In the National Capital Region the useful cut through search data is not the country and rarely the city — it is the language of the query. This article works through the Search Console and SERP views in the Semalt panel from that starting point.
Most reporting problems in Ottawa are not caused by missing data. A verified property hands over impressions, clicks, click-through rate and position daily, and a rank tracker adds the competitive picture beside it. The difficulty is that the standard slices — country, device, month against previous month — were designed for markets that do not look like this one.
An association serves members in both official languages. A federal supplier publishes every page twice. A firm in Ottawa answers to a client in Gatineau who searches in French on Monday and reads the English page on Tuesday. Run a country filter across that and you get one bar labelled Canada. The fourteen analysis views inside the rebuilt Semalt panel deserve better questions.
Why the country filter is the least informative slice you own
A .ca domain serving the capital region shows Canada as an overwhelming majority of impressions, with a thin tail from the United States. That distribution is stable, it moves for reasons you cannot influence, and no decision has ever come out of it. Check it twice a year, and otherwise leave it closed.
The division that explains your traffic runs through language instead. Same organisation, same service, same week: one set of searchers types federal procurement consultant Ottawa and another types the French equivalent, and those populations often overlap in the same building — frequently in the same person, who searches in French at home and in English at the office because that is what the workstation is set to.
This catches teams out, because neither Search Console nor a rank tracker gives you a language dimension. Country and device are built in; language is not. You construct it yourself, from the query strings and the page paths, and the moment you do the report stops being generic.
Associations and professional bodies
Membership, certification and event pages are searched in both languages by the same audience, with no geographic separation.
- Split by query language first
- Compare funnels, not the total
Suppliers to federal departments
Capability and past-performance pages must exist in both languages before they can be cited in a submission at all.
Professional-services firms
Practice-area pages draw English queries from Ottawa and French ones from Gatineau, often for the identical service.
- Watch the CTR gap between versions
- Check which URL Google chose
Universities and institutes
Long, specific programme queries, where one missing translated page removes an entire branch of demand.
Search Console and SERP tracking answer different questions
The two sources in the panel are routinely treated as interchangeable, and they are not. Search Console is the record of what happened to you: queries that produced impressions, clicks those impressions earned, positions your URLs held when shown. It is your own history, and nothing outside it exists.
SERP tracking is the opposite instrument. It samples the results board as it currently stands for a defined keyword set, whether or not you appear on it. That is where competing domains become visible, with domain authority and shared keywords recorded, and where you find terms you rank nowhere for and therefore never see in Search Console.
| Question | Search Console | SERP tracking |
|---|---|---|
| Which French queries reached us last month? | Answers it directly | Only for tracked keywords |
| Who else ranks for our tender-related terms? | Silent | Answers it directly |
| Did our English page lose position after the rewrite? | Yes, with a two-day lag | Yes, on the tracking schedule |
| Which terms do we not appear for at all? | Structurally invisible | The core use case |
| Which URL did Google choose for a bilingual pair? | The pages view | Per tracked keyword |
Search Console analytics is organised as eight views: a dashboard across every verified site, a site overview, keywords, pages, devices, countries, traffic and keyword dynamics. SERP adds six: a cross-site ranking score, average position, keyword counts, top keywords by volume, best pages and competitors. Background workers keep both current, so nothing on screen waits on a manual refresh.
Impressions, clicks, CTR and average position on their own
Each of the four headline figures is honest in isolation and misleading in isolation. On a bilingual property each fails in a specific way that a single-language site never encounters.
- Impressions count exposure, not interest. Rising impressions with flat clicks usually means you have started matching queries that were never yours.
- Clicks are real but unattributed. A click says someone chose your result. It does not say which language version they wanted, or whether they went looking for the other one.
- CTR depends on position before it depends on your copy. Falling with position, it is arithmetic. Falling while position holds, it is a signal about the title, the description or a rival's snippet.
- Average position averages things that should not be averaged. Third for twenty queries and thirtieth for twenty others reports as sixteenth — a position you hold for nothing.
The correction is the same in each case: never read one of the four without another beside it, and never read a blended figure when a split is available. In practice that means position bands rather than an average — how many keywords sit in the top three, the top ten, the top thirty — and the movement between them.
That movement is what the keyword dynamics view is for: rather than reporting where you stand, it records entries into and exits from each band. On a bilingual property this is the earliest reliable warning you have, because a translated page losing ground shows up as a band exit weeks before it shows up as lost clicks.
The query view and the page view describe different problems
The keywords view answers a demand question: what people ask for, and how much of that reaches you. The pages view answers a supply question: which of your URLs Google chose to show. The two come apart most visibly on bilingual sites, where the supply side has a structural problem the query side cannot see.
The characteristic case: a service published at an English URL and a French URL, both accurate. For a query either could answer, Google picks one and shows it consistently. If it picked the English page for a French query — which happens when the French version is thinner, newer or linked from fewer places — the keywords view reports that query performing acceptably while the searcher lands in the wrong language and leaves.
| Symptom | Visible in | Likely cause | What it points to |
|---|---|---|---|
| French query, English landing page | Pages view | Weak or late French version | Internal links and depth on the French page |
| Both versions ranking for one query | Keywords view | Unclear language signals | Language annotations and canonical setup |
| Strong impressions, poor CTR on one side only | Keywords view | Translated title reads as machine output | Rewriting titles, not the body copy |
| Tender PDF outranking the HTML page | Pages view | The document accumulated the links | A landing page hosting the document |
| One version updated, the other stale | Pages view | Review stalled mid-pair | Publishing rule, not an SEO fix |
None of these show up if you read only the keyword list, and the last two are not SEO problems at all — they are publishing problems surfacing as SEO symptoms. Naming them correctly is most of the value of the pages view.
Language, device and the splits that are actually built in
The device split is genuinely useful here, and it behaves unlike the consumer pattern. Government-adjacent audiences search largely from managed desktop machines during business hours. A mobile CTR gap on a supplier site is worth investigating; a mobile traffic share that looks low is often just the audience.
The country split, as established, does very little. Its one legitimate use is confirming that a sudden impression spike came from outside Canada — which usually means you have started matching a query that has nothing to do with you.
Site tags do the second half of that work. A tag acts as a global filter across the panel, so one applied to the French section of a property carries through the dashboards and into exports without rebuilding the segment. Where several properties are managed together — a national association with provincial chapters — the tenancy model shares individual sites to named email addresses rather than the whole account.
Beyond a quick look, exports are where the language split gets done properly: heatmaps and sortable tables cover the built-in dimensions, and a CSV with a language column added covers the one that matters. The reporting and export tooling renders PDFs server-side with your own logo and colours, which is what a steering committee will accept.
Why last year's same month beats last month
Month-on-month comparison is the default in every analytics tool, and here it is close to useless. Demand around the capital follows the federal fiscal year — April to March — and the tender cycles attached to it. Spending concentrates as year-end approaches, procurement activity clusters, and the quiet stretches are structural rather than a sign that anything is wrong.
A supplier comparing February against January sees a rise and congratulates itself on work that had nothing to do with it. The same supplier comparing August against July sees a fall and starts changing pages that were performing normally. Both readings are artefacts of the calendar.
The headline comparison
The same month one year back holds the fiscal position constant, so what remains is what changed about you.
- Use it for progress reporting
- Needs one clean year behind it
The breakage detector
The short window spots something broken this week, and should never be presented as a trend.
- Two-day data lag is built in
- Pair it with keyword dynamics
The panel's date ranges cover the routine cases with 28 and 90-day presets, and the two-day reporting delay is already accounted for, so a range ending today does not silently include two empty days.
One caution: the annual comparison only works once you have a clean year behind you. A site relaunched or newly translated in the interim is not comparable to itself, and saying so is more useful than a percentage nobody can act on.
From a weekly reading to work that survives review
Here is where most analytics advice fails this audience outright. Recommendations written for a fast-moving site assume a page can be edited the same afternoon. In an association or a federal supplier it cannot. Changes go through committee, and anything published in one language triggers an official-languages review of its counterpart before either can ship.
That changes how analysis should be packaged. Reactive one-off fixes are the wrong unit of work: the overhead per change is fixed and high, so one title tweak costs the same in review time as thirty. The right unit is a batch — a ranked, dated list assembled over a reporting period and submitted once, in both languages.
- Pair every change. An English edit without its French counterpart stalls at review. Write both into one ticket so the pair moves together.
- Attach evidence, not conclusions. Committees approve what they can verify. The position band and the query list travel better than a recommendation alone.
- Set thresholds before the reading. Decide in advance what counts as a real move, and keep everything below that line out of the report.
- Plan against the review calendar. If sign-off runs four weeks, work identified in September lands in October. Say so rather than letting someone assume otherwise.
The campaign side of the panel is built around the same two speeds, so choosing between its tiers is really a choice about who reviews what.
AutoSEO — the automated track
For organisations whose pages are stable and whose work is mostly discovery and off-site.
- Automatic keyword discovery. Candidates come from Search Console, live SERP data and your own seed terms; each is approved, rejected or deferred individually.
- Automated backlink building. Placements run against a partner network of more than 230,000 sites.
- On-site AI suggestions. Recommendations arrive as proposals, which suits a team routing them through review anyway.
FullSEO — with a human review step
For government-adjacent work where nothing changes without a named person signing it off.
- Manual keyword selection. You choose the terms, with automatic fallback if the list runs short.
- Manual placement with a DR target. Donor quality is set deliberately rather than accepted as configured.
- Human-review mode for on-site changes. Every edit waits for approval — the only mode compatible with an official-languages review.
- A team behind the automation. SEO specialists, developers and writers, not the tooling alone.
Between cycles, the Stream assistant keeps the running record: AI answers, automatic reports, newly placed backlinks with donor authority and traffic, to-dos and campaign news in one chronological feed with full-text search. A router model decides per question which data blocks to load, between none and three, so a question about French keyword performance pulls the Search Console and SERP blocks and leaves the rest alone. Our service pages set out how that fits a review-bound process.
Frequently asked questions
Can I see which language a query was made in?
Not as a built-in dimension. Search Console reports country and device; language has to be derived from the query text and from the path of the page that received the impression. Once you have classified queries once and tagged the French section, the split becomes a saved filter you reuse rather than rebuild.
Why do Search Console and the rank tracker disagree on my position?
Because they measure different things. Search Console averages the positions at which your pages were shown, weighted by impressions. SERP tracking samples the board for a fixed keyword list on a schedule, whether or not you received impressions. Both are correct; neither checks the other.
How long before campaign work shows up in the data?
First measurable movement typically appears after four to eight weeks, and that clock starts when the change goes live, not when it is approved. On a month-long review cycle, plan a full quarter between identifying work and reading a reliable result.
Should we track French keywords separately from English ones?
Yes, as separate tracked sets. The competitive field is often not the same — a French query here can surface Quebec-based organisations that never appear on the English equivalent — and averaging the two produces a ranking score that describes neither.
Is month-on-month reporting ever appropriate here?
For detecting breakage, yes: a sharp drop inside a 28-day window is worth investigating at once. As a measure of progress, no. Use the same month a year earlier, which holds the fiscal cycle constant and shows what changed about your site.
Where the numbers stop and judgement starts
Analytics will tell you that a French page is losing ground in the top ten, that a tender document outranks the page it should sit on, and that your English CTR fell while position held. It will not tell you whether that page is worth rewriting this quarter, or whether the committee has capacity before the fiscal year turns. Those are the decisions, and they are yours.
What the panel does is remove the excuses for not making them: data arrives current, splits are reproducible, exports suit a room that wants evidence, and the competitor view supplies the outside picture your own history cannot contain. What it cannot know is that a page is bound by a policy nobody wrote down.
To see the eight Search Console views and the six SERP views against your own property before committing to any of this, open the panel and connect a verified site — the Google consent flow covers Search Console and Analytics in one step, and the first reading costs an afternoon. The rest of this series continues on our blog.