Client collaboration platform: four coupled decisions, one order to settle them

A client collaboration platform for a creative agency is not four features you buy separately. It is four decisions that constrain each other: how a client's access is scoped, what proofing binds feedback to, whether a file carries its own version history or gets re-uploaded every round, and who counts as a paid seat. Each one can pass its own demo in isolation. Bought as a system, you find the join where a producer spends Friday afternoon re-exporting cuts and re-pasting links.

The numbers below assume an agency of 5–30 internal people, 8–15 active clients at a time, and at least one reviewed deliverable per client per month, with no dedicated ops hire. Every figure is an input you should replace with your own count — none of it is a benchmark.

The couplings, and what each one charges you

The useful question is not "does it have a client portal" but "if I choose this, what does it force next". Here are the five couplings that decide most purchases:

Three of those couplings carry the real money. Take them in order, because the first one sets the constraints for the others.

Coupling 1: portal scope decides your seat model

There are only two ways software can hide internal work from a client: by assignment (this person sees the items they are attached to) or by container (this person sees this folder, board or project). Nearly every seat-pricing difference follows from which one the platform uses.

In kloudboard, scoping is by assignment. Guests without the "view all cards" permission see only the cards they are assigned to, per-member permissions are set in Settings > Team, and a board can be restricted to a member list. Owners and members count as paid seats; guests are free and do not. That combination is why client reviewers do not inflate the bill — and it is also the tax you pay: an unassigned card is invisible, so someone has to keep assignments honest. There is a full permission walkthrough in giving clients project access without exposing internal work.

Two things to check while you are here, because they are where the model leaks. First, does the client's drive access follow the same logic? Enabling the Drive page for a guest does not automatically grant Viewer access, and sharing a folder changes which files they can open but does not upgrade Viewer to Editor. Second, does the guest tier allow uploads at all? If your clients send you footage, a view-only portal means those files arrive by transfer link and get filed by hand.

Coupling 2: proofing depth decides your version model

A timestamped comment is table stakes. What separates platforms is whether that comment knows which cut it was written on. If it does not, you compensate with file naming: v1, v2, v3_FINAL, and a new link each round.

Run that compensation as arithmetic. Twelve deliverables a month, three review rounds each, is 36 files and 36 links a month instead of 12. Add the reconciliation: 20 minutes per round spent working out which notes still apply to the new cut is 12 hours a month at that volume. Your producer is now a version-control system.

In kloudboard, review is attached to the card rather than to a standalone link: open the card and use the Review action beside the attached media file or PDF, which gives timestamped comments and @mentions on videos and documents. Files keep version history, so uploading a new version to the same file preserves the old ones and reviewers move between versions instead of receiving a new link.

The honest tradeoff: a card-based review surface means the client has to be inside the workspace, not clicking an anonymous public link. That is a deliberate exchange — fewer places to look and a real audit trail, at the cost of an onboarding step and a permission setup you have to do correctly. If your clients will not log in, this coupling is the one that fails. The shape of the record you need is set out in the approval workflow with file version history.

Coupling 3: version model decides the handoff and storage bill

Once you commit to one canonical file per deliverable, that file is the deliverable, and where it lives matters. In kloudboard, files attached to a card are stored in the workspace Drive, so the draft an editor uploads and the file a client downloads are the same object, with version history intact. Guests can upload directly to folders shared with them when their Drive access is Editor.

Two consequences worth pricing before you sign. Storage is finite and every kept version counts against it, so a team that keeps 40 takes of every cut is buying storage whether or not the plan says so. And download links expire after a few hours — a link pasted into an email thread last week is already dead. The handoff has to be "open it in the workspace", which is a habit change, not a feature.

Seat economics is the output, not the input

Most teams start with the per-seat price. It is the last number that matters, because the first three couplings decide how many seats you need. Work it through on one roster: 9 internal people, 12 active clients, 2 reviewers per client.

That is 24 client-side access-holders, 33 people who need to open something. At an illustrative $25 per seat per month, billing everyone is $825 a month. Under a free-guest model the billable population stays at 9 — about $225 a month — and grows only when you hire. The gap is roughly $600 a month, or $7,200 a year. Insert your own quoted rate; the point is the size of the exposure, not this figure.

Now cost the other side, because free guests are not free. Assignment hygiene at 20 minutes a week is about 17 hours a year; at an illustrative $60 an hour that is roughly $1,040 — comfortably less than $7,200. But if the free guest tier can view without approving, those approvals fall back to email, and reconciling them costs maybe 15 minutes per round: 3 rounds × 12 deliverables × 15 minutes is 9 hours a month, or 108 hours a year. That single check can erase the entire saving. Confirm what a guest can actually do — comment, approve, upload — before you treat the seat tier as a win.

One caveat on any roster: a headcount tells you who needs access, not what you will be charged. Plan quantities, historical changes and adjustments only appear on a real invoice. Check the live pricing page and the billing section of your account rather than inferring the total from a team list. The wider comparison is in free client seats vs paid user seats.

The order to settle these decisions before you trial

Settling them in the wrong order is how teams end up trialling a tool whose seat model already rules it out. Scope first, then proofing, then versions, then seats — each answer narrows the next.

Platform decision sequence checklist

Settle the four coupled decisions in an order that keeps later choices open, before you spend a trial on the wrong seat model.

What a failed coupling looks like in month three

Three symptoms are worth recognising, because each looks like a training problem and is actually a decision problem.

• Shared logins. A client who cannot get into the portal is given the account lead's password. That is the signature of an onboarding step too heavy for the value on the other side, not a difficult client.
• "Final" means three things. Two clients approve by email, one approves in the tool, and nobody can say which version was signed off. Version binding was never enforced, so the record has gaps.
• A dormant account with a working link. Offboarding was done in one place. Removal has three surfaces — board membership, folder access, and the seat on the plan — and revoking one leaves the other two live. Do the full removal during the trial, then open the old link to confirm it is dead.

To act on this today, pick the deliverable that generated the most revision rounds last quarter. Write down the four decisions as you currently make them, by hand, then run the one deliverable through the platform and see which coupling breaks first. If the client reaches their file without your help, notes land on the frame, the approved version is named, and no client needs a paid seat, you have a platform worth configuring properly. If exactly one link in the chain fails, price the workaround in hours before renewal — not after.

For the configuration side once the decisions are made, see how to choose and configure client collaboration tools; this article deliberately stops at the evaluation.