How fragmentation happens
Fragmentation is rarely a decision. It builds up. An extra ad set is added to test an interest, another for a second city, another because a colleague wanted a different age range, another as a copy of something that once worked. Nothing is removed, so within a few months the budget is divided many ways and every division looks reasonable on its own.
The damage sits in how Meta optimises. Delivery improves as an ad set accumulates the conversion event you selected, and that history belongs to the ad set alone. Ten ad sets sharing one budget therefore build ten weak histories rather than one strong one, and each has to relearn whenever it is edited or paused.
Why fragmentation matters
The visible symptom is instability. Cost per result swings from one week to the next, ad sets drop in and out of the learning phase, and a unit that looked promising cannot be trusted because it produced too few results to be meaningful. Decisions made on those numbers are guesses dressed up as analysis.
There is an auction cost too. Narrow ad sets built from similar interests usually contain many of the same people, so the account bids against itself and pays more to reach an audience it already had. In a smaller market such as Nepal, where the pool matching any one interest is limited, that overlap arrives much sooner than advertisers expect.
Where fragmentation goes wrong
The trap is that fragmentation looks like diligence. Separate ad sets feel like control and like testing, and a granular report feels more professional than a simple one. But a test only works when each arm gets enough traffic to produce a readable answer, and most fragmented accounts never reach that point on any arm.
The other failure is repair by addition. When results fall, the instinct is to launch another ad set, which takes more budget out of the units that were closest to working. Fragmentation gets worse precisely when the account is under pressure.
How to act on it
Count the ad sets that spent money last month and divide the budget between them in your head. If any is receiving too little to buy a sensible number of conversions, the account is fragmented and the fix is consolidation, not another test. Merge the overlapping and under-fed units, widen the targeting to cover what they collectively reached, and keep the strongest creatives from each.
After that, set a rule for yourself: a new ad set is created only when you can name the question it answers and can fund it properly. Test creatives inside a well-funded ad set instead of splitting audiences, and turn off anything that has been paused for months rather than leaving it to be duplicated later.
