Project management for agencies: keeping many clients apart without chaos
An agency with eight people is running work for eleven clients. Four of those are monthly retainers with no end date, five are fixed projects with deadlines, and two are in that uncomfortable state where the contract is signed but nothing has started. Every client has a different contact, a different approval process, and a different opinion about how often they should hear from anyone.
The problem is not any single project. Each one, on its own, would be straightforward. The problem is the portfolio: eleven streams of work, one pool of people, and a set of questions that nobody can answer quickly. Who is overloaded next week. Which client is waiting on the agency and which is waiting on themselves. Whether the fourth round of revisions on a logo was ever agreed to. This article is about structuring the work so those questions have answers.
Why agency work breaks generic setups
Most project management advice assumes one team working on one product. An agency violates that assumption in four specific ways, and every structural decision follows from them.
People are shared, projects are not. A designer works across five clients in a week. Any view organised strictly by project hides the only number that matters for planning, which is how much of that designer's week is already spoken for.
Confidentiality runs between projects, not around the organisation. Client A must not see client B's work, and often must not even know client B exists. This is the opposite of a product team, where everything internal is visible to everyone internal.
A large share of the work is waiting. Agency schedules are dominated by feedback rounds, approvals and assets that have not arrived. Work that is blocked by the client is not the same as work in progress, and treating them the same makes the board a lie.
Work ends, and then arrives again. Projects finish and clients leave. Retainers run indefinitely with the same six recurring tasks every month. One structure has to hold both without becoming a graveyard of dead boards.
One board per client, and the conventions that make it work
The default structure that holds up is one board per client, not one board per project and not one board for the whole agency. A single shared board makes confidentiality impossible. A board per project creates twenty boards for eleven clients and buries the small jobs.
A board per client works only if the boards are consistent, which requires three conventions agreed once and then defended.
Identical stages on every board
The columns must mean the same thing everywhere. If one board reads Brief, In progress, Internal review, With client, Approved, Done, then every board reads that. Someone moving between four clients in a day should never have to work out what a column means on this particular board. Consistency also makes it possible to look at every board at once and understand what is in front of you.
The two stages agencies most often leave out are the ones that matter most: Internal review and With client. Merging internal review into in progress hides quality problems until the client finds them. Merging With client into in progress makes the agency look slow for time it did not spend.
The same naming on every card
A card title that reads Homepage is useless three weeks later. A convention such as project, then deliverable, then round works: Website relaunch, home page, round 2. It takes no longer to type and it makes search work, which matters more than it sounds once there are hundreds of cards.
One place for the client's assets and access details
Every agency loses hours to hunting for a login, a brand guideline or the name of the person who signs things off. A single pinned card at the top of each client board, holding the links and the contacts, removes that cost permanently. Credentials themselves belong in a password manager rather than on a card, and the card can point to the entry. This is a small convention that pays back every week, and it is the one most often skipped because it produces nothing visible on the day it is done.
A board for work that belongs to no client
Pitches, the agency's own website, hiring, invoicing and the equipment that keeps breaking are real work that consumes real capacity. Left off the boards entirely, they become invisible, and the planning conversation concludes that everyone has two free days that they demonstrably do not have. One internal board fixes this.
Seeing capacity across every client at once
This is the question a board per client does not answer by itself, and it is the question agencies most need answered.
The mechanism is to view the same cards grouped by person rather than by stage. Whether a tool offers this as a dedicated view, a filter across boards, or a personal list of assigned work, the requirement is the same: one screen showing what each person has in flight everywhere, not per client.
Two habits make that view honest.
Every card has one owner. Not two, not a team. A card with two owners is a card that neither will start.
Every card in progress has a date. Without dates, a person with nine cards looks equally busy whether those nine are due this week or spread over two months.
With those in place, the weekly planning conversation changes from a round of reassurance to a short reading of the screen. The most useful single number is not how many cards someone has, but how many are sitting in With client, because that is the work that will land back on the agency without warning and does not appear in anyone's estimate of the week.
A timeline view of the same cards adds the other half of the picture, showing where two clients' deadlines collide before the collision happens rather than afterwards. Tools differ a great deal in whether that view costs extra, which is worth checking on the features page of anything under consideration.
What clients should see
There are three workable models, and mixing them by accident is what causes the awkward conversations.
Nothing. The board is internal. The client receives a weekly note and a call. This is the least work and the least transparency, and for clients who do not want to be involved it is genuinely the right answer.
A shared client board. The client is a guest on their own board and sees everything on it. This is efficient, and it demands discipline, because every internal remark about a budget or a difficult stakeholder is now visible. Agencies that do this well keep internal discussion out of cards entirely and treat the board as a shared workspace from the start.
A split board. The internal board carries the working detail, and a second, simpler board shows the client only what they need: what is with them, what is coming, what is approved. It is the most work and the most controlled, and it suits clients who need to see progress but should not see the mechanics.
Whichever model is chosen, two practical points matter when comparing tools. Whether permissions are per board or per workspace, because per workspace permissions make client access an all or nothing decision. And whether view only guests count towards the price, because an agency with eleven clients and two contacts each is adding twenty two people who will never create a card.
Retainers and projects are different shapes
A fixed project has a beginning, a middle and an end, and its board empties as the project completes. A retainer never empties. The same six or seven tasks recur every month, and a board that treats them as one off cards fills with hundreds of completed items that all look the same.
The workable pattern for a retainer is a board where recurring work is created fresh each cycle from a template card or a checklist, rather than reopening the same card forever. That way the history shows twelve monthly reports rather than one card with twelve comments, and the question of whether last month's work was actually done has an answer.
It also makes the harder conversation possible. A retainer's value is only visible if the work done under it is countable. A board where each month's work is a distinct set of cards produces that count without anyone keeping a separate log.
Making revision rounds and scope visible
Scope creep in an agency rarely arrives as a request for something new. It arrives as a fourth round of small changes on something already approved.
Two mechanisms make it visible without turning anyone into the person who says no.
Number the rounds in the card title. Round 3 on a card is a fact, not an accusation. It appears in every list, and when the contract specified two rounds, the conversation has an obvious trigger rather than a feeling.
Give blocked work a label rather than a column. A card with a blocked label stays in the stage where the work genuinely is, so the board still shows the shape of the project. A blocked column, by contrast, becomes a parking area where cards go to be forgotten.
For teams that want to go further, capping how many cards may sit in each stage at once is the practice that keeps flow honest. The Atlassian guide to kanban covers how work in progress limits work and why they matter more than they first appear. In an agency the cap that pays for itself fastest is on internal review, because that is where quality is lost when everything is urgent.
What to check when choosing a tool for an agency
| Requirement | Why it matters for an agency | What to look for |
|---|---|---|
| Per board permissions | Client A must not see client B | Access granted per board, not only per workspace |
| Guest pricing | Client contacts multiply fast | Whether view only guests are counted as seats |
| Board count | One board per client plus internal | Whether the plan limits the number of boards |
| Cross board view of a person's work | Capacity is the real constraint | A view of assigned work across every board |
| Timeline of the same cards | Deadlines collide between clients | Whether a Gantt or calendar view is included or costs extra |
| Archiving | Clients leave, work should stay searchable | Archive rather than delete, and whether archived boards count |
| Per seat cost as freelancers come and go | Headcount changes every month | Monthly billing and easy removal of a seat |
The last row is the one agencies underestimate. A tool that is inexpensive for a fixed team of eight can become awkward when three freelancers join for six weeks. Comparing the per seat price and how boards are counted is worth doing properly, and the pricing page of any candidate should answer both questions without a sales call.
What to change first
Standardise the stages across every client board this week, and add the two that are usually missing: Internal review and With client. Then create one internal board for the work that has no client attached, so capacity planning stops being fiction. If the structure is right and the remaining problems are per board client access, boards that outgrow a plan's limit, or a timeline view of the same cards, Pinateca is free for up to five people and ten boards.
Q1. Is it better to have one board per client or one per project?
One per client, in most agencies. A board per project produces far more boards than clients, scatters small jobs, and makes it hard to see everything happening for one account. If a single client has several large concurrent projects, splitting that one client is reasonable, but it should be the exception.
Q2. How should client access be handled without exposing internal discussion?
Decide the model before inviting anyone. Either keep the board fully internal and report separately, share the board and keep internal remarks off it entirely, or run a simplified client facing board alongside the internal one. The failure mode is sharing an internal board and then trying to remember to be careful.
Q3. How can capacity across many clients be seen at once?
By viewing the same cards grouped by person rather than by stage, across every board. It works only if each card has exactly one owner and each card in progress has a date. The count of cards sitting with clients is the best early warning of a busy week ahead.
Q4. How should a monthly retainer be tracked on a board?
Create the recurring work fresh each cycle from a template card or checklist rather than reopening the same card each month. That produces a countable record of what was delivered every month, which is what makes the value of the retainer visible at renewal.
Q5. What is the most common structural mistake?
Leaving out the stage for work that is sitting with the client. Without it, time the agency did not spend appears as agency delay, the board overstates how much is genuinely in progress, and nobody can say which clients are holding up their own projects.
