Two separate jobs hide inside this one request. The first is deciding what counts as a hook for your purposes, which is an editorial call nobody else can make for you. The second is getting back to the first few seconds of every ad you have delivered, across clients, across years, across whatever drive or project file each one ended up in. The second job is the one that decides whether the catalog gets finished, and there are four practical ways through it.
Watch and log
One spreadsheet: client, date, source file, hook type, one line describing what happens. Someone opens each ad, watches the top, writes the row.
This produces the best catalog of the four, because the taxonomy is yours and the descriptions are written in the language your team actually uses in a pitch. The cost is linear in the number of ads, and it does not care that half of them are ten seconds long. For fifty ads it is an afternoon. For the delivered work of an agency that has been running six years it is a project someone has to be paid for, which is why most of these catalogs stall in the second week.
A stills board
Screenshot the first frame or two of every ad into a shared board, grouped however you want. It is fast, it is scannable, and a board of two hundred opening frames is genuinely useful in a pitch meeting.
What it drops is everything that is not the first frame. Audio hooks vanish. So does anything built on motion, which includes most of the good ones. A whip pan into a product, a hard cut on a beat, and a slow push are the same still image.
Tagging at ingest
If you already run an asset system, add hook fields and fill them as new work lands. From the day you start, the catalog maintains itself. The back catalogue does not, and retroactive tagging is the same watch-and-log job with extra database steps. Tag vocabularies also drift, so the terms you used two years ago stop matching what you search for today. Where the automation in these systems actually helps is narrower than the marketing suggests, which is worked through in what automatic tagging can and cannot label.
Search the footage by description
The newer option is to put the delivered ads into a layer that indexes their content once and then answers questions about it, so instead of maintaining a hook taxonomy you ask for the thing you want when you want it. Queries look like a description of the moment rather than a keyword: the shot before anyone speaks, the point where the product first appears, the frame where pricing comes up on screen.
In Vivu, each client sits in its own project and a search runs inside one project at a time, so this builds a hook library per client rather than one across the whole roster. What comes back is a set of time ranges with a line of reasoning attached to each, and when the match is something written on screen, that line paraphrases the text it saw, which is usually enough to tell at a glance whether it is the frame you meant.
Two limits to plan around. Results are candidates, and anything you are going to put in front of a client needs opening and confirming, particularly when the reasoning paraphrases on-screen text that you would not want to quote secondhand. And a search over a project full of very short ads tends to return whole spots rather than the specific seconds inside them, since a fifteen-second video is close to being one moment already.
What none of this gives you
Performance. A hook catalog tells you what you made and never what worked, because what worked lives in the ad account next to spend, audience, placement, and the twelve other things that moved that week. Any catalog that starts sorting hooks into good and bad is smuggling in an opinion that the footage cannot support. Build it as a craft reference and a pitch asset, and leave ranking to the numbers.
It also does not solve getting to the delivered file in the first place, which for older work is often the real blocker and is covered in tracking down work you already delivered.
When you don't need a catalog
If you are building it for one pitch next month, pull fifteen ads by hand and skip the system. A catalog earns its keep when the same question gets asked repeatedly by different people, and when the back catalogue is large enough that nobody carries it in their head. Below that, the maintenance costs more than the lookups save. The same logic applies if your team turns over rarely, since a stable creative director is a functioning index.
One more thing before you upload anything. If the catalog is going to include ads you did not make, competitor work or reference spots you admire, check what your usage rights actually allow, because putting someone else's footage into a hosted system is a different act from watching it on the internet.
So the decision is mostly about volume and repetition. A small body of work, consulted occasionally, wants a spreadsheet someone actually maintains. A large one, queried by people who were not there when it was made, wants something that reads the footage, and you accept that it works one client at a time and that every result gets checked before it leaves the building.
FAQ
Can I catalog competitors' ads the same way?
Technically yes, legally it depends. Watching a public ad is not the same as downloading it and uploading it into a system you pay for, and rights to reference material vary by platform, by rights holder, and by whatever your client agreement says about third-party content. Check before you build a library on top of it. A safer version for competitor work is a link list with notes, which keeps the footage where it was published.
Does the hook start at the first frame or the first shot?
Pick one definition and write it down, because a catalog with two definitions in it is worse than either. The common working answer is the span before the viewer has decided to stay, which on most feeds runs a few seconds and may cover several shots. Whatever you choose, record the time range rather than just a label, so the next person can watch it and disagree.
Do I need a separate catalog for each client?
For anything hosted, assume yes, since client material is usually kept in separate projects and searches run inside one of them. That constraint is mostly a feature, because it matches how the rights and the confidentiality work anyway. If you want a cross-client view, the thing that travels between them is your own summary layer, a document or board that references each client's material without copying it.
Can this tell me which hook to use for a new brief?
No. It tells you what you have made and where to find it, which is a reference problem. Choosing the hook for a new brief is a creative decision informed by the product, the placement, and the account data, and a catalog that pretends to answer it is giving you a popularity contest among your own past work.