Two sides. The front decides whether the thing you have in mind should be an artifact at all — and it will tell you no. The back is what to check before you hand one to anybody.
Almost everything you make is one of these. Most people have never noticed they were choosing.
Answer all three out loud, in your own words. Vague answers are the signal — that's the point of the exercise, not a failure of it.
Fill in the boxes above and this will tell you whether to build it.
| If this is true | Then |
|---|---|
| You'd open it less than once a month | Skip it |
| The answer changes what you do this week | Build it |
| You rebuild the same table by hand on a rhythm | Build it |
| You can't name the decision it drives | Skip it |
| The data is somewhere nothing can reach without a login you don't have | Fix the access first |
Run these before you send a live artifact to anyone. Most of the confusion comes from the first two, and almost nobody expects either of them.
"Live" means it re-asks — not that the answer is newer. The page fetches when you open it, so it's always freshly asked. But it can only show what's actually in the source: if the underlying data is three weeks old, so is your page. And the newest row in any dataset is usually incomplete — a part-month always looks like a collapse. Worth saying before someone panics at a chart.