Three Domains, No Marketing Department: SEO as Shared Work
A shop with three domains almost never has three people looking after them. It has one — an estimator, a shop manager, sometimes the owner — with an hour on a Friday, and often an outside contractor who makes the edits. This is what a shared workspace has to do so that hour is enough, and the four jobs it will not do for you.
"Team" in most SEO writing means a marketing department. Around the Capital Region that fits almost nobody, and the domains exist anyway — usually more of them than there are people paid to think about them. A fabricator ends up with a main site, a parts catalogue that grew its own domain because the numbering system needed one, and a service-area site for the towns the crews cover. An equipment dealer has a corporate site plus a brand site the manufacturer insists on. Nobody planned a portfolio; it accumulated.
The failure there is not analytical — anyone who reads shop drawings can read a traffic chart. What breaks is the record: what was noticed on the parts site in March, what was decided, whether the contractor did it, and where that conversation went when the person who had it changed jobs. In the rebuilt Semalt panel the data, the discussion about it and the resulting task sit in one stream.
Nobody here has a marketing department
Start with an inventory of who touches the site. The list is short, none of the people on it were hired for this, and that is the condition any tooling has to survive.
Knows what buyers ask
Reads every enquiry and can tell within a sentence whether a lead was serious.
- Best read on lead quality in the building
- Available between quotes, not on a schedule
Owns the site by accident
Inherited the login when someone left. Updates a capability page when a machine lands, then leaves it for months.
- Knows the products better than any agency
- Has an hour on a Friday, at best
Signs off, rarely edits
Wants one answer: is this producing quote requests. Will read two pages, never a dashboard.
- Decides what gets funded next quarter
- Needs the summary, not the data
Makes the changes
A freelancer or small studio with the CMS credentials, paid per engagement.
- Fast when told exactly what to change
- Access must end when the work does
Three requirements fall out of that list, and they are the whole of this article. The work has to be written down where the next person finds it. A question has to be answerable without opening four tools. And a supplier's access has to end on the day the engagement does, without a password reset that locks out the shop manager too.
My SEO Stream: the feed that keeps the record
Stream is the conversational feed inside My SEO, one per project. The value is not the chat; it is that several kinds of event land in one chronological place instead of four.
One stream per project, five kinds of entry
What a colleague sees when they open a domain nobody has looked at since spring.
- Assistant answers. Questions about this project and the replies, kept in place rather than in one person's browser history.
- Automatic reports. Generated summaries dropped into the timeline beside the discussion of them, not filed elsewhere.
- New backlinks. Each placement posted as it goes live, with the donor domain's authority rating and traffic — the figures anyone will want when it is questioned later.
- To-dos. Tasks raised by the system or by a person, each carrying a state.
- Campaign news. Keywords picked up, candidates awaiting a decision, on-site suggestions ready for review.
The difference shows up at handover. A catalogue that has run unattended for two months is not a mystery to whoever inherits it: events are in the order they happened, and the odd-looking placement already has its figures beside it. Background workers refresh the data continuously, so the feed is not a stale log. The panel runs to more than thirty-five interface pages and eleven integrations; the stream is the one page you can send a colleague to without a tutorial.
An assistant that reads the project's own numbers
The assistant in the stream is not a general chatbot pointed at your site; it is bound to the project's real data. A router model reads each question and decides which data blocks to load — Search Console, ranking, campaign state, custom data — loading between zero and three of them by relevance.
Zero matters as much as three. Asking what a canonical tag does pulls no project data. Asking which pages lost positions this month pulls the ranking and Search Console blocks and leaves the campaign alone. Answers stay tied to your figures rather than drifting into generic advice.
Answers stream token by token, which sounds cosmetic and is not: within a second you see whether the question was understood, and can stop it if it was not. Up to twenty messages of history are kept, so a follow-up like "and on the service-area site?" carries the previous context. The Stream assistant also takes keyword and URL lists in bulk — how anyone with a catalogue will actually use it, pasting a batch rather than typing one query at a time.
Four filters, three states, and one search box
A chronological feed is honest and, after a few months, long. Four filters cut it down: All, Links, Files and To-do. Links gives the placement history; Files gives exports and uploads; To-do gives the work queue without the commentary around it.
To-dos carry three states, and the middle one is the useful one: active, deferred, or dismissed. Most systems offer only done and not done, which forces a decision nobody has made — so the item sits open for a year and quietly stops meaning anything.
| State | What it means | Typical shop-floor case |
|---|---|---|
| Active | Expected of someone, and visible until it is done | Rewrite the title on the guarding category page |
| Deferred | Worth doing, not now, reasoning still attached | Rebuild the dealer locator after the CMS upgrade |
| Dismissed | Declined on purpose, so it does not return as a fresh idea | Chase a consumer term that never precedes a quote |
Full-text search runs across every message, which is what makes the record usable rather than merely complete. Six months on, nobody remembers a date; they remember a word. Searching a part-number prefix or a town name returns the conversation, the report and the task together.
Sharing a site, and ending that access
Multi-tenancy is the unglamorous half of running several domains, and the half that does real damage when handled badly. The workspace is built on linked Google account groups: verification lives with a Google account, and a workspace holds several as a group. Google OAuth2 covers Gmail, Search Console and Analytics in one consent flow.
Individual sites are then shared with individual email addresses — the site, not the workspace, which is exactly the granularity a group of domains needs. The contractor rebuilding the parts catalogue gets the parts catalogue, not the main site's enquiry data, because nothing in the engagement requires it.
| Person | Sites they need | What they must not hold | Access ends |
|---|---|---|---|
| Shop manager | All three domains | Nothing — the standing owner | They leave |
| Estimator | Main and service-area site | Nothing critical | The role changes |
| Outside contractor | The site under contract | The other domains' figures | The engagement closes |
| Brand agency | The brand site only | Your catalogue and enquiry data | The programme ends |
Revocation is the point of the arrangement. Access granted per site is withdrawn per site, on the day the final invoice is settled, without changing a shared password and without locking out the people inside the business who still need to get in. Anyone who has run the alternative — one login circulated by email, changed in a hurry — knows what that is worth.
Site tags: main site, parts site, service area
Site tags are a workspace-wide filter. Tag a domain and every view — analytics, ranking, indexing, reporting — narrows to that set. It sounds like housekeeping; it decides whether a group of domains is legible.
The mistake is building a taxonomy: division, then brand, then region, then function. Deep trees fail because a domain rarely sits in one branch. A service-area site is regional, and a lead-generation site, and part of the fabrication line of business. Forced into one branch, it is unfindable from the other two.
Tags that stack
A few independent labels applied in combination, each answering one question.
- role: main, parts, service-area, dealer
- state: active, maintenance-only, retired
- who: in-house, contractor
Trees that break
One path per domain, chosen once, wrong as soon as the business changes shape.
- Fabrication → Alberta → Parts → Active
- Reclassifying means touching every branch
Six to ten tags across a whole group is plenty. A tag earns its place if you would ever filter by it alone: "parts" passes, because comparing catalogue domains against the rest is a question you will ask; "rebuilt in 2023" does not.
Tags also carry the commercial arithmetic. Campaign automation in the unified workspace is priced per domain, and a group rarely deserves the same level everywhere. Tagging by role makes that visible: the catalogue producing quote requests sits on the higher tier, a maintenance-only brand site lower or on neither.
Where the effort goes across a group of sites
The difference when one person supervises several campaigns at once.
- AutoSEO runs itself. Automatic keyword discovery and prioritisation, automatic link building, on-site suggestions, full analytics. Supervision, not operation.
- FullSEO adds people and control. Manual keyword selection with automatic fallback, placement against a target authority rating, human review of on-site changes, and a team of specialists, developers and writers behind it.
- Candidates are approved one at a time. The pool comes from Search Console, live results pages and your own seed terms — where a specifier's vocabulary gets separated from a student's.
- Placements draw on the partner network. Over 230,000 sites are available, and first measurable movement typically appears after four to eight weeks — not inside a reporting month.
Exports for analysis, PDFs for the owner
Reporting here has two audiences that want opposite things. One wants the data to work on elsewhere, in a spreadsheet or a quote-tracking sheet. The other wants two pages readable on a phone between site visits.
The export ceilings follow that split. CSV and JSON run to 10,000 rows, enough for a full keyword set on a catalogue domain. PDF is capped at 250 rows and rendered server-side, so it looks identical to everyone who opens it. The report builder configures the contents, and branding with your logo and colours matters when a file leaves the building.
| Format | Ceiling | Good for | Wrong for |
|---|---|---|---|
| CSV | 10,000 rows | Cross-referencing queries against real quote requests | Anything meant to be read |
| JSON | 10,000 rows | Feeding another system or a scheduled comparison | A colleague without tooling |
| 250 rows | The monthly page for the owner or a brand partner | Bulk data of any kind |
The visualisations serve the same split, each answering a different kind of question.
- Time series. Direction over a period — the only honest thing to show an owner about a four-week window.
- Metric cards and sparklines. A headline figure with a shape beside it, so a row reads as movement.
- Filterable, sortable tables. Fifty to two hundred rows per page, the size at which a person still scans rather than skims.
- Country and device heatmaps. Concentration, which on an Alberta industrial site is usually the finding: a narrow band of provinces and states, mostly desktop, mostly in working hours.
A week in practice
Here is the shape of it on a three-domain group — a main fabrication site, a parts catalogue and a service-area site — run by one shop manager with a contractor on retainer.
| When | Who | What happens | Recorded as |
|---|---|---|---|
| Monday | Nobody | Workers refresh the figures; automation posts overnight placements and new candidates | Stream entries |
| Wednesday | Estimator | Notes that two enquiries mentioned a capacity spec buried three clicks deep | One line, then a to-do |
| Friday, 20 min | Shop manager | To-do filter: four candidates approved, one deferred, two consumer terms dismissed | States, with reasons |
| Friday, 20 min | Shop manager | Asks which catalogue pages lost positions; the router loads two data blocks | Answer stays in the stream |
| Friday, 20 min | Shop manager | Sends the contractor two page changes on the parts site | To-do assigned and dated |
| Month end | Owner | Reads a 250-row PDF: direction, the pages preceding quotes, what is still noise | Branded report |
| Engagement end | Shop manager | Revokes access to the parts site; what the contractor wrote stays | Sharing withdrawn |
That is about an hour a week from one person, plus a few minutes from someone who never logs in. It works because the observation, the decision and the change are written in one place, in order, and because the search box finds them again when the same problem returns next spring.
The questions that come up
Three domains and one and a half people. Is a shared workspace overkill?
The case is stronger at that size than at ten domains with a department, because there is no colleague to ask. The value is not collaboration in the meeting-room sense; it is that the record survives someone changing jobs.
Can our contractor see only the site they are working on?
Yes. Sharing is per site, granted to an individual email address and withdrawn the same way. When the engagement closes, that grant is revoked without touching any password anyone else uses.
What does the router actually do, and why should I care?
It decides, per question, which of your project's data blocks the assistant may load — Search Console, ranking, campaign state, custom data — and loads none, one, two or three. That is the difference between an answer grounded in your figures and a plausible general one.
How many tags should we create?
Fewer than feels right. Six to ten flat tags, applied in combination, beat any tree you could design. Tag before reading anything, because every view afterwards inherits that structure.
Why cap the PDF at 250 rows when CSV allows 10,000?
They are different objects. The PDF is rendered for a person to read, and no owner reads a thousand rows of keyword data. If a report needs more than 250 rows, it is a dataset. More on structuring these sits on our blog.
That list is not a disclaimer added for balance; it is where the work is. Collection, aggregation, monitoring and much of the routine placement work have moved to machines. What has not moved is anything requiring knowledge of your own business: which fight to pick, why the figure moved, and how to explain a flat quarter to someone expecting better.
Against that division, a shared workspace earns its place for one reason: it compresses the mechanical half enough that an estimator with an hour on a Friday has time left for the half that matters, and it leaves a record so decisions are revisited rather than re-argued. If you want a domain group reviewed before you tool up, that is part of what we do.
To try it on your own domains, open a workspace and connect the Google account holding your verifications, then tag the sites before reading a chart. The tagging takes fifteen minutes and decides whether everything after it is legible — which, on three or four industrial domains run by one person, is the whole question.
