A company in this city rarely owns one website. It owns a marketing site, a documentation subdomain, a status page, a changelog, two campaign microsites nobody retired, and a domain left over from the thing it pivoted away from — and the reporting question is which of those deserve to be measured at all.
None of that inventory was planned. Each address was registered by whoever needed it that quarter, verified against whichever Google account happened to be signed in. Two years later the collection is real, uneven, and longer than anyone can name from memory.
The product surface and everything that fell off it
Advice about managing several sites assumes they are peers: five franchise locations, three regional brands, a set of client accounts. A software or hardware company here has nothing that tidy — one property that matters commercially, one or two that matter operationally, and a tail of addresses nobody was assigned to shut down.
The awkward part is that this is neither one website nor five businesses. Treat it as one and the docs subdomain drags every blended figure sideways. Treat it as five and you spend four hours a month reporting on a launch microsite that has served eleven visitors since March.
- The marketing site pays for the office. The one property where a ranking change has a number attached, and usually the only one with an owner.
- The docs subdomain outranks it. Documentation attracts demand the marketing team never targeted, and is often the first thing a buyer reads.
- Status and changelog are infrastructure. They exist for customers who already bought, and need only to load, resolve and not embarrass anyone.
- The microsites and the pivot domain are sediment. Nobody defends them and nobody deletes them, so they accumulate whatever a neglected address accumulates.
Tag by what a site is for now, not by who built it
Site tags work as a global filter: set one and every screen narrows to the matching properties until you clear it. The mechanism is simple; the taxonomy you feed it is where the decision lives, and most teams get it wrong first time by copying the org chart.
Tagging by department fails because ownership moves and tags do not. The microsite was marketing's, then nobody's. The docs site is engineering's until a technical writer is hired, at which point half of it is not. Tagging by brand fails where three of seven properties have no brand to speak of.
Lifecycle survives, because it answers the question you actually ask on opening the panel: how much attention does this address deserve this month?
| Tag | What sits here | What you watch | What acceptable looks like |
|---|---|---|---|
| live | Marketing site, docs, anything a buyer reads before deciding | Queries, clicks, competing domains, campaign progress | Movement in the direction you chose |
| legacy | Retired microsites, the pivot domain, superseded product pages | Errors, redirect health, anything embarrassing | Silence, and a plan to eventually fold it in |
| experiment | The site launched six weeks ago to test a positioning | Whether anything at all is being discovered | A verdict inside one quarter, then a retag |
| infrastructure | Status page, changelog, developer utilities | Availability and crawl behavior only | No surprises; measurement is optional |
Keep the scheme flat. A technical team's instinct is a hierarchy — product area, then lifecycle, then environment — and hierarchies force every site to have exactly one parent. The first ambiguous case breaks the tree, and the pivot domain that is both legacy and still collecting branded searches is always that case.
- A tag is a filter, not a folder. A site can wear several at once, so nothing needs to be filed in one correct location.
- Four labels is a workable ceiling. Past that, people stop remembering which one they meant and start inventing near-duplicates.
- Retagging is the maintenance. An experiment becomes live or becomes legacy, and making that call quarterly is the entire upkeep the scheme requires.
AutoSEO — priced per domain, which forces the question
For portfolios where only two or three of the addresses justify active campaign work.
- Per-domain billing maps onto the tags. Campaign work runs on the properties you marked live; the rest stay connected for reporting without carrying a campaign.
- Keyword discovery and link building run unattended. Candidates are found and prioritized automatically, with on-site suggestions attached.
- First movement takes weeks, not days. Four to eight weeks is usual, worth knowing before an experiment site is judged after ten days.
Who can still see the property you forgot about
Multi-tenancy sounds like an enterprise word until you count the people who have touched your addresses: the contractor who built the launch microsite, the agency that ran one quarter's campaign, the founder who verified the domain against a personal Google account and has since left.
The panel handles this in two layers. Google accounts are linked as a group through one consent flow covering Gmail, Search Console and Analytics, so properties verified under different accounts stop being scattered across browser profiles. Then individual sites are shared to specific email addresses — one site, not the workspace — and any grant can be withdrawn.
Linked Google accounts
One consent flow pulls the verified properties, their search data and their analytics from several accounts into one group.
- Ends the sign-in shuffle between profiles
- Brings orphaned verifications back into view
- The consent covers the data, not the ownership
Per-site sharing
A single property is granted to a named address. The contractor working on the docs site sees the docs site and nothing beside it.
- Scoped to one site, not the account
- Given at the start of an engagement
- Withdrawn at the end of one, as a routine step
Form the habit of revoking on a schedule rather than after an incident: when a contract ends, the grant ends the same week. That is easier when withdrawal is one action against one site rather than an argument about a shared password.
The feed that remembers what happened to this site
A dashboard describes the present. Where the last person to touch a property has left the company, the more valuable artifact is a record of what happened and when. My SEO Stream is that record: a chronological feed per project carrying assistant answers, generated reports, newly placed backlinks with the donor's authority and traffic, to-dos and campaign news, in one column rather than four screens.
Four filters cut across it, and each answers a different kind of question.
- All — the chronology. Useful after a gap, because it reads as a history rather than a state.
- Links — what was placed. Each entry carries the donor's authority score and traffic, so the record is checkable rather than a count.
- Files — what was produced. The exports and documents generated for this project, which matters when somebody asks what went to the board in April.
- To-do — what is still open. The short list. On a legacy property it should be nearly empty; if it is not, the tag is wrong.
Full-text search runs across every message. It sounds like a minor convenience and turns out to be what makes a portfolio survivable, because the question you ask six months later is never what the state of a site is — it is whether anyone ever looked at the redirect from the old domain, and the answer is a phrase somebody typed in February.
A router decides how much of the project each question needs
The assistant in the feed is not a general chat window with your site's name pasted into the prompt. A router model reads each question first and decides which data blocks to load — search performance, ranking, campaign state, custom sources — between zero and three of them, by relevance.
Zero is a real outcome and a useful one. Ask what a canonical tag does and no project data is fetched, because none is relevant. Ask why clicks on the docs subdomain fell and the relevant blocks arrive. Ask something spanning performance, competing domains and campaign progress and three are assembled. The ceiling is deliberate: past three, an answer stops being grounded and becomes a summary of everything.
Answers stream token by token rather than arriving complete, which changes how you use it: within a sentence you can tell whether the question was understood, and reframe instead of waiting. Up to twenty messages of history are kept, so follow-ups work without restating which site you meant.
Two habits make it more useful. Name the property and the window, because a portfolio account cannot guess which of seven addresses you meant. And hand it lists in bulk — keyword and URL lists are accepted a batch at a time, which beats a conversation for anything over a few rows.
Active, deferred, dismissed — and why the third one carries the weight
Every to-do sits in one of three states, and the design decision worth understanding is that dismissal is an outcome rather than a deletion.
Active means a decision you intend to make. Deferred means the right question at the wrong time — a suggestion about the experiment site that only matters if the experiment survives the quarter. Dismissed means read, decided against, and on the record with a date attached.
On a portfolio of offcuts, dismissal keeps the system honest. A legacy microsite generates correct, well-reasoned suggestions indefinitely, every one technically right and none of them worth acting on, because nobody is investing in the property. Without a way to close them the list grows until people stop reading it, and once a list is unread the two items that did matter are invisible with the rest.
Deferred, used properly
A parking place with a reason — the right state for anything blocked on a decision somebody else makes first.
- Waiting on a launch, a hire or a rename
- Revisit when the site is retagged
- Not a polite way of saying no
Dismissed, used properly
A closed question, rejected for a reason that will still be true next quarter.
- Correct advice for a property nobody funds
- Work already done outside the panel
- Leaves a dated record of the decision
A week on a seven-property account is smaller than people expect. Monday: filter to live and read the feed since Friday. Midweek: link entries arrive on their own schedule and need attention only if a report moved. Thursday: pull whatever export the team asked for. Friday: empty the to-do list into the three states — ten minutes weekly, forty if left for a month.
| When | Filter | What you do | What you skip |
|---|---|---|---|
| Monday | live | Read the feed since Friday, note anything sharp | Anything tagged legacy or infrastructure |
| Midweek | live, Links | Check placements against the donors listed | Chasing a single quiet day |
| Thursday | per property | Export what the team joins against its own data | Rebuilding a chart somebody will not open |
| Friday | To-do, all sites | Every item into active, deferred or dismissed | Leaving anything in an undecided middle |
| Quarterly | all | Retag: experiments become live or legacy | Redesigning the tag scheme |
FullSEO — a review gate in front of the changes
For the property where an unreviewed on-site edit would be an incident rather than an improvement.
- Human review before anything lands. On-site changes wait for approval instead of appearing on the page, which is the difference between a marketing site and a documentation set under version control.
- Keywords chosen by hand, with a fallback. You select the terms; automatic discovery continues underneath so nothing stalls while a decision is outstanding.
- Placement against an authority target. Links placed manually to a stated Domain Authority threshold, with specialists, developers and writers attached to the account.
Your team will open the CSV before it opens the dashboard
This matters more here than in most markets, because the people receiving the report can write SQL. Hand a technical team a chart and they ask for the rows. Hand them the rows and they join the data against signups, trials and revenue in their own tooling, producing something better than a dashboard would have been.
So the export is the product, and the visualizations are how you decide what to export. Both exist: time series, metric cards, sortable and filterable tables at fifty to two hundred rows per page, sparklines, and country and device heatmaps. The tables earn their keep on a portfolio account, because sorting seven properties by one column is how you find the one that moved.
Read the row ceilings as guidance rather than limits. Ten thousand rows of query data is a real analysis; two hundred and fifty rows in a PDF is not a truncated version of it but an admission that a document nobody can filter should be a summary. Choosing which rows those are is most of the work of writing a good report. The configurable report builder also carries logo and color branding, which matters when the file leaves the building.
Questions from teams running more than three properties
Should the legacy domains be connected at all, if we are not investing in them?
Yes, and tagged so they never appear in a working view. Connected costs nothing and gives you somewhere to see errors and residual branded searches. Disconnected means that when somebody links to the old domain from a conference talk, nobody notices for a quarter. Connected does not mean attended to.
Can a contractor see one site without seeing the rest of the portfolio?
That is what per-site sharing is for. A named email address is granted one property and sees only that property's data and feed. Withdraw the grant when the engagement ends. This is separate from any access the same person holds directly in Search Console, which has to be removed there.
Does the assistant have access to everything in the account?
It answers within one project's feed, and the router loads at most three data blocks per question from that project's connected sources. It will not compose a cross-portfolio answer; comparing properties is what the filtered views and exports are for.
We already have dashboards internally. What does another workspace add?
Mainly the parts your dashboards lack: campaign state, link placements with donor detail, the decision record in the feed, per-site access management. If your tooling already reads the search API, use the JSON export as its input and keep the panel for the operational half. Duplicating a chart you own is not the argument.
Four things that stay on your side of the line
Everything above moves work off a person: collection, filtering, retrieval, formatting, the mechanical parts of keeping seven properties visible at once. It is worth being precise about what does not move, because a workspace that implies otherwise produces worse decisions than a spreadsheet.
- Prioritization. Ranking by impact is arithmetic. Ranking by what your quarter can absorb is judgment, and the panel cannot know the docs subdomain is being rebuilt.
- Cause-finding. The record narrows the candidates. Choosing among them requires knowing what your team shipped, which happened in a room rather than in the index.
- Goal-setting. A trend can be extrapolated. Whether the extrapolation is a plan or a wish is not in the data.
- Explaining it. The export states scope and date. Someone still has to say what it means and what happens next.
A sensible first pass takes an hour. Link the Google accounts so scattered verifications land in one group, tag every property live, legacy, experiment or infrastructure, then filter to live and see what is left. Most teams find a working set of two or three addresses and are relieved rather than disappointed. From there, the project feed and its filters give each a running record, and the export limits decide which format fits which audience.
If your portfolio grew the way most of them do here — by accretion, one quarter at a time — the useful step is inventory before analysis: open the panel and link a Google account to see every verified property in one list, including the two nobody remembered. Then tag them, share what needs sharing, revoke what does not, and dismiss what will never be acted on. Where the cleanup is a project rather than an afternoon, our engagement outlines cover how that is scoped, and the shared workspace keeps the result legible to more than one person.
