good reframe. the thing you are actually capping is how much interruption a person puts up with before they leave, so it is a user-experience call wearing an ad-ops hat. the cap that earns most this session and the cap that keeps the audience are not the same number. work backward from the experience that makes people come back and set it there. squeeze for max impressions and you monetize hardest right up until the people stop showing up.
the frequency-cap question has a practical answer under all the it-depends, which is to start conservative and loosen carefully while watching retention, because it is far easier to raise a cap that turns out too tight than to win back users a too-loose cap already drove off. so err tight and loosen with data. so begin with a conservative frequency cap and increase it only while your retention holds, because the asymmetry matters, an overly tight cap costs you a little revenue you can recover by loosening, while an overly loose one costs you users you may never get back, so the safe path is starting from restraint and testing upward rather than starting greedy and testing down, which means the sane frequency-capping process respects that the two kinds of error are not equal, since undercapping is cheaply fixable and overcapping is expensively permanent, and the operators who tune caps well move up cautiously from a tight starting point watching the retention numbers rather than starting loose and discovering the damage only after the overexposed users have already decided the site is not worth the annoyance and stopped coming back.
there is no single frequency-cap number, and anyone who hands you one is selling a simplicity that does not exist. the right cap depends on your traffic, your formats, how much your audience tolerates, and what you are optimizing for. it is a test result, you run a few caps and watch both revenue and retention to find your line.
Under all the it-depends there is a practical answer: start conservative and loosen carefully while watching retention. The two errors are not equal. A cap that is too tight costs you a little revenue you get right back by loosening, but a cap that is too loose drives off users you may never win back.
the frequency-capping question has a geo dimension, because audience tolerance for ad frequency varies by geo and market, so the right cap can differ across your geos, which means a single global cap is often wrong for at least some of your markets. so ad-frequency tolerance varies by geo, a single global cap is wrong for some markets. so consider that the right frequency cap varies by geo because audience ad-tolerance differs across markets, because a cap right for one geo can be too tight or too loose for another with different tolerance, so the operators who cap well set it per geo rather than applying one global number, which means the frequency cap has a geo dimension, since ad-frequency tolerance varies by market and a single global cap fits some geos poorly, and setting the cap per geo based on each market’s tolerance rather than one global number is more precise, making the per-geo capping worth considering, because audience tolerance for ad frequency genuinely varies across geos and markets, and a single global frequency cap that is right for one market can be wrong for another, so setting the cap per geo based on each market’s actual ad-tolerance rather than applying one number everywhere is what fits the frequency cap to the varying tolerance across the different geos you serve.
frequency tolerance isn’t the same across geos, so one global cap is always wrong somewhere, too tight in one place and too loose in another. capping per geo is a bit more work but you stop leaving money on the table.
there isn’t one right number. it depends on your traffic, your creative, how you’re monetizing. anyone quoting you a universal cap is just guessing. run it on your own traffic and see.