Search data arrives as rows, not as answers. In Quebec City the gap between the two is wider than most dashboards admit, because the same word reaches you spelled two ways and the same service is named differently on two continents.
If you are assessing a Quebec City supplier from outside, or you are the one anglophone manager in a company whose customers, staff and search traffic all run in French, the reporting problem here is not a shortage of numbers. Once a property is verified, four figures arrive daily — impressions, clicks, click-through rate, average position — and a rank tracker supplies the competitive board alongside them.
What is missing is the instruction manual for a market where the working language of the customer is not the working language of the report. Taking that as the starting point, this piece walks the analytics sections of the rebuilt Semalt panel — fourteen views across Search Console and rank tracking — asking of each number what it measures, what it conceals alone, and where a person has to take over.
Accents are not a typography problem, they are a search problem
Open any query report for a site in the Capitale-Nationale and you will find pairs. cegep and cégep. quebec and québec. reparation and réparation. hotel vieux quebec and hôtel vieux-québec. Real people type both, and which form they use depends less on their French than on the keyboard in front of them.
Google is reasonably good at folding accents when it matches a query to a page, so both forms usually return similar results. Your reporting does no such folding. They sit as separate rows, each with its own impressions, clicks and average position, and each looking smaller than the demand it represents. A term that would be your third-largest query appears twice at ranks eleven and twelve, and is treated as marginal.
The duplication runs through the rest of the stack, with different consequences at each level. In the query data it splits demand. In the page title it changes how the result reads to someone who cares. In the URL it can produce two live addresses for one document.
Where the accent splits your demand
Both spellings collect impressions independently, so any ranking of your top terms is wrong until they are combined.
- Sum the pair, then rank
- Check whether one form converts better
Where the accent is judged
A title that reads Quebec in a French result tells a local reader something about the page before they click it.
- Write titles correctly accented
- Do not chase the unaccented query in the tag
Where the accent multiplies the page
Accented paths are percent-encoded in transit, and a legacy migration often leaves the plain form live as well.
- Pick one form as canonical
- Redirect the other, do not link to both
Where the accent is your own name
Companies with an accent in their trade name see their branded queries split the same way, which flatters or deflates brand reporting.
Search Console and SERP tracking do not answer the same question
The panel carries both, and teams routinely use them as two views of one dataset. They are not. Search Console records your own past: queries that produced an impression, the clicks those impressions earned, the positions your pages held when shown. Everything outside that history is invisible to it, including every term you rank nowhere for.
Rank tracking runs the opposite way round, photographing the results board at fixed intervals for a keyword set you choose, whether or not your domain is on it. Rival domains appear there with their authority and shared-term counts, and so do terms that have never once sent you an impression. About your clicks it knows nothing.
| What you want to know | Search Console | SERP tracking |
|---|---|---|
| Which spellings of a term actually reached us | Answers it directly | Only what you chose to track |
| Who else ranks for soumission rénovation Québec | Silent | Core use case |
| Did our page slip after the redesign | Yes, with a two-day lag | Yes, on the tracking interval |
| Which terms we are absent from entirely | Structurally invisible | The reason to run it |
| Whether a French-from-France site outranks us locally | Cannot see it | Visible in the competitor view |
The rule is short: Search Console tells you what happened, rank tracking what is possible. A finding needing both — fourth place on a real-volume term, still almost no clicks — is usually the most valuable one you get all month.
- Band movement. The dynamics view records entries into and exits from the top three, top ten and top thirty — far more legible than watching a position column twitch.
- Who else is there. Rival domains in your term space arrive with authority scores and shared-term counts, enough to separate a Lévis competitor from a national aggregator.
- The reporting delay. Console figures are two days behind. Presets allow for it; conclusions drawn inside the current week do not.
Impressions, clicks, CTR and average position
Each of the four headline metrics is honest on its own terms and misleading alone. Impressions count how often a URL was shown, not how often it was seen — a result at position 19 is an impression nobody scrolled to. Clicks are the only metric describing a human action, which makes them the most trusted and the least diagnostic: an outcome with no cause attached. A click count that fell can mean lost position, a lost snippet, a changed title, seasonal collapse in demand, or a competitor who wrote a better line.
Click-through rate catches experienced people out, because it is a ratio and both of its terms move. CTR falls when you gain impressions on weak terms, which is growth, and rises when you lose them, which is decline. Read as a quality score it gives the wrong instinct about half the time; read as a comparison between two pages at a similar position it is genuinely informative.
| What you see | Tempting reading | More likely explanation |
|---|---|---|
| Impressions up, clicks flat | Growth in progress | New visibility on low-intent or off-market terms |
| CTR down, position unchanged | The title got worse | The mix widened, or a new SERP feature took the top |
| Position improved, clicks fell | Reporting error | Seasonal demand fell faster than rank improved |
| Clicks up on one page only | The rewrite worked | Consolidation — another URL of yours stopped ranking |
| A term appears at position 40 | Not worth attention | Often the accented twin of a term you already rank for |
Query view and page view describe different things
The query view answers a question about demand: what were people looking for. The page view answers a question about supply: which of your documents Google chose to show. Reconciling the two is where most real findings live, and the arithmetic never balances neatly — one page serves many queries, and one query can be served by several of your pages on different days.
On a French-first site this has a specific shape. You will regularly find a blog post, a PDF or an old landing page holding the impressions a service page was written for. On sites that have been through a translation project, the URL shown is sometimes the English one for a French query — a signal worth acting on at once.
The product page loses to the glossary
Explainer content on assurance de dommages or REER outranks the page that actually sells the service.
- Check which URL holds the query
- Link the explainer to the product deliberately
Programme pages buried under news
A press release from three years ago holds the query that a current programme or clinic page should own.
- Compare publication date against demand
- Redirect or retire, do not duplicate
One query, four project pages
Portfolio entries compete with the service page for the same intent, splitting position between near-identical URLs.
The expired page still ranks
Last season's programme page keeps the impressions while this season's page has no history behind it.
Quebec French is not French French, and the data knows it
In most markets the country filter is decoration. Here it is load-bearing, for reasons of vocabulary rather than geography. French is searched by tens of millions of people who are not your customers, and their professional vocabulary is not the one used in Quebec City.
A homeowner in Charlesbourg asks a contractor for a soumission; one near Bordeaux asks for a devis. A Quebec employer offers assurance collective, a French one a mutuelle. Retirement savings here means REER; in France the acronym is different entirely. Food companies here work in transformation alimentaire where French firms say agroalimentaire. These are not stylistic variants — they are the words that do or do not appear in the query.
The failure mode is cheap to describe and expensive to live with. A keyword set is assembled from a general French-language volume source, most of whose data comes from France. It looks credible, it carries impressive volumes nobody in Lévis has ever typed, and it omits the terms your customers use. You then spend a quarter ranking for the wrong dictionary.
- Filter to Canada before you read anything. An unfiltered French keyword table is a table about France with a Quebec rounding error attached.
- Trust your own query data over any volume estimate. Search Console shows what was really typed at your domain, which is the only source that is definitionally local.
- Watch for France-based results in the SERP view. Where a French-from-France site outranks you on a local term, the fix is usually specificity rather than volume.
Device deserves a sentence too. The desktop and mobile heatmaps in the reporting section are not about layout; they tell you which population you are reading. A contractor checking a supplier from a truck types differently from a hospital administrator at a full keyboard, and both sit in your data under one average.
Date ranges, seasonality and the tourist trap
Every comparison claims that two periods are otherwise alike, and in Quebec City that claim is often false. Tourism and heritage demand peaks in summer and around the winter festival season, loudly enough to drown the steadier business demand underneath — insurance renewals, clinic appointments, construction quotes that continue at their own rhythm all year.
Two rules keep comparisons honest. Use a 28-day window against the same 28 days a year earlier to judge progress, and month-against-month only to detect breakage. And if the site contains a seasonal section, segment it out before comparing anything site-wide, or the season will be credited with work it did not do.
The panel's date presets already allow for the two-day reporting lag, which removes the commonest self-inflicted error: comparing a complete period against one still filling in. No preset removes the temptation to set two fourteen-day windows side by side and call the difference a trend.
Where a reading becomes work
A reading that changes nothing is entertainment. The bridge is the keyword pool, fed from Console history, live SERP sampling and seed terms you supply, then worked candidate by candidate. Each is approved, rejected or deferred on its own — and that is where your judgement about vocabulary and spelling enters the system.
AutoSEO — the automated tier
For a small team that wants reading, prioritisation and link building running without a weekly meeting.
- Candidates found for you. Terms surface and sort themselves by priority; no list has to exist before you start.
- Links placed automatically. The placement network behind it runs to over 230,000 partner sites.
- Page-level AI suggestions. Recommendations written against your own pages rather than a generic audit checklist.
- The analytics in full. Every Console and rank-tracking view, plus the live chat, with nothing held back for the upper tier.
FullSEO — automation with people behind it
For a site where wording is reviewed before it ships, which describes most regulated businesses in this region.
- You pick the terms. Selection is manual, with the automatic route filling any gap you leave — this is the tier where Quebec vocabulary can be enforced.
- Placements to a DR floor. You name the donor quality you want instead of inheriting whatever the default provides.
- Nothing ships unreviewed. Every on-site edit waits for sign-off, which matters where the French wording is the version that legally counts.
- People behind the software. Specialists, developers and writers sit alongside the automated part of the work.
Both figures are the published USD prices; the Canadian amounts are approximate conversions for orientation only. Whichever tier you take, the Stream assistant maintains one chronological feed per project: answers, generated reports, each new link with its donor's authority and traffic, open to-dos and campaign notices, all searchable as text. Behind each question a routing model picks up to three relevant data blocks — sometimes none. Our service pages describe how that fits a bilingual publishing process.
Frequently asked questions
Do I really need to worry about accents if Google matches them anyway?
Matching and reporting are separate. Google generally folds accents when deciding what to return, so both spellings tend to reach a correctly written page. Your query table does not fold them, so your measure of demand stays split until you combine the rows. The risk is not ranking badly — it is misjudging which terms deserve attention.
Should the URL carry the accent or not?
Either is workable; having both live is not. Accented paths are percent-encoded in transit, which makes them awkward in logs and shared links, so many teams standardise on the unaccented form. Whichever you choose, redirect the other permanently and let internal links and the sitemap carry only the canonical form.
Can I use French keyword volume data from a general tool?
Only with a country filter applied, and even then with suspicion. The bulk of French-language search volume comes from France, where the professional vocabulary differs. A term can look strong globally and be nearly absent in the Capitale-Nationale. Your own Search Console history is the more reliable starting point precisely because it is local by definition.
Why do Search Console and the rank tracker disagree about my position?
Because the two are measuring different things. One reports an impression-weighted average across the occasions your pages were genuinely displayed. The other photographs the board at set intervals for a fixed list, impressions or no impressions. Both can be accurate simultaneously, and neither verifies the other.
How soon after a change should I expect to read a result?
Expect the first measurable shift somewhere in the four-to-eight-week range, counted from publication rather than from approval. Where French copy has to clear a review step, budget a full quarter between spotting the work and trusting the number that comes back.
The point at which a person has to decide
The reporting can establish that two spellings of one term together outweigh everything else on the site, that a discontinued programme page is sitting on a query the current one should hold, and that a competitor writing for readers in France ranks above you on a word your customers use differently. All three are findings. None of them is a decision.
The decisions are whether that page is worth rewriting this quarter, whether the vocabulary your marketing department prefers is the one your customers type, and whether a modest-volume term is worth pursuing because the people searching it sign the contracts. No dashboard holds that.
What a unified panel takes away is the excuse. Background workers hold dashboards current with no manual refresh; a split defined once is reproduced rather than rebuilt each month; exports reach 10,000 rows in CSV or JSON and 250 in PDF; and the competitor listing contributes the outside view your own history cannot hold. Together, these analytics sections reduce a heap of rows to roughly five sentences a month worth saying aloud.
To see those views against your own property before committing to anything, connect a verified site and open the dashboard — one Google consent flow covers Search Console and Analytics, and the first honest reading takes an afternoon. The series continues on our blog.