count, chart, invoice and export.
- Offer one button that removes all of it, and confirm what will be deleted.
- Stop seeding accounts that arrive through an import, since they will have real data in minutes.
The failure that hurts most is silent. A user starts editing the sample records instead of creating their own, treats them as real, and six weeks later their workspace is half fiction. Badge everything, and consider making sample records read only until the user duplicates one.
When it fits
- Products whose value comes from the relationship between many items rather than from one item.
- Trials where evaluation happens before any real data has been imported.
- Complex surfaces like automations or dashboards, where a working example teaches faster than a description.
- Products where the user's own data arrives slowly, such as anything that waits on inbound email or events.
When it backfires
- When sample records leak into totals, charts, exports or billing, which destroys trust in every number on the screen.
- When clearing the sample takes more than one action, turning a helpful gesture into cleanup work.
- When the sample is more polished than anything the user can build, setting an expectation the product cannot meet.
- When the account was created by importing real data, so the seed is clutter on top of the real thing.
Products using it
Notion
New workspaces have long arrived with example pages already in them, including a sample database rather than an empty sidebar.
Trello
New accounts have been given a starter board whose cards describe how lists and cards work, so the board itself is the lesson.
Asana
Project templates create a populated project with example tasks and sections rather than an empty task list.
The psychology under it
The takeaway
If your product is unreadable when empty, ship it non empty, badge the sample, and make removing it one click.
Finished the teardown? Bank it and the day counts toward your run.