Strategy · Conversational ads

There is no day 7: conversational ads break the durability clock

By Orcool · July 29, 2026 · 6 min read

The clock every media buyer plans around

Ask anyone who buys performance media how long a creative lasts and you will get a number, usually somewhere between two and six weeks. That instinct is well founded. We pulled a panel of 1,061 live Meta ads and measured how long each one had been running as a proxy for lifespan. The overall median sat at 39 days. Split by format the gaps were large: static image reached 101 days, video 66, and dynamic creative just 33, despite making up 57% of the volume teams were shipping.

Slice the same panel by category instead of format and the range gets wider still, with subscription products holding roughly six times longer at the seventy-fifth percentile than ecommerce DTC. We wrote that one up separately in the durability floor by vertical, and the mechanism behind spend concentrating on a small surviving set in the 6% problem.

All of that is real. It is also a description of one delivery mechanism, not of advertising itself. That distinction has been cheap to ignore for fifteen years, because for fifteen years almost every dollar of performance budget went into a ranked feed.

Why the clock exists at all

Creative fatigue is not a mysterious property of images. It is arithmetic. A ranked feed is a push surface: it decides what to show you, and it will show you the same thing repeatedly because repetition is how it maximises the odds of a response. Frequency climbs, novelty falls, click-through drops, cost per action rises, and at some point the buyer swaps the asset out.

Every practice built on top of that follows from the same arithmetic. Refresh cadences. Creative calendars. Fatigue alerts. The whole apparatus exists to manage a decay curve that the delivery mechanism itself produces.

push the feed picks  →  repetition same person, again  →  decay the clock starts

A chat impression is a different object

Now look at what happens when an ad appears inside an assistant. Someone types a question. The answer comes back, and alongside it sits a placement that relates to what was just asked. The trigger is intent expressed seconds ago, not a slot in a ranked queue that has to be filled with something.

Three properties follow from that, and each one breaks a piece of the feed model.

Frequency is not the delivery lever. The same person does not get served the same message forty times in a week because they were bucketed into an audience. They get served something when they raise a matching need. The exposure pattern that manufactures fatigue is simply not the mechanism here.

Context is unusually specific. A feed placement knows an interest cluster. A conversational placement sits next to a sentence the person just wrote. That is a much narrower target, and it rewards a much more specific message than a broad brand promise would.

The competitive set is one deep, not a scroll. In a feed your ad competes for attention against the next thumb-flick. In a chat the relevant question is whether your message fits the moment better than nothing at all.

The transplanted playbook

Look at the guidance that has appeared for these placements so far and it is remarkably familiar: keep five to ten creatives per ad set, refresh every two to four weeks, watch for fatigue. Sound advice, in a feed. Here it is a habit being carried across without anyone checking whether the thing it manages is present.

The platform reporting reinforces the habit, because it looks like every other ad dashboard: impressions, clicks, click-through rate, cost per click, conversions. Familiar columns invite familiar routines. Nothing in that table tells you which moment an impression happened in, which is the one thing that actually matters here.

Fatigue is a property of the delivery mechanism, not of the ad. Move the ad to a surface that does not repeat, and the clock you have been planning around quietly stops running.

What replaces durability: coverage

If lifespan is the wrong axis, something has to take its place. The honest replacement is coverage, meaning how many of the real moments in your category your current set of angles actually fits.

A meal-kit brand does not face one buying moment. It faces someone asking what to cook with three ingredients, someone comparing weekly costs against a supermarket shop, someone looking for a way to feed a family without an evening of planning. Those are three different states of mind, and a single well-made asset that covers one of them is not durable in the useful sense. It is simply narrow.

Which is why the asset stops being the unit that has to hold up. The angle becomes it. An angle can keep working for as long as the moment keeps recurring, and moments recur far more stubbornly than novelty lasts. The failure mode changes as well. In a feed you fail by wearing out. Here you fail by being absent from moments your category is full of.

This is also where the metric we care most about survives the move. Time-to-Winner, the number of calendar days from starting a cycle to a concept that clears the advertiser's own KPI gate and keeps clearing it, does not depend on frequency at all. In a feed, holding means surviving repetition. In a conversational surface, holding means the angle keeps fitting the moment it was written for. The clock stays honest either way, because what it counts is decision speed, not output.

What to do with this now

Stop importing the fatigue calendar by default. A two-week refresh rhythm in a surface with no repetition mechanism is spending on motion rather than on the placement itself. Let the data from the new surface tell you when a message stops working, instead of assuming it will on a schedule borrowed from elsewhere.

Write down your moments before you write copy. List the questions a real person asks right before your product becomes relevant. Ten to fifteen of them, in the words they would actually use. That list is the brief. Your app-store reviews and support tickets already contain most of it.

Audit for gaps, not for freshness. The most useful review is not which ad is tired. It is which moment has nothing pointed at it. Gaps are invisible on a fatigue dashboard, because a moment you never addressed produces no declining line to notice.

Keep the feed discipline where the feed is. None of this retires the durability data. Most budgets still sit in ranked feeds, and there the clock is real, measurable and worth planning around. The point is to stop treating one surface's physics as universal.

Frequently asked questions

Methodology note: the lifespan figures cited here come from our own panel of 1,061 live Meta ads, first published in the 6% problem and sliced by category in the durability floor by vertical. Days-still-running is a proxy for lifespan, not a controlled measure of decay, and the panel covers Meta placements only. The argument about conversational surfaces is structural rather than measured: these placements are new, public performance data is thin, and nothing here should be read as a benchmark for them. We are publishing the reasoning so you can weigh it before the numbers exist.