Many Client Domains, One Workspace: Running an Agency Portfolio

An agency running twenty client domains has a different problem from a company running twenty of its own. The company can decide who sees what. The agency cannot — each client owns their data, expects to be able to take it with them, and will at some point stop being a client. Every arrangement that ignores this works beautifully until the first departure.

Zurich has a dense population of small consultancies in exactly this position: three to six people, a portfolio of clients across German and English, and an operating model that depends on being able to hand work over cleanly. The tooling question for them is not which dashboard is prettiest. It is which arrangement survives an offboarding without a week of unpaid admin.

What follows is a way of setting up a multi-client workspace so that joining, running and leaving are all ordinary operations. The mechanics refer to the Semalt panel; the structure applies to any tool that separates ownership, visibility and permission.

Why the usual arrangement breaks

The common setup grows by accident. A client is onboarded, someone connects their Search Console using an agency account, a report template is copied from the previous client, and it works. Twenty clients later the agency owns the connection to twenty properties it does not own, on one login shared by four people, and nobody can say which reports were sent when.

• Ownership confusion The connection sits on an agency account; the property belongs to the client. Nobody wrote down which is which.
• Shared credentials Four people, one login, no record of who changed a campaign setting and when.
• Reporting by copy-paste Each client report is a duplicated template that drifts from the others over time.
• Departure as a crisis Handing back access means untangling a connection that was never separable.

None of this is negligence; it is the natural end state of a setup optimised for onboarding speed. Fixing it costs half a day once, and the half day is best spent before the portfolio reaches the size where it matters.

Ownership, visibility and permission are three different things

Almost every multi-client problem dissolves once these three are held apart. They tend to arrive fused together because a single login grants all three at once, and separating them afterwards feels like extra work with no immediate payoff — until the first client transition.

The payoff is visible the first time a consultant leaves the firm. Their access is withdrawn, the properties stay connected, the history stays intact and the handover is a reassignment rather than a reconstruction. The same applies in reverse when a new consultant joins mid-retainer and needs to know what was decided last spring.

Tags decide whether the workspace is usable in a year

Labels look like housekeeping and turn out to be the structural decision of the first day. They are the filter that applies across analytics, reporting and the portfolio view, and a portfolio of twenty domains labelled by instinct becomes a list nobody filters.

Three dimensions is the practical ceiling. More produces combinations nobody filters by, and half-applied labels are worse than none because they make the filter untrustworthy. The rule that keeps it working is that every property carries exactly one label in each dimension, checked whenever a contract changes.

What changes in the working week

The most noticeable effect is an absence: nobody assembles anything. Background workers keep the data current, so a cross-client view exists before someone needs it rather than being built when they do. For an agency whose consultants each spent half a day a month compiling reports, that is the largest single recovery of time in the whole exercise.

The second effect is on the order of attention. A portfolio overview showing clicks, impressions, click-through rate, trend and health flags across every connected property means the monthly review starts with the account that has moved most, not the account that happens to be first alphabetically. Over a quarter that changes which problems get found at all — and finding a client's problem before the client does is most of what a retainer is for.

The third is the project stream: assistant answers, generated reports, new links with their source details, tasks and campaign notes in one chronological record per property. For an agency this is the artefact that makes handovers survivable, because the answer to «why did we decide that in March» lives on the account rather than in a former colleague's mailbox.

The assistant in that stream is worth a note of its own, because it changes who can answer a client question. It decides per question which blocks of project data to load — none, one, two or three depending on relevance — and keeps up to twenty messages of conversation history, so follow-up questions do not have to restate the context. In practice that means a consultant who does not own an account can still answer a routine question about it correctly, which in a firm of four people is the difference between a same-day reply and a scheduling problem.

Reports clients actually read

Client reporting has two failure modes and they are opposites. Too little, and the retainer looks like nothing is happening. Too much, and it is filed unread and the same impression results. The workable shape is short, consistent and comparable month to month, with the detail available on request rather than delivered by default.

Reports carry the agency's own logo and colours, which settles the presentation question, and the server-rendered PDF holds up to 250 rows. That ceiling is a feature in this context: it enforces the discipline of choosing what goes to a client rather than forwarding a data dump. Where a client genuinely needs the full dataset, the raw export handles it at up to 10 000 rows in CSV or JSON.

Onboarding and offboarding as routine operations

An agency's real test is not how it runs an account but how it starts and ends one. Both should be checklists rather than projects, and both are short once the three separations from section two are in place.

The baseline snapshot in the first card repays itself repeatedly. Twelve months later, when the question of what the retainer achieved comes up, the argument rests on a dated record made before any work started rather than on a reconstruction from memory. Agencies that skip it end up arguing about a starting point that nobody wrote down.

Offboarding deserves the same deliberateness. A client who leaves and receives their property, their data and a clean handover within a day tends to speak well of the firm afterwards, and in a market this size that matters more than the retainer being retained. The technical portion of that handover is covered under technical services.

What a shared workspace does not solve

It does not resolve positioning conflicts. If two clients in the same city compete for the same terms, a portfolio view makes that visible in a way separate accounts never did — and then the agency has a commercial decision to make, not a technical one. Seeing the conflict earlier is genuinely useful; being told what to do about it is not something any tool provides.

It also does not create capacity. Twenty properties generate twenty times the task list of one, and the tasks are not completed by living in one place. The recovery is in preparation and reporting time, which is real and substantial; the execution time is unchanged. An agency of four people running twenty accounts remains an agency of four people running twenty accounts.

Choosing tiers across a client portfolio

Most agencies end up with a mixed configuration, and the sensible split follows the engagement type label rather than the client's size. Accounts on an analytics-and-reporting engagement need the data, not the automation. Accounts on a full retainer where the agency selects targets deliberately are the ones where the managed tier earns its cost.

A useful sanity check is to compare the annual licence for an account against the annual retainer it sits inside. Where the licence is a small single-digit percentage of the retainer, the discussion is not worth the meeting. Where it approaches a tenth, the engagement is probably underpriced for what it is being asked to deliver — and that is a conversation about scope rather than about tooling. The surrounding market picture, useful in both cases, sits in the generated market research.

Setting it up in four steps

Sequence matters more than speed. Connecting properties before deciding the label system produces twenty properties in no order, and the ordering then has to be done twice.

• Step one List every client domain with an owner, an engagement type and a state — on paper, before touching the tool.
• Step two Fix the three label dimensions and their permitted values. Write them down; they will outlive the person who chose them.
• Step three Connect properties in order of importance, with the client as owner in every case.
• Step four Run one account fully for a month before rolling the pattern out to the rest.

Step four is the one that gets skipped and the one that saves the most. The first account is where the label values turn out to be wrong, the report shape turns out to be too long and the monthly rhythm turns out to fall on a bad day of the week. Discovering that once is cheap; discovering it across twenty accounts simultaneously is a fortnight.

If you want to see what your own portfolio looks like from one place, connect two client domains in the Semalt workspace — the largest and the one you know least about. Comparing those two views says more about the real state of your account management than any internal review, and it takes an hour. Further pieces on running multi-domain work sit in the blog, with the editorial side under content.