Publishing and Media: You Own an Archive, Not a Warehouse

A publisher's back catalogue is the only asset in the business that appreciates while nobody is looking at it. It cost money to create, it has already been paid for, and every year it sits there it accumulates a little more scarcity value as the people who made it move on.

It is also, in most independent publishers and media businesses, almost entirely unsearchable. Which means it functions as a warehouse: something you pay to store and occasionally retrieve from, rather than something that earns.

This article is part of our guide to modernising a legacy business.

What is actually in there

Most publishers underestimate the archive because they think of it as the finished products. The finished products are the smallest part.

The works themselves. Books, articles, episodes, recordings. Usually catalogued at title level, rarely at content level, which means you can find a book but not a passage.

The rights position. Who owns what, in which territories, for how long, with what reversion terms. Frequently spread across contracts in filing cabinets and nobody's current understanding covers all of it.

The production material. Transcripts, raw footage, images, research notes, interviews that were cut. Often the most commercially interesting material and almost never catalogued.

The performance history. What sold, to whom, when, through which channel, at what price. The one piece usually in a system and usually never analysed.

The three things that unlock it

1. Content-level search

The jump from "we can find the title" to "we can find the passage" is the one that changes what is possible.

Once an archive is searchable at content level, several things become viable that were not: licensing extracts, repackaging into collections, answering rights enquiries in minutes rather than days, and supplying material to the people who make your new work.

This is the same technical problem as any document archive, covered in scanned, filed and forgotten. The difference in publishing is that the answer is more reliably worth it, because the archive is the product rather than the evidence.

2. The rights position, in one place

Unglamorous and frequently the single highest-return piece of work available.

Most independent publishers cannot answer, quickly, which works they hold which rights to. That uncertainty costs money in three ways: deals not done because the position could not be established in time, deals done on worse terms than necessary, and rights quietly reverting because nobody tracked the dates.

Extracting rights terms from historic contracts into a structured record is a defined project. For a catalogue of a few thousand titles, expect £15,000 to £50,000 depending on how consistent the contracts are. It frequently pays for itself on the first deal it makes possible.

3. Knowing what actually sold

Performance data usually exists and usually has not been joined to anything. Joined to subject, author, format, price and channel, it tells you what to commission, which is the decision that determines the next five years.

The uncomfortable question about AI

Publishers have an unusual position here, because they are simultaneously a potential user of these tools and a party with a direct interest in what has been done with published material.

Two things worth separating.

Using AI on your own archive, inside your own environment, to make it searchable and useful, is straightforward and does not require you to give anything away. This is the version worth pursuing.

Licensing your archive for model training is a commercial decision with real money attached and real long-term consequences, and it is a rights question before it is a technology question. Whether you can license it depends on what your contracts say, and a great many older author agreements do not contemplate it at all. Establish the position before entering any conversation.

The distinction matters because vendors sometimes blur them.

What to do first

Establish the rights position. Everything else is constrained by it, and it has independent value.

Make one collection searchable at content level. Not the whole archive. One well-defined slice, to prove the extraction quality and the usefulness before committing. Three to six weeks.

Join performance data to subject and format. A data project, usually a few weeks, and it informs commissioning immediately.

Then decide about the wider archive, on evidence rather than enthusiasm.

Key Takeaways

• A back catalogue appreciates while unattended. Unsearchable, it functions as a warehouse you pay to store rather than an asset that earns.
• The valuable material is not just the finished works. It is the rights position, the production material and the performance history.
• Content-level search is the change that makes licensing, repackaging and fast rights answers possible.
• Getting the rights position into one structured record is usually the highest-return project and it often pays for itself on the first deal.
• Separate using AI on your own archive, which is straightforward, from licensing your archive for training, which is a rights question first.

Frequently Asked Questions

Our contracts are inconsistent across forty years. Is extraction realistic?

Yes, with a caveat. Extraction gets you a structured record plus a flagged list of the ones that need a human to read them. That list is usually 10 to 20 per cent of the total, which is a manageable amount of expensive time rather than all of it.

Should we license our archive for AI training?

It is a genuine commercial option and it is not a technology decision. Establish what your author and contributor agreements actually permit, take advice on the ones that are silent, and be clear that it is generally not reversible. We would not advise entering that conversation before the rights position is documented.

What does content-level search cost for a mid-sized catalogue?

For a few thousand titles already in digital form, typically £20,000 to £60,000 including the interface. Material that exists only as print or poor scans costs considerably more, which is why the slice test matters before committing.

Is any of this relevant to a small media business rather than a publisher?

Yes, and often more so. A production company's raw footage, a podcast network's back episodes and a magazine's photo library have the same characteristics: expensive to create, already paid for, and invisible. Our work with FilmWaffle and Storyroom touched on exactly this.

---

Sitting on an archive that nobody can search? Talk to Halo Technology Lab. Our AI solutions for media start with one collection and a rights position, not a platform.