How we check Atlanta hotel rates — our method

Most “cheapest hotels in Atlanta” pages are a ranked affiliate list with a paragraph on top. This one is not, and the difference is worth explaining in full, because a median is only useful if you know exactly what went into it.

What a median on this site means

For each of the 24 bases we collect the nightly price of every bookable double room inside the neighbourhood boundary, for mid-week stays inside a sixty-day window, and take the middle value. Not the average, which one $600 suite distorts; not the cheapest, which is usually a single room in a hostel and tells you nothing about what you will actually pay.

Prices are recorded before tax and before parking, and both are stated separately because they are large: roughly 16% in combined city and state lodging tax, and $25 to $45 a night for parking in Downtown and Midtown. We keep them out of the headline number so that two neighbourhoods are compared on the same basis, and we say so every time the number appears.

The procedure, step by step

01 Fix the dates Two mid-week nights, Tuesday to Thursday, sampled across a rolling sixty-day window so that one convention week cannot dominate a neighbourhood’s number. 02 Draw the boundary We use the boundaries people here actually use, and we do not let a hotel two neighbourhoods away claim an address. Where the edges are genuinely blurred, the page says so. 03 Collect every bookable double All available double rooms inside the boundary at the time of checking, chains and independents alike, including short-let rooms where they are the bulk of the stock. 04 Take the middle value The median of that list, rounded to the dollar. One expensive suite or one distressed rate cannot move it the way it would move an average. 05 Strip the extras, then name them Tax and parking come out of the headline number and go into the page in words: roughly 16% lodging tax, $25–45 a night to park downtown or in Midtown. 06 Date it, and re-check weekly Every figure carries the week it was recorded. If a number is more than a week old it gets re-collected, and the ones that moved get rewritten rather than quietly left.

How this site makes money, and what that buys

Booking links on this site go to a search partner and we may earn a commission when a booking completes. That is the whole commercial arrangement, and it is why the rest of it matters: no hotel, neighbourhood or operator can pay to appear here, to be ranked higher, or to have a paragraph softened. There is no sponsored content, there are no paid reviews, and nobody sees a page before it is published.

The practical test we hold ourselves to is that we will tell you not to book. We say the overnight houseboat is a bad deal on our Kerala site and we say Buckhead is the wrong base here, and both cost us bookings. If a page never recommends spending less, it is an advertisement.

Three things this method gets wrong

It is a snapshot. A median recorded this week is not a forecast. Atlanta rates move with the convention calendar more than with the season, so a number on this site tells you what the market looked like the week we checked and roughly where a neighbourhood sits relative to its neighbours — not what you will be quoted in March.

Neighbourhood boundaries are arguable. We draw them the way people who live here talk about them, not the way the city planning department does, and the edges of Midtown, Old Fourth Ward and Poncey-Highland genuinely overlap. A hotel two hundred metres one way or the other can move a median by a few dollars.

Small samples are noisy. Cabbagetown and Virginia-Highland have very little hotel stock — mostly short-let rooms — so their medians swing more than the big ones. Where the sample is thin we say so on the neighbourhood page rather than pretending the number is as solid as Midtown’s.

Corrections

If a figure here is wrong, tell us and we will change it and say that we changed it. Prices carry the week they were checked precisely so that an out-of-date number is visibly out of date rather than quietly wrong. Everything else on the site — the walks, the journey times, the verdicts — was done by someone writing it, and where we have not been somewhere recently the page says when we last were.

In short

Mid-week doubles, sixty-day window, median not average, tax and parking excluded and stated, re-checked weekly, no paid placement, and a note wherever the sample is thin. If a number on this site has no date attached to it, that is a bug — report it.

Tell us about a wrong number ↗