What is a good fill rate?

There is no universal figure, because it depends entirely on your floors. A high fill rate at a low floor is worse than a moderate one at a healthy floor. Judge it against revenue per thousand requests instead, which combines both effects into a single number you can actually optimise.

Why did my fill rate drop after adding a demand partner?

Most likely the denominator changed rather than the demand. If the new partner increased the number of counted requests without a matching increase in wins, fill rate falls while revenue holds or improves. Compare revenue per thousand requests before and after to see what really happened.

Does unfilled inventory hurt anything besides revenue?

It costs latency and user experience. Each unfilled pass adds time before a passback or house ad resolves, and slow ad slots hurt page performance metrics. Very low fill also signals to some partners that the inventory is hard to monetise, which can affect future bidding.

Fill Rate Calculator

Calculate the share of ad requests that returned a paid impression. Useful for diagnosing a demand or floor problem — but treated on its own, fill rate is one of the easiest metrics in ad ops to optimise in exactly the wrong direction.

Paid impressions

Ad requests

Fill rate

—

Enter both figures to calculate.

Runs in your browser · nothing sent, stored, or logged

Formula

Fill Rate = (Paid Impressions ÷ Ad Requests) × 100

Impressions divided by requests, as a percentage. Both numbers need to come from the same system and the same point in the delivery chain, or the result is meaningless.

Fill rate alone is a vanity metric

Fill rate is trivially maximised by dropping your floor: at a low enough price, everything fills. That makes a fill rate improvement meaningless without the revenue figure beside it. A hundred percent fill at a $0.10 floor earns $100 per million impressions; forty percent fill at a $2.00 floor earns $800 on the same inventory. Publishers who chase fill as a target rather than a diagnostic reliably end up with a fuller, cheaper, less profitable stack. Read it next to eCPM, always, and treat the pair as the actual metric.

Decide which requests you are counting

The word 'request' is used for several different events in the chain: the ad server request, the bid request sent to each exchange, and the individual bidder responses. Counting bid requests instead of ad requests can divide the denominator across a dozen partners and produce a fill rate that looks alarming while nothing is wrong. Before investigating a fill rate drop, confirm the definition did not change — a new header bidding wrapper or an added demand partner will move the number without any change in delivered revenue.

Frequently asked questions

What is a good fill rate?: There is no universal figure, because it depends entirely on your floors. A high fill rate at a low floor is worse than a moderate one at a healthy floor. Judge it against revenue per thousand requests instead, which combines both effects into a single number you can actually optimise.
Why did my fill rate drop after adding a demand partner?: Most likely the denominator changed rather than the demand. If the new partner increased the number of counted requests without a matching increase in wins, fill rate falls while revenue holds or improves. Compare revenue per thousand requests before and after to see what really happened.
Does unfilled inventory hurt anything besides revenue?: It costs latency and user experience. Each unfilled pass adds time before a passback or house ad resolves, and slow ad slots hurt page performance metrics. Very low fill also signals to some partners that the inventory is hard to monetise, which can affect future bidding.