A search dashboard records what happened. It does not say whether the cause was the season, the field of competitors, or something your own team shipped last Tuesday. On this coast, separating those three is most of the work.

Every verified property hands over the same four columns — impressions, clicks, click-through rate, average position — plus filters and a date picker. What it does not hand over is a decision. Between the two sit three choices: which view to open, which slice to trust, and what to compare against.

Demand here runs on two clocks most reporting templates ignore. One is the season: a rental fleet in Deep Cove, a hotel on the harbour and a gear shop in Kitsilano all live through a curve nobody controls. The other is the announcement: a change to property rules moves search volume in an afternoon.

The question behind every chart. Before reading a number, say out loud what would have to be true for it to change your plan. A figure with no answer to that is context, not evidence.
Starting point · Two clocks

Why the month-on-month reflex fails on a seasonal coast

Almost every analytics tool defaults to the previous period: twenty-eight days against the twenty-eight before. The arithmetic is sound and the conclusion frequently worthless, because on a seasonal curve the previous period is a different business.

Take a shop renting paddleboards and selling wetsuits. Its September always looks worse than its August, whatever anyone did to the website, so the same team is congratulated in spring and investigated in autumn for reasons unrelated to their work. Against the same month a year earlier, the seasonal shape cancels out and what remains is the part you influenced.

  • Month over month measures the calendar, not the campaign. On a swinging curve the season dominates the comparison and buries the question you asked.
  • Year over year isolates the change you made. Same month, same weather, same visitor patterns, one year of work in between. Any gap left over is yours to explain.
  • Some queries ignore the calendar entirely. Property and rental-rule searches spike on news, so the useful comparison is the week before an announcement against the week after.
  • Brand terms drift slowly and hide everything else. If your name carries a third of your clicks, a total including it looks stable through a serious non-brand loss.

None of this argues against short windows, only against a single one. Twenty-eight days catches problems while they can still be fixed; year on year says whether the business is further ahead.

Where year-on-year breaks down. It assumes last year was normal. A closure, a rebuild or one unusual season poisons the baseline, and no dashboard will warn you. Someone has to remember what happened.
Two sources · Two questions

What Google already shows you, versus what the market looks like

The Search Console side of the panel is a record of your own appearances. It knows which URL was shown for which query, on which device and in which country. It knows nothing about who else was on that results page — the blind spot behind most wrongly explained declines.

Rank tracking is the other half. It samples the results page on a schedule with standardised queries, whether or not you appear, and reports the field: which domains hold the places, how strong they are, how much of your keyword space they share. Eight Search Console views and six SERP views sit in the same analytics workspace, which makes the pairing practical rather than theoretical.

Question you are askingSearch ConsoleSERP tracking
Have our clicks fallen?Yes, with impressions and CTR beside themNo click data at all
Who took the position we lost?InvisibleBy domain, with authority and shared keywords
Which of our pages did Google pick?Exactly, per queryOnly the URL that ranked
Is a booking platform sitting above us?Not shownYes, it appears as a competing domain
Personalised results?Yes, this is real user dataNo, standardised queries

The two will disagree about position, and neither is broken. One averages what real people saw from their own locations and devices; the other runs a neutral query. Treating one as an audit of the other wastes an afternoon.

8
views of your own data
6
views of the field
28
day portfolio trend curve
2
day data lag, built in
Measurement · Four columns

Impressions, clicks, CTR and position: how each one lies on its own

The four headline figures are not four measurements of one thing. They are four partial views, each with its own failure mode. The habit worth building is never to quote one without the figure that constrains it.

NumberWhat it saysWhat it hidesNever read it without
ImpressionsHow often you were shownWhether the query had any commercial valueClicks, and the query itself
ClicksHow much traffic arrivedWhether demand or your ranking changedImpressions over the same window
CTRHow convincing the listing wasThat it falls automatically as you rank for moreAverage position for that row
Average positionRoughly where you satA spread of very different placementsThe count of terms inside the top ten

Impressions flatter most easily. A tour operator that begins appearing for hundreds of informational questions about weather, parking and ferry times posts an impressive chart and takes no bookings from any of it. Impressions without clicks describe reach into an audience that was never yours.

Clicks are the honest number and the slowest. They cannot separate a ranking loss from a demand drop, and in a seasonal market those look identical. Set them beside impressions and the ambiguity narrows: if impressions held and clicks fell, the market was there and your listing lost the argument. If both fell, demand moved.

A useful reflex. When clicks fall, look at impressions first. Falling impressions point at ranking or seasonality; flat impressions with falling clicks point at your title, your description, or something new above you.

Click-through rate catches growing sites in particular. Its denominator grows every time you start appearing for a new, weaker query, so an expanding site watches site-wide CTR fall while every page performs as well as before. At query level, with the position beside it, CTR is diagnostic; as a site-wide average it mostly measures how broad your footprint has become.

Average position is the worst offender. It weights by impressions rather than value, so a definitions page far down a high-volume query moves the number more than the commercial term that produces enquiries. It treats every place as equal, when eleventh to sixth is a business event and thirty-fourth to twenty-ninth is nothing. And it punishes growth: every new query enters low and drags the average down.

No recorded position is the true one. Location, device, personalisation and a layout that shifts through the day turn every figure in either source into an approximation. Direction is reliable; exact standing never is. A threshold built on a decimal place fires on noise.

The replacement is counts rather than averages. The keyword views with their position history track which terms move into and out of the top three, top ten and top thirty, and those counts compare cleanly because a term arriving at position forty lands in the outer group instead of dragging a mean.

Attribution · Query or page

The query view and the page view describe different problems

These two views cover the same traffic and answer opposite questions. The query view asks what people wanted; the page view asks which URL Google sent them to. Most reporting only opens the first, which is why the second is where the surprises live.

Demand

The query view

What the market asked for, in its own words, including phrasing you would never have written yourself.

  • Finds seasonal vocabulary shifts
Supply

The page view

Which URL earned the impression. The only place a cannibalisation problem is visible at all.

  • Reveals swapped URLs
Conflict

One query, two of your pages

A seasonal landing page and an evergreen category page trading places month by month, each holding the other down.

Breadth

One page, many queries

A strong guide absorbing dozens of terms it was never built for — an argument for splitting it, not a success.

The local version involves a page built for a season. A hotel publishes summer packages in May; by November it still outranks the general rooms page for queries that now carry winter intent, and the visitor lands on an offer that ended in September. Nothing in the query view shows this. The page view shows it in one line.

Grouping matters as much as the rows. A retailer with a page per North Shore and Fraser Valley location should read the whole set together, because Google often swaps which one it favours for a city-level query — a rotation a page-by-page reading records as two unrelated moves.

Splits · The border and the screen

Country and device: two slices that change the answer

Search does not see the border. Seattle is two hours away, the results pages overlap heavily, and a Lower Mainland business on a generic query is often up against a Washington domain and a listings platform rather than anyone it has met. Those impressions land in your totals whether or not you serve that market.

So the country split is not an optional refinement here. Left blended, a healthy impression trend can consist largely of Americans who will never book, while the Canadian slice underneath declines quietly. The country and device heatmaps in the reporting section of the panel exist for this reading, and it usually changes what counts as a good month.

  • Decide once whether US demand counts. For a cross-border logistics service it plainly does. For a Burnaby dental practice it does not. Either way, report it on its own line rather than inside a total.
  • Country is not language. French results differ between Quebec and France, and both can land under one country row. If you serve francophone visitors, that split runs through URL structure, not the country filter.
  • Watch which competitor set you are in. On travel and accommodation queries the field is often platforms rather than businesses, and losing a place to an aggregator is a different problem from losing it to the operator down the street.
  • Check devices inside each country slice. Blended numbers hide gaps of several positions on one query, and that gap is where template and speed work is justified.

The device split has a seasonal character of its own. Summer demand skews towards someone on a sidewalk with a phone wanting an answer in twenty minutes; winter planning skews towards a desktop session with a decision weeks away — the same query meaning two different things.

A split worth keeping permanently. Two saved readings — domestic mobile and domestic desktop — explain more decisions over a year than any blended chart in the same report.
Time · Comparisons that hold

Comparing dates: the season, the announcement and the twenty-eight-day window

Date ranges come with presets at twenty-eight and ninety days, and the two-day reporting lag is accounted for. The discipline is in choosing what each window is measured against, because that choice decides the conclusion before anyone reads a chart.

SituationComparison to runWhat it tells you
Tourism, hospitality, outdoor retailSame month, previous yearProgress with the season removed
Any site, weekly operations28 days against the prior 28Fresh problems, early enough to fix
Property or rule-driven queriesWeek before against week after the newsThe size and shape of the spike
After a template or content releaseFixed date forward, not a rolling windowWhether the change did anything
Reporting to owners or a board90 days, plus the same 90 last yearTrend rather than weather

Policy-driven demand obeys no calendar. When a rule on short-term rentals, foreign buyers or development permits is announced, the related queries move within hours, run hot for a week or two, and settle at a level that may be permanently different. A monthly comparison averages that spike into invisibility; a before-and-after around the announcement date captures the surge and the new baseline.

Year-on-year work needs the history to exist. Search Console keeps a finite window, so a comparison you want next spring is one you capture now. Exports run to 10,000 rows in CSV and JSON and 250 in PDF, rendered server-side, which makes a monthly archive of the query and page tables a five-minute habit with a payoff twelve months later.

Archive before you need it. One export a month, stored somewhere dull and permanent, is the only way to compare a January to a January once the source window has rolled past it.
Operations · Who reads, who acts

Turning a weekly reading into work someone owns

A routine that survives the busy season is short and has its limits written down in advance. Thresholds set after the data is seen bend towards whatever the week produced.

Alarm

A commercial term leaves the top three

Same-week investigation. Those places carry most of the clicks, and the loss reaches revenue before it reaches the weekly total.

Pattern

Three or more terms leave the top ten

One term slipping is variation. A cluster in one week points at a single page, template or release nobody logged.

Snippet

CTR down a fifth, position flat

The listing stopped convincing people. Rewrite the title and description before touching anything structural.

Field

An unfamiliar domain in the top ten

Check it in the competitors view before promising gains. A platform and a rival across the border need different answers.

What happens after the reading is a separate decision, and it comes down to whether a person approves each keyword and each on-page change before it happens. The two campaign tiers in the My SEO section divide on exactly that line.

My SEO · Tier 1

AutoSEO — the reading and the routine run themselves

For an owner or a one-person marketing team who wants the analysis maintained without maintaining it.

149 USD / month · per domain
  • Keyword discovery and prioritisation. The pool is fed by Search Console, live SERP results and seed terms you supply, and every candidate is approved, rejected or deferred one at a time.
  • Full analytics access plus automation. All fourteen analysis views, alongside automatic backlink building and on-site suggestions from the AI.
149 USD
per month, per domain
4–8
weeks to first movement
My SEO · Tier 2

FullSEO — selection and review by hand

For sites where the seasonal calendar and the wording of an offer matter too much to be automated.

500 USD / month · per domain
  • Manual keyword choice with automatic fallback. Terms are picked by hand, placements aim at a chosen domain rating, and a team of SEO specialists, developers and writers stands behind the automation.
  • Human review before anything goes live. Proposed on-site edits wait for approval, which keeps a seasonal page from being rewritten in the middle of the season it was built for.
500 USD
per month, per domain
230,000+
partner sites for placements

Both figures are US dollars, as the Semalt list quotes them — approximately 200 and 680 Canadian dollars a month at recent rates. The tiers share one multi-tenant structure: site tags as a global filter, grouped Google accounts, and individual sites shared to a named address, so an owner and an outside agency work from the same figures.

Frequently asked questions

Should a seasonal business ever use month-on-month comparisons?

Yes, for operations: a twenty-eight-day comparison catches a problem while it can still be fixed. It should not be the number in a report about progress, because on a seasonal curve it mostly measures which month it is.

Why do Search Console and rank tracking give different positions?

Each measures something else. Search Console averages the placements real people saw, personalisation and location included; rank tracking runs a standardised query on a schedule. Both can be correct at once, and neither verifies the other.

Our impressions include a lot of US traffic. Is that a problem?

Only if it sits inside a total you make decisions from. Decide whether that market is one you serve, then keep it on its own line in the countries view either way. Cross-border logistics businesses usually want it; a local service business finds it inflates everything.

How long before a change shows up in the data?

Search Console runs about two days behind, so the last two points on any curve are incomplete. For campaign work, first measurable movement typically appears after four to eight weeks. Judging a release inside the first fortnight produces false alarms.

What do we do when a listings platform outranks us on our own service?

Treat it as a field problem rather than a page problem. The competitors view shows which domains share your keyword space and how strong they are, which reveals what is realistically winnable and where you are better off being present on the platform as well.

Where the numbers stop and judgement starts

The honest limits. Very low-volume queries are withheld from detail rows to protect user privacy, so those clicks sit in your totals and in no keyword row. Position is always an approximation, and the keyword-dynamics view reports that six terms fell while staying silent on why. None of it weakens the analysis; it sets the limit of what you can defend under challenge.

Everything a panel does ends in the same place: it describes what happened and how the field is arranged. It cannot tell you that the drop began the week a ferry schedule changed, that last February was unrepresentative because of a closure, or that the American domain now above you serves a market you never intended to enter. Those facts live with people.

What the data does reliably is narrow the plausible explanations until few remain, in an hour rather than a quarter. Impressions against clicks separates demand from ranking. Query against page separates a market shift from cannibalisation. Country against device separates the audience you want from the one that arrived. Year against year separates your work from the weather.

More on the subject sits in our blog, with the implementation side set out under our services. To see both data sets against your own domain, connect your property in the Semalt dashboard and give it a full season. The reading that changes a plan here is rarely the average position — it is the year-on-year line for a single month, next to the list of domains that were not on the page twelve months ago.