Your editor has uploaded a new cut. The producer is looking at yesterday’s version. The client has emailed comments about a shot without a timecode, and nobody knows whether the person with final sign-off has watched it. The files are shared. The work is still stuck.
Every one of those people could be somewhere different. Gallup’s latest figures put 52% of remote-capable US employees in hybrid roles and another 26% fully remote. Ericsson reports that video now makes up around three quarters of all mobile data traffic. Your contributors are everywhere, on whatever screen is in front of them.
We first wrote about this in 2017. The operational advice from then still holds. What has changed is the business around it: bigger files, more outside reviewers, AI that people want to use but don’t fully trust, and buyers who have learned that a MAM can arrive with costs nobody mentioned in the demo.
The short answer: A media asset management (MAM) system for video collaboration should connect footage, versions, feedback, permissions and the approval decision in one workflow that people actually use. Judge it on two sets of tests. The business tests: what delay costs you, what reviewers cost, what storage and AI cost, and what risk you carry when content leaves the building. The operational tests: fit with the editor’s tools, feedback tied to the exact version, visibility of what’s blocking approval, and reliable review on the devices and connections people really have.
Part one: four business tests
Most MAM evaluations start with features. I’d start with money, risk and time, because that’s what decides whether the system is still earning its place in year three.
1. What does a stuck review cost you?
Fact: Review delays rarely show up as a line item, but they’re real. Every unclear note, every revision built on the wrong version and every round started before the decision-maker has watched is editor time you’re paying for twice.
Question: Do a rough calculation before the demo. As an illustration, if a team delivers 30 projects a month and one avoidable review round costs three editor hours each time, that’s 90 hours a month, more than half of a full-time editor. Use your own numbers. The point is to know what a faster, clearer approval is worth to you, so you can judge whether a system delivers it.
2. Who pays for the people outside your team?
Fact: Collaboration means agencies, clients, freelancers, talent and partners. If every reviewer needs a paid seat, teams ration access. Then the agency gets a download link, the client gets an attachment, and your governed library leaks out through side doors.
Question: Ask how guests are priced, how they’re identified, and what it costs to add 30 external reviewers for a two-week campaign. The cheapest route for the team should also be the governed route.
3. What will storage, transfer and AI really cost?
Fact: Video is heavy. An hour of 1080p footage in ProRes 422 HQ, at Apple’s published target rate of about 220 Mbps, is roughly 99 GB. A system that insists on copying your whole library into its own storage before anyone can collaborate makes you pay to move it, pay to store it twice, and wait months to start.
Question: Ask whether the MAM can index footage where it already lives, what moves and when, and who pays for egress. Then ask the AI question directly. Can you choose which content gets AI indexing, or is it everything or nothing? Will you see the cost before a bulk job runs? Surprise AI bills are one of the main reasons buyers have become wary of this industry.
4. What risk do you carry when content leaves the building?
Fact: Collaboration is the moment content leaves your four walls, so it’s where governance gets tested. In our recent piece on digital asset governance, I argued that trust has to be designed in, not bolted on after an audit.
Question: You should be able to answer three questions quickly: who saw this cut, who downloaded the master, and were we cleared to show it? Sectors that handle pre-release content already work to formal standards, such as the Motion Picture Association’s Trusted Partner Network. And as AI starts tagging faces and transcribing speech, people want to see what it did and correct it. Industry work on provenance, such as the C2PA’s Content Credentials, shows where those expectations are heading.
Part two: four operational tests
The business tests tell you whether a system is worth buying. These four tell you whether people will use it. Run each one with your own media and the people who really review it.
1. It fits the work, from finding footage to opening the edit
Start with a real task, not an integration logo. Ask an editor to find an interview moment, preview it, add it to a project and reach the correct original. Then ask the producer to share a cut for review. Count the screens, manual downloads and duplicate uploads.
AI search helps when filenames and tags are thin. A producer can search for a spoken phrase or describe a scene. But the useful result is a permitted asset, connected to the right project and version, that the editor can actually use. Don’t assume one plugin means support for every creative tool. Test the exact application and version you use.
The real test is adoption. If an editor can finish faster with a shared drive and a separate review link, the MAM hasn’t earned its place.
2. Feedback stays attached to the exact version and moment
“Use the other shot near the end” makes sense to the client who wrote it. It sends the editor hunting. Time-coded comments help, but only if they refer to the version the reviewer actually saw.
Ask what happens when a new cut arrives. Can you still see v1 comments? Can you tell resolved notes from open ones? Most importantly, does approval of v1 stay approval of v1, rather than silently becoming approval of v2?
Keep internal and external boundaries visible too. Your team may need private production notes while the client sees a narrower conversation. And a single comment stream doesn’t replace the producer’s job of deciding which conflicting note wins.
3. Review visibility tells you what’s blocking approval
A producer needs to know who’s outstanding before asking the editor for another revision: who was invited, who has responded, which version is under review and who can give final approval.
Be careful here. A recorded view is not approval, and it doesn’t prove someone watched the whole video. An activity counter and an accountable approval record are different things. Access has a lifecycle as well. Test expiry and revocation, and remember that a downloaded copy can’t be recalled by closing a link, so set download permissions deliberately.
4. Review works on the devices and connections people really have
A mobile-friendly page is only the start. A client should be able to open the cut, play it, jump to the right moment and leave a note, on their own device and on a slower connection. At 99 GB an hour, nobody should be downloading masters to review a cut. Proxies make remote review possible, while originals stay available to the people who need them.
Test the limits. Can a reviewer resume after a dropped connection? What happens when an invitation expires while the page is open? Does removing a guest cut access on both laptop and phone? And remember that a native mobile app is a different capability from a usable mobile browser.
Smooth proxy playback is right for editorial feedback. It isn’t automatically right for final colour, audio or technical sign-off. Decide which approvals need a controlled viewing setup and say so.
Run one pilot that covers both
Bring a rough cut, a second version, the linked edit project and one restricted asset. Give the producer, editor and an external reviewer their normal tasks, and run the same sequence with every shortlisted vendor.
How Evolphin X fits
We rebuilt Evolphin X, our AI-native MAM, around these tests. I’ll be clear about what’s available now and what’s coming.
Core is the engine. Every Evolphin X deployment runs on Core. No engine, no car; no Core, no MAM. Core owns what collaboration depends on: access control, data integrity, AI integration and an open API. Index in Place connects to storage you already own, cloud or on-premises, indexes it, and keeps the record in sync when files change. Selective AI indexing lets you choose which files, folders or file types get AI processing. And when moving content does make sense, Evolphin X can transfer selected collections to new storage. That’s the business tests covered at the foundation: no forced migration, no all-or-nothing AI bill.
The Collaboration module is where review happens. Launched on 1 October, it gives your team, agency and clients focused, expiring workspaces to comment, approve and upload, without an account and without seeing the rest of your library.
More modules are coming. Portals arrive in early 2027, so you can share, license or sell content to buyers and partners from your own storefront. We’re also building Adobe plugins so editors can work with the library from inside their creative tools, and a Mobile app for teams who don’t work from one place. Analytics follows. Over the next year, we’re also building autonomous capabilities such as rights and expiry enforcement and workflow routing, so the system acts on your rules rather than waiting to be asked. The AI takes the work, not the job. Creative judgment stays with your people.
Whoever you evaluate, including us, ask the vendor to show each capability working in the current build, on your media. A roadmap slide is not a workflow.
Bring us a real review loop
Don’t judge a collaboration MAM on a polished demo. Bring one real project: a rough cut, a second version, an editor, an outside reviewer and a deadline. We’ll run it through Evolphin X with your team and show you what the current build does, what it costs, and what’s still to come. Email me at brian@evolphin.com.
.avif)