Flight budget calculator

A campaign booked from the 8th to the 21st runs for fourteen days, not thirteen. It delivers on the 21st: the flight stops at the midnight that opens the 22nd, not the midnight that opens the 21st.

That is not a convention this page invented. It is how a flight behaves in an ad account, and it is how the product represents one internally — an end date is stored as a day, and the instant the campaign stops is the start of the day after it. Subtracting one date from another gives thirteen, and thirteen is the answer to a different question.

The cost of getting it wrong is small and exactly quantifiable: one day of budget in the plan, and one day of delivery in the account. On a fourteen-day flight that is seven per cent of the money, which is more than most of the optimisations the same plan will spend a week arguing about.

It compounds, too, in the two places nobody checks. A plan of twelve monthly flights counted short is twelve missing days a year. And a reconciliation between what the plan said and what the account spent will be out by exactly one day's budget, which is usually written off as a rounding difference rather than traced to the one line of arithmetic that caused it.

Budget for the flight is the total for the whole run, in one currency.

First and last day of the flight are day numbers, counting the first day of the period as one. Inside a single month the dates themselves work — the 8th and the 21st. Across a month boundary, the rule is the same and the counting is one step longer. Leave the first day as 1. For the last day, add the days left in the starting month, counting the start date itself, to the date it closes on: 25 October leaves seven days of October (31 − 25 + 1), so closing on the 7th makes the last day 14. A flight across a year end works the same way. The arithmetic needs the distance and the rule that both ends count.

Daily budget you have set is optional, and it is what turns a planning tool into a reconciliation. It is the figure already in the account, or the one somebody is proposing.

Forwards is the plan: a total of 42,000 over fourteen days is 3,000 a day.

Backwards is the execution: a daily budget of 3,200 over the same fourteen days will spend 44,800, which is 2,800 more than the flight has.

Most plans are written forwards and run backwards, and the gap between the two is where flights quietly overspend. Nothing in an ad account announces that the daily figure and the flight total disagree; the campaign simply spends what it was told to spend until the money or the dates run out, whichever comes first.

Finding a 2,800 gap before the flight opens is a conversation. Finding it in the last week is a stop, and a stop inside a flight costs more than the money: an automated campaign that has its budget cut sharply may return to a learning state, and it will spend the rest of the flight recovering rather than performing.

Dividing a total by days is the honest default and rarely the best plan.

Demand is not flat. Weekdays and weekends differ in most categories and differ in opposite directions in some. Paydays matter. A flight built around an event deserves more money close to the event than three weeks before it. A flight that opens on a Thursday and closes on a Wednesday contains two weekends, and an even daily figure spends the same on both as on a Tuesday.

The practical method is to take the even figure as a baseline, weight the days deliberately, and keep the flight total unchanged. That way the pacing is a decision somebody made rather than the outcome of whichever days the auction happened to be cheap.

A flight of a few days carries a fixed overhead that a long one amortises.

Automated bidding needs a volume of conversions before it settles, and a short flight can end while the campaign is still calibrating. The cost per result over those days is closer to what the auction charges than to what the campaign could have achieved, and no amount of budget fixes it, because the constraint is time rather than money.

Short flights also concentrate spend, which raises the price paid for the same audience: buying a week's worth of impressions in two days means bidding further up the curve. The plan's cost per result should be expected to be worse than a longer flight's, and pretending otherwise is how a launch gets judged against a target it was never able to hit.

Check three things with the numbers on this page. That the day count includes the last day. That the daily budget in the account and the flight total agree, or that the difference is deliberate. And that the flight is long enough for the campaign to learn anything, which for an automated bid strategy usually means longer than anybody wants it to be.

None of those is difficult. All three are cheaper now than in the last week of the flight, which is the only other time anybody looks.