Printable planning poker cards

Most estimation now happens in a browser, and for a fully remote team that is the only sensible answer. The case for paper is narrower and it is about a specific room: the hybrid one, where four people are round a table and two are on a screen at the end of it.

In that room the people physically present are the ones a tool disadvantages. Everyone has to look down at a laptop to vote, which pulls four people out of the conversation they were having with each other. Cards keep their heads up. They hold a card face down, they turn it over on the count, and the reveal is a thing that happens in the room rather than a state change on six separate screens.

The second reason is unfaked simultaneity. A physical reveal is genuinely simultaneous and visibly so, and nobody has to trust that the software is not showing someone else’s number early. That matters more than it sounds like it should on a team where one person’s estimate carries obvious weight.

The third is workshops and training. If you are teaching estimation to a room of thirty people who have never done it, handing out paper decks gets you started in ninety seconds without a single question about links, logins or which browser tab.

The honest limits: paper cards have no timer, no history, and nothing to export. A hybrid team usually ends up running both, with the remote participants in a browser room and the people at the table on paper, and the facilitator reading the paper votes into the room.

Two settings decide whether this works. Set scale to 100 per cent rather than fit to page, because fit to page shrinks everything by a few per cent and the cards stop matching a standard sleeve. Leave the paper size on whatever your printer actually holds; the layout fits three across and three down on both A4 and US Letter, so nine cards a sheet either way.

Paper weight is the difference between a deck that lasts a quarter and one that lasts a week. Ordinary 80 gsm office paper is translucent enough that a bright ceiling light will show the number through the back of a face-down card, which defeats the entire point. Anything from 200 gsm up is opaque and stiff enough to hold. If you only have office paper, print on both sides with a solid back, or glue two sheets back to back before cutting.

Cut the full-width lines across the sheet first, so you end up with three strips, then cut each strip into three cards. Doing it in that order keeps the cuts square, which matters when the deck is stacked. A guillotine is worth borrowing from whoever in the office has one.

Store one set per person rather than one communal pile. Shared decks mean somebody is always waiting for a 5, and waiting is exactly the moment where a voter looks around the table to see what everyone else has picked.

1. 1

   Choose your decks and how many sets

   Pick the deck your team estimates with, then choose one set per voter. A team of five needs five sets of the same deck so nobody is sharing.

2. 2

   Print on the heaviest paper your printer takes

   Use 200 to 300 gsm card if you have it. Set the printer to 100 per cent scale, not fit to page, or the cards come out the wrong size. Print one test sheet before committing to a full team set.

3. 3

   Cut along the dashed guides

   A guillotine or a craft knife against a metal ruler gives a straight edge. Scissors work but the cards will not stack squarely. Cut the long lines across the whole sheet first, then cut each strip into cards.

4. 4

   Round the corners if you want them to last

   Square corners fray first. A corner punch takes a couple of minutes for a whole deck and roughly doubles how long the cards survive being shuffled into a bag.

5. 5

   Store one set per person, banded or bagged

   An elastic band, a small zip bag or an empty playing card box per set. Keep them in the room you estimate in rather than scattered across desks, or half the team will arrive without theirs.

Every deck here ends with two cards that carry no number, and both of them are doing a job that the numbered cards cannot.

The question mark means I do not have enough information to estimate this. It exists because the alternative is worse: without it, someone who genuinely does not know will pick a middle number to avoid looking lost, and that guess then sits in the distribution looking exactly like an informed opinion. A question mark is a request for the story to be explained again, and it is often the most valuable card played in a session, because the person holding it is usually not the only one who is unsure.

The coffee cup asks for a break. Estimation quality falls off sharply after about forty minutes, and the room usually notices too late. Giving people a socially cheap way to say so, without having to interrupt a discussion to ask, is far cheaper than the bad estimates a tired room produces. Treat a coffee card as a vote to pause, not a joke.

If you also run sessions in the browser, SprintsPoker treats these two cards the same way on screen: they count towards the agreement percentage, because abstaining together is still a form of agreement, but they are left out of the average, median and range. Averaging an admission of uncertainty produces a number with no meaning attached to it.

If your team has no existing habit, print Fibonacci. It is the default for a reason: the gaps widen as the numbers grow, which stops the room arguing about the difference between a 20 and a 21 and forces a real decision about scope instead.

Print T-shirt sizes if you are sizing work nobody understands well enough to put a number on, and print powers of two for infrastructure work and spikes where doubling is the only honest step. The deck comparison goes through the trade-offs properly.

Whatever you pick, do not print two decks and let people mix them. Points from one scale and points from another are not comparable, and a velocity history built from both means nothing.