Buying guides for digital asset management are written by companies that sell digital asset management, which is why every one of them concludes that you need it now.

Short answer: you need a DAM when people who are not the asset's creator have to answer questions about it that they cannot answer by looking at it. Not at a file count. Not at a storage number. If your assets are made by one or two people, used in one place, and carry no licensing exposure, the answer is no, and a well-named folder is the correct architecture. If you have crossed the line, Cloudinary Assets has a free tier you can test the premise on before anyone signs anything.

The four signals that actually mean yes

Ignore asset count. It is the metric vendors lead with and it predicts almost nothing. These four predict a lot.

One: a second group touches the assets. An agency, a reseller, a regional team, a freelancer. The instant someone outside the original team needs to find and use an asset correctly, the knowledge that was in one person's head has to be written down somewhere, and a folder cannot hold it.

Two: a wrong use has a cost. An expired stock licence in a paid campaign. An unapproved product photo on a retailer feed. A person whose model release lapsed. If any of those would generate a real bill or a real letter, you need approval and expiry as fields on the asset, not as institutional memory.

Three: the same question gets asked more than once a week. "Is this the latest version." "Can we use this externally." "Do we have a landscape crop of this." Each instance is a person interrupting another person for a fact that should have been attached to the file. Count them for a fortnight; the number is usually startling.

Four: something other than a human is fetching assets. A build pipeline, a storefront, a feed generator, a coding agent. Software cannot recognise an image on sight and cannot ask a colleague, which is a much harder constraint than most teams anticipate. That case is covered properly in what an AI agent needs from your DAM.

One signal is a maybe. Two is a yes.

When the answer is genuinely no

I want to be specific about this, because the failure mode of buying too early is worse than the failure mode of buying too late. A library nobody uses is a subscription plus a migration plus a quarterly meeting about why nobody uses it.

Do not buy yet if:

• One or two people create everything and both of them know where it is. Overhead with no payoff.
• Your assets are all first-party and all owned outright. No licensing exposure means the highest-value field set does not apply to you.
• There is one destination. If everything goes to one website and nowhere else, your CMS media library plus a naming convention is probably sufficient. The boundaries are worth understanding anyway: DAM, CMS and PIM explains which system should own what.
• Nobody has complained. Not a joke. Genuine asset pain is loud. If nobody is frustrated, you are solving a hypothetical.
• You cannot name the person who will own the taxonomy. This is the one that sinks implementations. A DAM without an owner degrades faster than a folder, because it has more places to be inconsistent.

That last point deserves its own sentence. Software does not create discipline. It gives discipline somewhere to live. If there is no discipline to house, you are buying an empty building.

What about "we will need it eventually"?

Probably true, and a bad reason to buy today with one exception.

The exception is when migration cost scales with the mess. Every month you defer, you add assets with no description, no approval state and no version lineage, and someone eventually has to reconstruct all of it. The reconstruction is the expensive part of any migration off shared drives, not the file copying.

So the reasonable middle position: do not buy the platform, but start capturing the metadata now, in whatever you already have. Embedded IPTC fields travel with the file regardless of where it lives, which means a photographer filling in creator, description and rights today is doing work that survives any future decision. That is free insurance and it is the single highest-return thing a pre-DAM team can do.

Sizing the decision honestly

If you have decided yes, the next trap is buying for the organisation you imagine rather than the one you have.

Ask three questions before you look at a single vendor page:

• How many people will log in weekly? Not seats purchased. Weekly active. This number is usually a fifth of what the project plan assumes and it determines which pricing model hurts.
• How many assets need describing by hand? Automated tagging covers a lot now, but the fields that matter commercially, approval and rights, are human judgements. Budget the hours.
• What breaks if this project stalls at 40 percent? If the answer is "nothing, we carry on as now", the project will stall at 40 percent. Successful rollouts have a surface that becomes the only way to do something people already need to do.

The money side of that, including the costs that do not appear on a pricing page, is in what a DAM costs and what not having one costs.

Test the premise cheaply

Before a procurement process, run a two-week experiment. Take your two hundred most-used assets, put them in a free tier somewhere, describe them with five fields, and see whether the questions stop.

Free tiers are real now and adequate for this. Cloudinary publishes its pricing openly with a free plan and no card required, and other platforms including Canto and Bynder will run a trial. Two weeks of real use tells you more than three vendor demos, and it tells you the thing demos cannot: whether your team will actually fill in the fields.

If the questions stop, you have your answer and a decent shortlist starting point in DAM platforms compared by who they are actually for. If they do not, you have saved yourself a year. Either way, what digital asset management actually is is worth reading first so you are testing the right thing.