A real 2014 campaign, for a Marvel release.

This is the oldest real thing on this entire site: a Google Ads campaign built for a Marvel Avengers theatrical release in 2014, before Adostrophe was formally incorporated but under the same working identity. It's kept here as a genuine historical record rather than rewritten into something newer than it is.

A quick note on naming Marvel directly: this describes an actual marketing technique used on a real campaign, not merchandise, not implied endorsement, just a factual account of how the keyword structure was actually built.

How the keyword list actually got built

Manually, and at real scale. Starting from seed lists (cast names, character names) run through spreadsheet concatenation formulas to generate every reasonable variant, checked against search volume data, then expanded again from whichever variants showed real volume. The first pass produced over 5,000 keyword variants; manual relevancy filtering narrowed that down to roughly 500 worth actually bidding on. Free, public keyword tools still don't handle that kind of volume in one pass today; it took real, deliberate spreadsheet work.

Because the studio tightly controlled creative assets for brand consistency, the keyword targeting itself carried most of the optimization weight. There wasn't much room to differentiate on ad copy, so getting the audience segmentation right mattered more than usual.

Structuring around an ensemble cast

For a film with this many lead characters, a flat keyword list doesn't capture how differently people actually search. Separate ad groups were built for each major axis:

Genre and franchise: "upcoming action films," "marvel movies 2015," "comic book movies," capturing broad category interest.

Cast: individual actor names paired with "movie" and character-specific variants, a distinctly different search intent from someone searching the franchise name generally.

Characters, the highest-priority cluster: "iron man movie," "captain america movie," "thor movie." These consistently converted best, since they capture the most specific, committed search intent.

Production and crew: directors, writers, and the studio itself, for the smaller but real audience specifically following the production.

The negative keyword list mattered as much as the positive one

Phrase-match exclusions for "free," "watch online," "torrent," "download," and "streaming" kept budget away from piracy-adjacent searches with zero commercial value. "Tickets" started as a negative during the awareness phase, then flipped to an actively targeted term the week of release. "Merchandise" and similar terms were excluded entirely and routed to separate post-theatrical campaigns instead, once that market actually existed.

The same phased structure still applies today

Pre-release (90–120 days out) optimizes for awareness: trailer views and site traffic, broad franchise and cast terms. Release week shifts hard toward conversion, with bids pushed toward "showtimes" and "buy tickets," terms close to an actual purchase decision. Post-theatrical pivots again, toward streaming and merchandise demand. What's changed since 2014 is mostly the tooling: Smart Bidding now handles a lot of the manual bid escalation, and Performance Max can pull video, display, and search into one campaign instead of three separate ones.

What hasn't changed

Automation is only as good as the structure feeding it. A badly segmented account with AI-generated keywords and copy layered on top just automates the waste faster. The old discipline of understanding who's actually searching for what, and why, still does most of the real work; the platform just executes on it differently now.

This applies well beyond this one campaign: certifications and spend-based partner tiers are a baseline, not proof of results. Consistent, verifiable growth in an account, month over month and quarter over quarter, is the only real evidence that matters.