The Approver Is Not the Operator: Many Websites, One Workspace

A cardiac service line, a research center and a single-product trial site have almost nothing in common, except that the same two or three people answer for their search performance — and that the only person qualified to approve a sentence on any of them has fifteen minutes a month to give you.

Most writing about portfolio SEO assumes a marketing department: an operator who does the work, a manager who signs it off, both paid to care about traffic. That arrangement is rare here. What exists instead is a set of institutional estates owned by departments that never agreed to share anything, and an approval step belonging to a clinician, a principal investigator or a general counsel.

Those two facts change what a shared workspace is for. It is not primarily a place to look at charts. It is a place to hold a precise question until the person who can answer it has a spare quarter of an hour, and to keep that answer where the next person finds it.

The approver is not the operator, and the approver is a specialist

Ask who has to say yes before a page changes. On a restaurant site it is the owner. On the estates that dominate this market it is somebody whose day job has nothing to do with the web: the physician who knows whether same-day evaluation is really available in November, the PI who knows which claim the readout supports, the counsel who decides whether a practice page may advertise the service at all.

That person is not obstructive; their caution is why the page is worth reading. But they will never open a dashboard, and no process depending on their curiosity survives its second month.

The second condition is structural. A hospital system's estate is a main domain plus service line sections, a research institute, a foundation and whatever a department commissioned directly. A university has schools and centers with separate budgets and separate content systems. A biotech has a corporate site under investor-relations discipline and a trial microsite under a different review regime entirely. Each has its own owner, and none of them asked to be part of a portfolio.

One project, one feed, in the order things actually happened

The project feed inside the rebuilt Semalt workspace is called My SEO Stream, and its organizing idea is plain: everything that happens to a property lands in one chronological conversation instead of being scattered across a reports tab, a notification bell and somebody's inbox.

Four filters cut across that column — All, Links, Files and To-do — answering four different people's questions from one record. Links is what a compliance-minded reviewer wants. Files holds exports and uploaded lists. To-do is the queue.

To-dos carry three states rather than two, and the middle one does most of the work here: active, deferred, dismissed. Most task systems offer only open and closed, which forces a decision nobody has made and turns every unanswered specialist question into a lie or a backlog.

Full-text search runs across every message, which is what turns a log into institutional memory. Someone asks in October whether the department ever approved a particular claim; searching the phrase returns the exchange, the date and the person who answered. The argument ends in thirty seconds.

An assistant tied to your project rather than to the open web

A chat window over a marketing dataset is useful only if it is bound to that dataset. Here the binding is explicit. Each question first meets a small routing model whose entire job is deciding how much of the project the answer needs. Four kinds of material are available — Search Console figures, ranking data, campaign state, custom data you supplied — and it attaches none, one, two, or at most three, by relevance.

Loading nothing is a legitimate outcome, not a fault. A question about what impressions mean needs no rows from your property; a question about which pages lost positions is meaningless without them. Keeping the two apart is why answers are not padded with irrelevant figures.

Answers stream a token at a time, and up to twenty messages are kept. That history is what makes the tool usable here, because the useful sequence is never one question. It is a rough worry, then a narrowing, then a second narrowing, until what remains is small enough for somebody with fifteen minutes. The assistant also takes keyword and URL lists in batches, which matters when a department's input arrives as a spreadsheet.

Linked accounts, one site at a time, and the right to take it back

Portfolio tools fail inside institutions for reasons that are never technical. The first request they make is impossible: give one team visibility over everyone's numbers. A department head who has spent a decade protecting a budget line will not have their performance appear in a view the development office can open, and is right to refuse.

Note the asymmetry. Sharing is granted per property and the history stays with the property: when the person who negotiated approved wording leaves, their access ends and the exchange approving it does not. That is the failure institutional web work suffers most — the knowledge was real, and it left with a laptop.

Two labels per site: the department, and the person who approves

Site tags act as a global filter: pick one and every surface narrows to the properties carrying it. The temptation is to model the organization chart. Resist it — the chart is wrong within a year, and was never accurate about who approves anything.

Flat labels beat a deep hierarchy for one reason: a site belongs to several categories at once, and hierarchies force a single parent. A trial microsite belongs to a therapeutic area, a review owner and a temporary lifecycle. A tree makes you pick one and lose the other two; flat tags let all three apply and combine at query time.

The approval-owner tag pays for itself. Filtering by it assembles, in one action, every open question waiting on one person across the whole estate — the difference between four scattered interruptions of a clinician's week and one prepared conversation.

What you hand to somebody who will never log in

Exports come in two shapes with two purposes, and confusing them is the commonest reporting mistake here. CSV and JSON run to 10,000 rows and exist for analysis: joining search data against a CRM, an admissions system or a referral log. PDF is capped at 250 rows, rendered server-side, carrying your logo and colors through a configurable builder.

Read the 250-row ceiling as guidance rather than restriction. A document meant for fifteen minutes of a specialist's attention should never have held 10,000 rows; if the monthly PDF will not fit, it is an export dressed as a report.

The visual vocabulary in the report builder is deliberately small: time series, metric cards, sortable and filterable tables, sparklines, and heatmaps split by country and device. Each earns its place here for a particular reason.

• Time series set against a year earlier. With a population that turns over between September and May, the only honest line is the same window twelve months back; a quarterly curve plots the academic calendar.
• Metric cards for the three numbers that survive the meeting. Clicks against last year, terms held in the top ten, questions awaiting this approver.
• Sparklines beside table rows. Direction per page without a second chart, so twenty program pages read in one pass.
• Country heatmaps on research-facing sections. Foreign attention on a protocol page is normal and often valuable; the heatmap sorts it from noise faster than a list.
• Device heatmaps as an audience check. A mobile majority on a page written for specialists means it draws a different reader than intended.

A week across nine properties

The routine below assumes one operator, nine properties across four departments, and nobody else logging in. Its output is counted in answered questions, not tasks closed.

• Monday — filter by tag, read the exits. One department tag at a time, look at which terms left the top ten. Recent losses are the cheapest recoveries available, and need no approval when the fix is technical.
• Tuesday — the Links view. Work through new placements with their donor Domain Authority and traffic; in regulated sectors this doubles as the review someone asks for later.
• Wednesday — narrow the questions. Turn four vague concerns into four one-sentence questions. The twenty-message history earns its keep here, because the first version of each is always too broad to answer.
• Thursday — the fifteen minutes. One prepared conversation per approval owner, five questions at most. Every answer enters the feed as a to-do: active, deferred with its reason, or dismissed with the objection recorded verbatim.
• Friday — one PDF per approver, not per property. Two branded pages showing this month against the same month last year, plus the questions waiting on them, get read; a portfolio report does not.

Questions that come up

The parts a workspace will never do for you

Everything above is bookkeeping in the best sense: it collects what happened, keeps it addressable, and puts a few well-formed questions in front of the people entitled to answer them. That is a large improvement on a spreadsheet of logins, and it remains bookkeeping.

Four things stay human, and they decide whether any of it was worth doing. Prioritization — that the trial microsite matters more this quarter than the news archive — turns on politics no tool observes. Cause-finding remains inference: a report shows nine terms fell and never mentions that a competitor hired three writers in March. Goal-setting means knowing what your organization could service if the demand arrived. And communication, where your counterpart holds a doctorate in the subject, means answering in their register rather than yours.

Automation covers the volume beneath those judgments. AutoSEO at $149 per month per domain handles keyword discovery and prioritization, link building across a partner network of more than 230,000 websites, and on-site suggestions. The FullSEO tier, at $500 per month per domain, adds hand-picked keywords with automatic fallback, placement against a Domain Authority target, human review of on-site changes, and a team of specialists, developers and writers. Expect four to eight weeks before movement is measurable, longer where every sentence passes a reviewer. Our service overview sets out how that cadence runs on an institutional estate, and the blog archive continues into analytics and indexing.

To see the shape of your own portfolio rather than argue about it, connect one property with Google sign-in, tag it by department and by approval owner, and put one question into the feed this week: open the Semalt dashboard and connect your first site. The most valuable thing a workspace produces here is not a chart. It is a dated line saying that on the fourth of the month, the person allowed to decide said yes.