How the planning figures on this site are produced

What counts as a record

One record is one request: a quote form, a chat, a phone call or a message that came to us through one of our group-transportation websites or phone lines. Duplicates from the same person on the same date are collapsed into one, and anything our filters flag as spam or as a test submission is thrown out before we count anything. A request is not a reservation and not a completed trip — it is somebody asking us about a date. We keep that line bright because collapsing the two would quietly inflate every figure on the site.

Fields used and fields never used

Sample size and suppression rules

A small sample can say almost anything, so we set the bar before we look at the answer rather than after. These rules are applied by the calculation itself, not by whoever is writing the page.

• • We publish a figure only when at least 25 qualifying records sit behind it. A narrow slice — one city crossed with one event type, or one city crossed with one vehicle size — needs 30.
• • Any cell under 10 records is suppressed, and so is any second cell that could be subtracted to reveal it.
• • Percentages are rounded to whole numbers, and to the nearest five when fewer than 100 records are behind them. We keep exact denominators private and write “about” or “roughly” in the public sentence, because a figure printed to one decimal implies a precision this kind of data does not have.
• • Malformed values are rejected rather than guessed. If a passenger count was typed into the event-type box, that record drops out of the event analysis instead of being cleaned up into whatever we assume was meant.
• • Party buses, Sprinter vans, limousines, shuttles and motorcoaches are never pooled together, and requests are never pooled with reservations. Mixing either pair would produce a tidier number and a less true one.
• • When a figure fails any of these rules, the page shows a wider figure and labels it, or shows nothing at all. It never shows a filled-in estimate.

Geography attribution

We attribute a request to a city through the market website it arrived on. A request that came in through our Grand Rapids site counts toward Grand Rapids. We chose that over guessing from an IP address or a phone area code, because people plan trips from out of town constantly — the bachelorette weekend booked from three states away belongs to the city where the bus actually rolls, not to wherever the maid of honor happened to be sitting. We mapped every site to its market by hand rather than by pattern-matching domain names, and where a domain was genuinely ambiguous we left it out instead of assigning it. When a city does not have enough qualifying records, the page falls back to the state or province, then to the nationwide figure, and says on the page which one you are reading.

What these figures will and will not tell you

Used the way they are meant to be used, these figures are genuinely handy: they tell you when demand piles up, how far ahead groups like yours tend to lock in a date, how many hours people actually end up booking, and what size vehicle a group your size usually lands on. That is enough to plan a night around. Here is what they are not.

• • Not your price. We do not publish historical quote amounts, and they would not describe your trip anyway. Your price comes from your date, route, hours and vehicle, and we put it in writing.
• • Not availability. What is free on your date comes off our calendar, and we will tell you on the phone in about a minute.
• • Not industry market share. These are requests to us. We are one company, not a census.
• • Not reviews or satisfaction scores. Those live on our reviews page and are kept completely separate from this data.

How often we rebuild it, and how we handle mistakes

The whole set is recalculated at least every six months from the current records, and every published sentence is re-checked against the rules above at that point. A statement that no longer clears the sample-size bar comes down rather than being carried forward on last year’s strength. If you spot something that looks wrong, tell us: we would rather correct a number early than defend it. Corrections go in at the next scheduled rebuild, or immediately when a figure turns out to be flatly wrong.