In Washington the person receiving your marketing report probably spends the rest of the day weighing evidence for a living. That decides which figures you may show them, and how each has to be labeled first.
Verify a property and you are handed more measurement than a small firm can consume. Access was never the difficulty. The difficulty is deciding which figures still mean something once they land in a meeting where somebody is paid to ask where they came from.
Your report goes to people who read for a living
Consider who sits in the room. A trade association's communications director came out of committee staff. The managing partner of a K Street firm has read depositions for twenty years. A think tank's outreach manager is surrounded by researchers who publish with footnotes and expect them back. Evaluating a source is not a skill these people admire from a distance. It is the job.
That inverts most persuasion advice. The usual counsel is to lead with a confident claim and cut the qualifiers, because qualifiers dilute. Here the unqualified claim draws the challenge, and once found loose you get no second reading. Hedged, sourced and specific beats bold and vague with this audience, every time.
- An unattributed number reads as invented. Not doubted — discarded, along with the rest of the page it sat on.
- Precision reads as honesty. “Up 18 percent against the previous 28 days” survives scrutiny; “significant growth” invites the question that ends the meeting.
- Overstatement is permanent. A reader who catches you inflating once discounts your next twelve reports without mentioning it.
- The qualifier is the credibility. Naming what a figure does not cover is how you show you understand what it does.
Two records that were built for different questions
Two families of screens sit side by side and look as though they describe one quantity. They do not, and the instinct to call one of them broken when they disagree costs an afternoon each time.
The Search Console side reports what Google's serving machinery recorded on a domain you have verified: how often your listing was shown, and how often somebody picked it. Inside its walls it is authoritative; outside them it is silent. When a Bethesda competitor overtakes you, nothing lights up. Your line bends downward and you supply a story to explain it.
Rank tracking works the opposite way. It reads the results page from outside under fixed conditions, so it can name the domains occupying the slots, attach a Domain Authority to each, and count the terms you hold in common with a rival. Past the click it sees nothing. Eight views sit on one side of that line and six on the other, and the case for keeping both records under a single login is that whatever one answers cleanly, the other cannot answer at all.
| Question in the meeting | Which record answers it | Why the other one cannot |
|---|---|---|
| Did anyone choose us? | Clicks in the logged record | An outside check sees no decisions |
| Who else holds this term? | Competitors view, with authority scores | Rivals leave no mark in your own log |
| What do people actually type? | The query breakdown | Volume figures are modeled, not observed |
| Worth contesting? | Keywords by search volume | Impressions are exposure, not demand |
| Did we drop, or the page? | Both together, never one | Half a cause and a story |
When you cite a figure, name which of the two produced it. “From the logged record, 28 days to Tuesday” and “from an external check of the results page” are different claims with different reliability, and a reader who works with evidence wants to know which you are making.
Four numbers across the top, four ways to be wrong
Impressions, clicks, click-through rate, average position. Every one of them is defensible; every one of them, read by itself, sends you somewhere unhelpful. The four fail in different ways, so they need taking apart separately.
Impressions
A tally of appearances on results pages somebody loaded. Not of reading, not of interest, not of intent.
- Rises when you appear for terms you do not want
- Falls when a results page adds a feature above you
Clicks
The one figure recording a person deciding something. Also the smallest, and small numbers move for trivial reasons.
- Eleven to seventeen is not a 55 percent gain
- One conference week can carry a whole month
Click-through rate
A ratio, so either half of it can drive the change. A rising rate on a shrinking base is a worse position, not a better one.
- Rewards shedding irrelevant exposure
- Meaningless below a few hundred impressions
Average position
A mean across every query, device and location where the listing appeared. It describes no search anyone actually ran.
- Improves when weak terms stop appearing
- Hides a first-place term inside a mediocre mean
None of the four is a result. As a set they describe a shape: exposure up while decisions stayed flat, or exposure flat while decisions climbed. Two situations needing two responses — and either can produce the same unhelpful movement in the average. Reading the four together is the only defensible way to quote any of them.
Query rows and page rows are not two angles on one thing
Two breakdowns of the same period refuse to add up, and every analyst who checks discovers it within a week. It is not a defect: query rows and page rows count different objects. One search can involve several pages; one page serves hundreds of queries.
They also part ways because queries typed by very few people are withheld to protect those people. The query table is a large, useful, deliberately incomplete sample, not a census. For a specialist firm — a boutique handling debarment matters, a vendor with one compliance product — much of the valuable traffic sits below that threshold and is never listed. The page view still counts it.
| Breakdown | What one row represents | Best question for it | Known limitation |
|---|---|---|---|
| Queries | A string somebody typed | What words does this market use? | Rare queries withheld |
| Pages | One address on your site | Which documents carry the work? | Silent on why |
| Devices | A class of hardware | Where is this audience reading? | Three coarse buckets |
| Countries | An inferred location | Are we seen abroad? | Domestic detail collapses |
- Query rows give you vocabulary. How associations, agencies and counsel actually phrase what you sell, which is rarely how you phrase it.
- Page rows show what is load-bearing. Often an old explainer nobody has touched in three years rather than the service page the site was built around.
- Cross them before concluding. Filter to one page, read its queries, and you learn what that document is taken to be about — frequently not its stated subject.
- Never show the two totals together. They differ, someone notices, and the meeting becomes a discussion of sampling.
Three jurisdictions collapsed into one metro-level line
The country breakdown is near useless for most firms here and quietly important for a few. A local trades business sees one row holding 97 percent of everything. An association with international members, a university with applicants abroad, a consultancy with overseas offices — for those the same view is where the first real finding appears.
The device breakdown carries more weight. A page read on a desk monitor and the same page read on a phone on the Orange Line are one document meeting two levels of patience. High phone share of exposure with low phone share of decisions is a specific problem in a specific place.
Then there is the split that matters most here and appears in no menu. Washington is not one market; it is three legal regimes sharing a subway map. The District, Virginia and Maryland differ on licensing, on how services are taxed, and on which registrations a business must hold before it can work. Someone in Silver Spring and someone in Arlington typing the same three words are asking different questions, and a metro-level report averages them into one line correct for neither.
| Where the reader is | What differs for them | What the page must not do |
|---|---|---|
| District of Columbia | City registration, local licensing, District tax treatment | Quote a suburban process as though it applied |
| Northern Virginia | Commonwealth licensing, separate filings, county requirements | Assume a District credential transfers |
| Suburban Maryland | State licensing, its own registrations and tax rules | Treat “the DMV” as one regulatory area |
| All three at once | Nothing — no such reader exists | Write for an average of the three |
Date comparison, and the two days that are always missing
Nothing here means anything without a window attached, and the window has a known gap. The most recent two days are incomplete by design, which the presets account for. Quote a seven-day figure ending today and roughly a third is missing; the same figure quoted next week will contradict you, in writing, in front of the person you gave it to.
The presets exist because they are defensible. Twenty-eight days absorbs the weekly rhythm of a market that goes quiet on Friday afternoons and quieter still when the city empties in August. Ninety days shows direction. An arbitrary eleven days against an arbitrary nine produces a number nobody can check, including you.
Comparison needs a like-for-like frame: equal length, equal weekends, similar weeks. Keyword dynamics is the better lens anyway. Instead of a movement in a mean it reports which terms entered or left the top three, the top ten and the top thirty — a countable claim, and countable claims are what survive a careful reader.
A report that cannot name its source and its date is worse than none
Worse, not merely weaker. A report with no provenance still circulates, still gets quoted back at you months later, and still counts as your professional opinion. If the figure came from a screenshot of unknown vintage, the damage is not to that figure. It is to everything else you have said.
Know the export ceilings before promising anyone a file: 10,000 rows in CSV or JSON, 250 in a server-rendered PDF. The builder takes your own logo and colors — useful less for vanity than because a branded, dated document can be filed and reopened a year later by someone who still knows what they are looking at.
AutoSEO — the automated tier
For a firm that needs the reporting to run without anyone tending it daily.
- Candidate terms surface on their own. Drawn from logged search data, live results pages and seed terms you supply; you approve, reject or defer each one.
- Link placement proceeds unattended. Across a partner network numbering over 230,000 sites.
- Both records reported in full. Every view on either side, with on-site suggestions and a live assistant alongside.
FullSEO — automation with people behind it
For organizations where a change to a public page has to pass a human first.
- You pick the terms; automation covers the rest. Manual selection with an automatic fallback behind it.
- Placement with an authority target. Manual backlink work against a stated Domain Authority floor rather than whatever is available.
- Human review before anything ships. On-site changes queue for approval — the only workable arrangement when counsel signs off on public wording.
Both tiers offer add-on placements — Wikipedia slots at $10 each, private-network slots at $1 each — and both feed the same assistant, Stream: a chronological feed of answers, generated reports, newly placed links with their donor authority and traffic, and open to-dos. A router decides per question which blocks of data to load, between none and three, so an answer about last month's clicks comes from the record rather than being improvised. In a city where the follow-up question is always the same one, that distinction is the whole point.
Where the numbers stop and judgment starts
Every one of these views describes what already happened on ground you already hold. None tells you whether a term is worth pursuing, whether a page should exist, or whether the other two jurisdictions need their own documents. Those are decisions, and decisions are not in the data.
What the record settles
Questions with a countable answer and a date on it.
- Which terms entered or left the top ten
- Which pages carry the exposure
- Which rivals share your keyword space
What stays a judgment call
Questions where the honest report is a recommendation, labeled as one.
- Whether a contested term is worth the year it takes
- Whether a jurisdiction deserves its own page
- Whether an inquiry was worth having
The market-research views help at this boundary without pretending to cross it. A competitiveness score sorts the field into top-tier, mid-tier and niche rivals; the model reads a domain's positioning and flags pages worth expanding or linking internally; content gaps get named. All of it is generated, all of it prompts a decision rather than replacing one, and labeling it as such is not weakness. With this audience it is what makes the rest believable.
None of this is separable from the writing. A page that hedges honestly and cites what it claims does better here than one that asserts confidently, and that is not a moral point. It is an observation about a metropolitan area where an unusual share of the population evaluates sources professionally and stops reading the moment a claim outruns its support. The same discipline belongs on your service pages.
Questions that come up in the first review meeting
Our clicks doubled but our average position got worse. Which one do I report?
Both, in that order, with the reason. The usual cause is that you began appearing for a broad new term where you sit far down the page. It contributes many low positions and drags the mean, while the pages you held kept converting. Report the clicks as the result and the position as a mix effect, naming the queries that entered.
Why do the query totals and the page totals disagree?
They count different objects, and rare queries are withheld to protect the people who typed them. Neither table is wrong. Use the query view for vocabulary and the page view for volume, and keep the two totals off the same slide.
Can I see how Virginia performs against Maryland?
Not from the geographic breakdown, which resolves to country level here. The answer is structural: give each jurisdiction its own page and the page breakdown becomes the split you wanted. It also fixes the underlying problem, since licensing and tax treatment differ across those lines and one combined page is wrong for most of its readers.
How far back should a comparison go before it means anything?
Use the 28-day and 90-day windows and compare like against like: equal length, equal weekends, and preferably not one half sitting in a stretch when the city empties out. Below four weeks you are mostly reading noise, particularly on clicks, which are small enough that one unusual week distorts them.
A partner wants one number for the quarterly update. What do I give them?
Give them a count, not a rate: the number of terms holding a top-ten position, with the window stated. It is checkable, it does not move for reasons you cannot explain, and it survives the question about where it came from. Rates and averages shift when the mix changes, and defending them takes longer than the update itself.
Start not with a report but with one question asked of your own property: for every figure circulating inside your organization, can you name the source and the window? Connect the domain, put the two records side by side, and check your existing claims before writing anything new. Open the analytics views for your own domain, and see the blog for the longer version. The usual finding is not a ranking problem. It is three or four numbers in circulation that nobody can trace — in front of this city's readers, the expensive kind.