The first two seconds are a thumbnail problem

Most briefs describe the opening of a mobile game ad as a moving sequence: a hook that builds, a reveal at second two, a payoff at second four. That's the wrong mental model for the part of the ad that decides whether anyone sees the rest. On a feed, the first two seconds land as a still. The viewer is scrolling, the ad has only partly entered the viewport, autoplay hasn't necessarily started and the sound is off, so the decision to stop gets made on a single glance at what amounts to a poster.

The thesis is uncomfortable for people who love motion design: build and test the opening as a thumbnail first, and only then as a sequence. Most hook-rate problems are composition problems, not pacing problems.

What the viewer actually sees

Consider the mechanics of delivery rather than the mechanics of the edit. Feed placements load the first frame before playback begins, and on slower connections, or when the platform is conserving battery, that still holds for longer than the creative team assumes. Rewarded and interstitial placements behave differently. On the social and discovery surfaces where most testing volume runs, though, the ad's first job is to survive a glance at roughly a fifth of the screen while the thumb is already moving.

That glance reads three things and almost nothing else. Contrast comes first: does a shape separate from the background at a small size. Then one recognisable object, a face or a hand, a grid, a character, a single large number. Then implied action, something mid-motion or mid-failure that the eye wants to resolve.

A hook that depends on a cut at frame twenty, a text overlay that types itself in, or a slow zoom into the play area is invisible at the moment of decision. It may be an excellent second two through six. It's a blank thumbnail.

The mechanism: why motion-first briefs keep producing weak stills

Creative teams review ads in a player, full screen, sound on, from frame zero. Every part of that review environment flatters motion and hides the thumbnail. The brief says "hook", the editor hears "opening sequence", and the review confirms the sequence is good; nobody in the room ever saw the ad the way the viewer does.

There's an incentive layer too. Motion is what editors get hired for and what looks like work. A strong still frame looks like a screenshot, and it's difficult to justify a day of production on something that could pass for a screenshot, so the effort flows toward the part of the ad fewest viewers will ever see.

The platforms have quietly acknowledged the same reality. Cover frame selection, custom thumbnails for video placements and first-frame guidance in creative best-practice documents all exist because the platforms observed the same thing: delivery starts with a still, and the still decides hold rate.

A second-order effect: thumbnail-strong ads change what "hook rate" measures

Once a team starts optimising the still, the hook-rate metric shifts meaning. A higher share of viewers now stop, including viewers the poster attracted who aren't the game's audience. Hold rate at three seconds rises while completion and click-through can fall. Teams read this as the hook getting better and the body getting worse, and they start re-cutting the body.

The truer reading is that the thumbnail widened the top of the funnel and the body now has to filter more aggressively. That's a trade-off, not a fault. A still that is honest about the game, one that shows the actual grid or the actual character, filters early; it yields a smaller but cleaner audience. A still that is purely arresting (a face in shock, a giant number) stops more thumbs and hands more of the filtering to seconds two through six. Neither is wrong, but the brief should choose, and the measurement should know which one it chose.

A working method: brief the still, then earn the motion

A practical process for a creative team that wants to test this.

First, before anyone cuts an edit, produce the opening frame as a static image at the size it will occupy in a feed on a mid-range phone. Put it on a board beside ten screenshots of ordinary feed content. If it doesn't separate at arm's length, it fails, regardless of what happens at second two.

Next, write the hook brief as two lines: what the still shows, and what the first motion resolves. "Still: a nearly complete grid with one wrong tile. Motion: the tile drops into place and the row clears." This forces the concept to exist as a poster before it exists as a sequence, and it stops the editor from hiding a weak idea behind a fast cut.

Then test stills as stills. A static placement or a paused-frame comparison in a cheap pre-test will tell you which composition stops the scroll, at a fraction of the cost of producing four full video variants; take the winner into motion. This inverts the usual order, where the team produces four videos and the still is whatever frame the encoder landed on.

Last, when reviewing finished ads, look at the frame at zero seconds, muted, at thumbnail size, before anyone presses play. Make it the first slide.

What this does not fix

A thumbnail-first discipline improves the share of viewers who stop. It does nothing for what happens next, and the desk's earlier piece on the hook, proof, promise structure for the first six seconds still applies to the sequence that follows. The two disciplines fit together: the still buys the viewer, the sequence keeps them.

The place to be careful is the temptation to make every still a face in shock because faces stop thumbs. That works until it becomes the account's signature, at which point it decays like any other overused pattern. Rotate the object in the still as deliberately as you rotate the hook, and keep a count of how many live ads open on the same visual idea. Three is usually the point where the fourth one stops working.

Related archive reading

These articles provide related context and remain subject to their stated review status.

• Hook, proof, promise: a three-beat structure for the first six seconds
• Why the first playable level should not be level one