Scenarios

Loading

    The result at a glance

    Why it moves

    Every metric

    Each row is the change against the same world without this card, over the whole window. Hover any heading or question mark for what it means.

    Each world columnA world that roseA world that fellToo small to tell from simulation noiseThe typical world

    Watch one world play out

    One world with every assumption at its middle value, run twice with the same people: once as things are, once with this card switched on at day 0.

    Day0

    Press Play. The world warms up for 180 simulated days first, which takes a few seconds.

    One world is one draw: a small change can be chance here, which is why the table above uses many worlds. Values are a trailing 7-day window.

    New luck changes the people and their chance events, not the assumptions. Replay uses the same buffered run instantly. Explore different assumptions.

    Run fresh worlds

    Draws worlds the stored batch has not seen (a new random seed each run), each with every assumption sampled once from its range. A world is kept only if its baseline reproduces Pinterest's published figures, so most draws are thrown away. Each kept world is then run with and without this card.

    Nothing running. 24 worlds at 300,000 people takes about three minutes on this machine.

    Is the baseline believable

    Run pinsim once to fill this.