About Penny's Window

A window is a small opening onto one thing at a time. This one opens onto the right month and activity.

I’m a mom who loves to travel and a scientist who loves data.

Are you the one who plans the family trips? The one who loses hours trying to work out whether you’re going at the right time of year? That was me.

So I started building this for myself — a way to check, month by month, whether the thing I actually wanted to do would work when we got there. Then, because that’s what scientists do, I checked how often it was wrong. That number is published, and so are the cases where the model and I disagree.

It got useful enough to be worth putting online for everyone else doing the same thing.

It’s named after our small deaf family dog who enjoys looking out the window. Yes, she has one beautiful blue eye.

What this is not

Not a booking site, not an itinerary builder, and not a list of the world's best places. Plenty of destinations here score badly for most of the year — and plenty score brilliantly and badly in the same month, depending on what you came to do. A coastline can be perfect for swimming in August and unwalkable in that same August heat, and both of those are true at once.

That is the point. “Is this a good idea?” depends on when you go and on what you want to do when you get there, and one number for the year throws away both answers. It is why every destination here is scored once per activity, twelve times over, and why a page here would rather show you four rows than one verdict.

Some pages carry a booking link and I earn a commission if you use one. Nothing here is scored, ranked or recommended because of it, and the pages with nothing worth linking to have no link at all.

The rules it holds itself to

Say when not to come. A destination page will tell you the months to avoid, and why. That is the part nobody selling a trip will ever write.
Never average away a catastrophe. Hurricane seasons, monsoons, wildfire closures and polar winters are listed against the months they affect and deliberately kept out of the score. September in the Caribbean is usually beautiful and occasionally ruinous; one number cannot say both.
An empty field beats an invented one. Forty-five destinations show no crowd figures, because the proxy used elsewhere did not pass its checks there. A crowd number nobody has checked is worse than none.
Crowds are measured against the place itself. 0 is a destination's own quietest month and 100 its busiest. The scale never compares one place to another, because visitor counts across countries are not comparable.
Coverage never follows commission. If the data says a place in a month and nobody sells a trip there, the recommendation stands anyway.

How much of this is checked

The model and the hand grades agree on 80% of cases and correlate at r = 0.88 across 322 destination-activity pairs. The model was built against 36 destinations and has since been applied to 102 more it had never seen — and the correlation rose rather than fell. Eighteen of the crowd figures are official government visitor counts rather than estimates, and eleven more are estimates with an official count to check them against, which they track at a median correlation of 0.91.

The whole method, including what it gets wrong, is written out in full: how a score is made.