The Segment Audit: Where and When Your Budget Actually Performs
Most accounts are judged on their averages. One cost per acquisition, one ROAS, one headline number for the whole thing. The trouble is that an account does not perform evenly. It performs very differently on a phone than a desktop, in one region than another, at 9am than 9pm, and the average smooths all of that into a single figure that hides where the money is really working and where it is quietly leaking. A segment audit pulls the average apart. Here is how to do it and what it finds.
Why the average lies
An average is a blend, and a blend hides its extremes. An account running at a healthy overall ROAS can be made of strong desktop performance and weak mobile, or a great home region and several that lose money, averaged together into a number that looks fine. The headline says the account is healthy. Underneath, part of your spend is subsidising another part that does not work.
This is the whole point of a segment audit: to stop trusting the average and look at how the account performs when you split it by the things that actually vary. The money hiding in the segments is some of the easiest to recover, because the fix is usually not building anything new, it is bidding down or cutting the slices that lose and letting the ones that work keep more of the budget.
Device: the split that catches most accounts
The single most revealing split is usually device. Most paid search traffic is now on phones, and mobile often behaves completely differently from desktop: different intent, different conversion rate, different value per visit. Yet many accounts bid the same across both, as if a mobile visitor and a desktop one were worth the same. They frequently are not.
Pull performance apart by device and the gap is often large. Sometimes mobile converts worse because the experience is poor, which is a landing page problem worth fixing rather than just bidding down. Sometimes mobile is cheap, high-intent and under-funded, and deserves more. Either way, treating devices identically leaves money on the table. The audit question is simple: does each device get a bid that matches how it actually performs, or is one quietly subsidising the other?
Location and time: the slices nobody checks
Geography is the next split. An account can serve a whole country at one bid while performance varies hugely by region, some areas profitable, some losing money, some where you do not really compete. A location audit finds the regions quietly spending at a poor return and the ones worth leaning into. It also catches targeting mistakes, like paying for areas you cannot serve well or missing ones you could.
Time is the one almost nobody checks. Performance shifts by hour of day and day of week, sometimes sharply. An account might convert well in business hours and poorly late at night, or do all its real work at the weekend, and if it spends evenly across all of it, part of the budget is going to hours that do not pay. Pulling the data apart by day and time shows whether spend is matched to when the account actually performs, or just dripping out evenly regardless.
How to run the audit without chasing noise
The one danger with segment analysis is slicing so thin that you are reading noise. Cut the data far enough, by device and region and hour all at once, and every tiny slice looks dramatic simply because it has too little data behind it. A good segment audit watches for this: it only acts on slices with enough volume to be real, and it treats a wild number from a thin segment as a question, not a finding.
So the method is to split by one thing at a time, start with device because it catches the most, then location, then time, and look for segments with real spend performing clearly above or below the account. Where a well-evidenced segment is losing money, bid it down or cut it. Where one is quietly outperforming and under-funded, give it more. Done with discipline, a segment audit turns a single flattering average into a map of where your budget actually works, and that map is usually worth real money, because almost nobody looks at it.
FAQs
What is a segment audit? A review that splits account performance by device, location and time instead of judging it on the overall average. It finds the slices of spend quietly underperforming and the ones worth more budget, which the headline number hides.
Why doesn't the account average tell me enough? Because an average blends strong and weak together. A healthy overall ROAS can hide weak mobile, loss-making regions or poor time slots subsidised by the parts that work. The average looks fine while part of your spend is wasted.
Which segment matters most? Usually device. Most traffic is on phones now, and mobile often converts and values differently from desktop, yet many accounts bid the same across both. Splitting by device frequently reveals the biggest gap.
How do I audit location and time? Split performance by region to find areas spending at a poor return or targeting mistakes, and by day and hour to see whether spend matches when the account actually performs. Both are slices most accounts never check.
How do I avoid reading noise in the segments? Only act on slices with enough spend and conversions to be real, and split by one thing at a time rather than cutting the data so thin every tiny segment looks dramatic. Treat a wild number from a thin segment as a question, not a finding.