Most performance video should be 15 to 30 seconds. Cold traffic works best at 15 to 20. Retargeting can carry 45 to 90 when there is something left to explain. Length is an output of how much you need to say, not a target you set in advance.
The useful question is never "how long should this be". It is "what is the shortest version that still makes the case", and the answer changes by placement and by how warm the audience is.
Length by placement
Placement changes viewing behaviour more than any other variable, so it changes the right length.
- Reels and Stories. 12 to 20 seconds. Vertical, sound on, thumb ready to move. Anything past 20 seconds needs to earn each additional one.
- Feed. 15 to 30 seconds. Sound often off, so the first three seconds have to work silently.
- Retargeting, any placement. 30 to 90 seconds. The viewer already knows who you are. Objection handling and detail belong here.
If you run one length across all placements, you are optimising for none of them. This is the cheapest fix on this page: cut a 15 second version of every winning ad and let it run in Reels.
Why shorter usually wins on cold traffic
Not because attention spans collapsed. Because on cold traffic a longer ad has more places to lose someone, and every extra second is another decision point where they can leave.
A 15 second ad has to be one idea, delivered once, cleanly. That constraint is doing most of the work. Long cold traffic ads usually fail not because of length but because length let three ideas into the same video, and no single one landed hard enough to move anybody.
When longer is genuinely right
Three cases justify it, and they are the same three every time.
High price. A $400 purchase requires more evidence than a $30 one. If the buyer needs to see materials, warranty and comparison before they act, a 15 second ad is not brave, it is incomplete.
Unfamiliar category. If the viewer has to understand what the product is before they can want it, you need the time. Trying to compress that produces an ad nobody understood.
Objection heavy purchase. Anything where the main barrier is "will this work for me". Testimonial and demonstration formats need room.
In all three cases, run the long version to warm audiences and a short cut to cold. Same footage, two edits, roughly one extra hour of work.
Read your retention graph, not the clock
Your own retention curve answers the length question better than any benchmark can.
Find where retention flattens. That is the point where the people still watching have decided to finish. If your curve flattens at 12 seconds and your ad is 45 seconds long, the last 33 seconds are being watched by a small committed group. That is not automatically waste, but you should know it is happening.
Find where retention drops sharply mid ad. That is a specific edit failing at a specific moment. Cut to that point and test the shorter version. It wins more often than not.
How to read the shape of that curve properly is covered in the five numbers that tell you if a creative is working.
The cut down is not a shorter version
Most teams make a 30 second ad and then trim it to 15 by deleting the end. That produces an ad with no payoff.
A proper short cut is re-edited from the same footage with a different structure. Hook, single strongest proof point, CTA. It is not the long ad with the last half removed, it is a different argument built from the same material.
This is a briefing problem more than an editing one. If the brief asks for one 30 second ad, you get one ad. If it asks for a 30 and a 15 as separate deliverables, you get two ads that each work. The difference costs almost nothing at the shoot or edit stage and a great deal if you skip it.
Practical defaults
If you want somewhere to start rather than a framework to think about, use these and adjust from your own data.
- Cold prospecting: 15 seconds, vertical, sound off legible
- Cold prospecting, considered purchase: 25 to 30 seconds
- Retargeting, objection handling: 45 to 60 seconds
- Every winner gets a 15 second cut for Reels regardless of original length
Then stop discussing length and start discussing what the ad is trying to say. Length arguments are almost always proxy arguments about clarity.