Digital Asset Management at Catalog Scale: A Working Setup

What a lost file actually costs

Count the library before judging the process. A 600 SKU catalog with seven images per SKU, delivered across five channel specifications, is over 20,000 renditions. Add video, pack shots, lifestyle sets and offline collateral and the number doubles.

Production stills cost Rs 300 to 900 each at volume for flat and mannequin work, and Rs 1,200 to 3,500 for styled or model shots. Losing the masters for 150 SKUs means Rs 1 lakh to 3 lakh of reshoot plus three weeks of calendar. The more expensive failure is quieter: a listing goes live on version two while the pack changed at version four, and the brand only finds out through a customer complaint about the label.

One naming schema, enforced at export

Pick a schema and never negotiate it again. Something like ZM-KUR-402-BLU-M_front_v03_2400.jpg carries the SKU, the variant, the view, the version and the long edge in one string.

• No spaces, no brand or agency names, no photographer initials.
• Views from a fixed list: front, back, side, detail, scale, inuse, packshot.
• Versions as v01, v02, v03. Never final, final2 or finalapproved.
• Dates only as YYYYMMDD, and only where a date genuinely matters.

The important part is where enforcement happens. Build the schema into the retoucher export preset so files arrive correct, and check the first delivery of every batch against it. That check takes two minutes. Renaming 2,000 files later takes two days and still leaves errors.

Masters, renditions and one folder spine

Keep exactly one master per view. Generate everything else from it with a script or an action, in a batch, on the same day.

Channel specifications differ enough that manual resizing guarantees drift. Marketplace listings want a square white background with the product filling most of the frame and at least 1,600 px on the long edge for zoom. Fashion platforms want a 3:4 portrait crop. Quick commerce onboarding wants a clean 1,000 by 1,000 pack shot with pure white behind it. All three come from the same master.

The folder spine stays boring: brand, then category, then SKU, then raw, masters and renditions. RAW files move to cold storage after 90 days at roughly Rs 90 to 350 per TB per month, while the working library stays on a synced drive of two to five TB. Never retouch a rendition, and never let a rendition become the source for a new edit.

Track usage rights or lose the asset

This is the gap that costs real money. Model, influencer and talent shoots come with a usage window, usually six or twelve months, a geography, and a split between organic and paid media. Music beds, fonts and stock images carry their own terms.

Maintain one rights register with columns for asset id, talent or licensor, start date, end date, permitted media, geography and a link to the signed contract. Add a 30 day expiry alert. Require the asset id in every media brief so a file cannot be scheduled into a campaign after its window closes.

An expired model image running in paid media is a legal notice waiting to be sent, and the cleanup usually burns a week of creative capacity at the worst time of the quarter. Renewing a usage window before it lapses almost always costs less than reshooting the set.

Wire the library to the channels

The library only pays off when it is the single place channels pull from. Maintain one row per SKU with the live asset links per channel, the current version number, and the date each channel last received an update.

Two triggers deserve automation. A version bump on any master creates a relisting task for every channel using that view. And a packaging change creates an asset review before the new pack reaches the warehouse, because the first order shipped in new packaging with old imagery is how return rates rise for no product reason at all.

Give the agency and the marketplace operations team read access to the same spine. Every parallel folder someone maintains in a personal drive is a future mismatch.

Add one weekly check that takes ten minutes. Pull the live primary image URL for the top 50 SKUs by revenue, compare the version in the filename against the current master version, and list any mismatch. On a catalog of 600 SKUs across four or five channels, this check normally surfaces two to five stale listings a month, each of which was silently showing an old pack, an old price sticker or a discontinued shade.

Buy software only after the schema works

Run a two week audit before signing any tool. Pick 20 live listings at random and try to find the approved master for each in under 60 seconds. Note how many attempts fail and why.

If more than three fail, the problem is naming, ownership and handoff discipline, and a DAM will simply hold the same mess in a nicer interface. Fix the schema, name one owner for the library, and make the export preset the gate. Then, if scale genuinely demands it, buy the software and migrate a clean library into it.