Most guides to search reporting assume your rivals are asleep. In Austin they are not, and that one fact changes which numbers deserve attention and which are only there to make you feel busy.
The reporting is not hard to reach. A verified property hands over more measurement than any small company can consume. The difficulty is deciding which figures still carry information once every serious competitor in your category has looked at the same screen.
Your competitors have read the same manuals you have
Between the semiconductor plants north of the river and the software floors downtown, a very large share of this city's business owners spent part of a career in marketing. The founder of a Cedar Park mechanical contractor did eight years in demand generation before buying the company. The Round Rock logistics firm has a marketing manager who ran paid search for a venture-backed product. They know what a keyword tool is, and they have already claimed the terms it recommends.
That inverts the standard advice. Elsewhere the counsel is to do the obvious thing first, because nobody has. Here the obvious thing is finished. Category head terms are held by firms with real budgets and real content teams, and a report showing you eleventh on one of them is not evidence of neglect. It is evidence that eleven capable organizations wanted the same string.
- The introductory wins are gone. Anything a free tool suggests on its first screen was suggested to your competitors too, years ago.
- Head terms cost more than they return. Contesting a two-word category phrase against a content budget is a project, not a task, and the traffic is the least qualified on the page.
- Specificity is the remaining advantage. The long, awkward query nobody optimized for is where a small firm still converts, and no tool suggests it.
- Half your market did not exist recently. Demand comes largely from people who moved here in the last five years and companies founded inside that window, so the historical record is thin.
The two records were built to answer different questions
Two families of screens sit beside each other and appear to quote the same quantity. They do not, and the reflex to decide one is broken when they disagree wastes an afternoon every time.
The Search Console side reports what Google's serving machinery recorded for a property you proved you control: the occasions your result was placed in front of somebody, and whether that person picked it. Within its walls it is definitive. Outside them it has nothing to say. When a firm in Georgetown displaces you, no indicator lights up; your curve bends and you invent an explanation.
Rank tracking works the other way around. It reads the results page from outside under fixed conditions, which is why it can name the domains occupying the slots, report Domain Authority, and count the terms you and a rival hold in common. It sees nothing past the click. Eight views cover the first side of that boundary and six the second, and the argument for holding both records in one place is that a question answered by one is unanswerable by the other.
| What you want to know | Which side can tell you | Why the other side cannot |
|---|---|---|
| Did anyone actually pick us? | The click column | Placement is not a decision |
| Which firms share our terms? | Competitors view, with authority scores | Rivals leave no trace in your own log |
| What do buyers really type? | The query breakdown | Volume estimates are modeled, not observed |
| Is this term worth the fight? | Keywords ordered by search volume | Impressions count exposure, not appetite |
| Did we slip, or did the page? | Both together, never one alone | Half a cause and a confident story |
One further distinction matters in a sophisticated market. Your own log describes ground you already occupy; the external view describes ground held by somebody else. Against able competitors, the second is where the strategy lives.
Every one of the four top-line numbers lies on its own
Four figures sit across the top of nearly every report: impressions, clicks, click-through rate, average position. Each is defensible, and each read alone points somewhere unhelpful. The failure modes differ enough to need naming individually.
Impressions
Counts the times a result of yours was rendered on a page somebody saw. A measure of exposure, and exposure is easy to manufacture badly.
- Rises when you appear for irrelevant strings
- Rises when a page slides into a wider result set
- Falling impressions with steady clicks is good news
Clicks
The only number matching a human choosing you over the alternatives on screen, and the one most sensitive to how the result reads.
- Says nothing about what happened afterward
- Concentrated in a handful of rows, always
- Small absolute counts swing wildly week to week
Click-through rate
A ratio, so it moves when either half moves. A rising rate can mean a better title, or that you quietly stopped appearing for a broad term.
- Meaningless above the query level
- Rate varies enormously by position
- Compare rows to their own history, not to a benchmark
Average position
The average of the best placement recorded on each occasion, across every query and location. An average of averages.
- One rare term at rank 60 drags the whole figure
- Improves when weak terms simply stop appearing
- Never actionable without a query filter applied
Read the four as a set. Impressions up and clicks flat means exposure on terms that do not want you. Clicks up and impressions flat means the result reads better. Position steady with click-through rate falling usually means a competitor rewrote a title above you — which happens here constantly, because someone over there is watching their own report just as carefully.
Queries describe the market; pages describe what you built
The same data is offered two ways, and the choice is not cosmetic. Grouped by query it records demand: the actual strings people typed. Grouped by page it records supply: the documents you published and how each performs.
Owners open the page view first, because it maps onto work they remember doing. That is why it is the weaker of the two for finding opportunity. It can only show pages that exist. Demand you never addressed leaves no row at all, and where demand is drifting, the missing rows are the point.
- Query view, impressions high and clicks near zero. Terms where you appear and get passed over — sometimes a titling problem, more often a mismatch with what the page offers.
- Query view, clicks on long strings. The specific, awkward phrasings that convert. Worth building a page around, and no keyword tool will hand them to you.
- Page view, filtered to one URL. Every string that page collects. One page pulling twelve unrelated phrasings is usually two pages not yet separated.
- Query rows with no matching page. The most valuable hour you will spend here. Demand you serve accidentally rather than deliberately.
Cross the two and it sharpens further. Take one high-value query and see which page Google chose to serve for it; often it picked one you would not have. That disagreement is a content decision waiting to be made, invisible from either view alone.
What a service was called in 2020 is not what people type now
Here is the second Austin problem, less obvious than the first. A market that grows by arrival gains customers who learned the category vocabulary somewhere else, from companies that describe the work in their own words, and who bring that phrasing with them. Meanwhile the categories keep splitting: a service that was one line item four years ago is now three products with three names.
So a keyword list assembled with real care eighteen months ago is now wrong in a way nobody notices, because nothing about it looks wrong. Every term still returns results. Volume estimates still populate. It simply stops describing what your buyers type, one phrase at a time, and the only record that catches the drift is your own query breakdown — because that record is observed rather than modeled.
| Signal in the query view | What it usually means | The move |
|---|---|---|
| New phrasings with tiny counts | A term entering use in your category | Watch a cycle; if it grows, write for it before it is contested |
| An old core term losing impressions | Vocabulary moved, not demand | Find the replacement phrasing before rewriting |
| Long strings you never targeted, with clicks | Demand served by accident | Give it a page and stop relying on luck |
| Company names you do not recognize | New rivals, or buyers comparing you | Check the competitors view before assuming |
| Questions phrased as full sentences | Buyers new to the category | Answer in the same register; these convert late |
This is the maintenance job that decays fastest when done by hand. A pool assembling itself from three inputs, then asking a human to rule on each candidate, survives vocabulary drift better than a spreadsheet.
AutoSEO — the keyword pool that refreshes itself
For a category whose vocabulary moves faster than anyone has time to track by hand.
- Three sources feed one pool. Search Console history, live results-page readings and your own seed terms, combined rather than kept in separate files.
- Every candidate is decided individually. Approved, rejected or deferred, so a phrasing that looks odd today can be parked and revisited when it grows.
- Discovery runs without prompting. Backlink work, on-site suggestions and the full reporting sit in the same account.
Where the category is contested and wording matters term by term, the manual tier exists for the same reason a senior person still reviews a contract a template produced.
FullSEO — selection stays in human hands
For categories where two near-identical phrasings separate a buyer from a browser.
- Manual keyword choice, automatic fallback. You pick the terms; the pool continues underneath so nothing stalls while a decision is pending.
- Placement with an authority target. Links placed manually against a stated Domain Authority threshold, from a network of over 230,000 sites.
- On-site changes pass a human review. Edits wait for approval rather than landing on the page, with specialists, developers and writers attached.
Where the person was, and what they were holding
Two breakdowns get skipped by nearly everyone, and both change conclusions. Austin's economy makes them load-bearing: many businesses here sell to people who have never been to Texas, and many others sell to people standing in a driveway.
Who is even eligible to buy
A software firm with international impressions has genuine reach. A local trade with the same profile has noise dragging every blended figure it owns.
- Filter to your served market before judging a rate
- Foreign impressions inflate exposure, depress click-through
- Country heatmaps show the pattern at once
What the moment looks like
A phone search for a residential trade is a person deciding now. A desktop search for a technical product is somebody building a shortlist for next week.
- Positions often differ by several places between devices
- Mobile rates fall on slow pages, not on weak rankings
- Split the two before rewriting anything
The device split has a second use. When a page holds a strong desktop position and a weak one on phones, the problem is almost never the wording. It is the page, and no amount of keyword work will move it.
Comparing two periods without deceiving yourself
Every report offers a date range and a comparison against the preceding one. Twenty-eight and ninety days are the sensible defaults, and both account for the two-day settling delay — the most common source of a Monday morning false alarm.
Comparison turns a static figure into a claim about change, and it is where the largest errors are made, because a difference between two periods has many possible parents and the report names none. A window straddling a major festival is not comparable to the weeks before it. A legislative session, a semester boundary, a competitor's launch and a change you shipped all produce the same shape of line.
- Compare like against like. Twenty-eight days against the preceding twenty-eight, or the same weeks a year earlier. Never short against long.
- Write the change down first. A dated note of what you altered turns a coincidence into something testable.
- Change one thing per window. Two edits in one period give one result and no attribution — how confident but wrong habits form.
- Beware the young comparison. If your company or category is younger than the window behind it, that earlier period may hold no usable signal.
That last point is the local one. Where a large share of firms did not exist five years ago, year-over-year comparison often measures a real business against a period when it barely had a website. Do not avoid the comparison; just say out loud what the earlier period actually was before drawing a conclusion. Our notes on measurement return to this more than once.
Questions that come up in the first month
Our average position improved but traffic fell. What happened?
Almost always the same thing: you stopped appearing for broad, low-value terms sitting far down the page. They dragged the average down while contributing almost no clicks. Losing them improves the number and cuts exposure at once. Check impressions on the queries that disappeared before calling it a win or a loss.
The rank tracker says position four and Search Console says nine. Which is right?
Both. The tracker takes one standardized reading under fixed conditions; your log averages every real occasion the result was served, across locations, devices and personalized variations. A consistent gap usually describes geography: strong where the reading is taken, weaker across the rest of the area you serve.
How long before a change shows up in the data?
The last two days of any window are still settling, so nothing recent is final. Beyond that, meaningful movement typically appears in four to eight weeks. A change judged after ten days is judged on noise — and here that noise includes your competitors' edits as well as yours.
Should we chase the terms our biggest competitor ranks for?
Look at them — the competitors view exists for that, with authority scores and shared-term counts. But a term held by a firm with a content team and years of links is a poor first target. Read the shared-term list backwards instead: terms they hold and you do not, filtered to the ones where their page is generic and yours could be specific.
How often should the keyword list be revisited?
More often than feels necessary. Where vocabulary is stable, once or twice a year is fine. Where products get renamed and buyers arrive carrying other words, a quarterly pass over the query breakdown catches drift while it is cheap. The tell is a small-count phrasing that keeps returning.
The point at which the numbers hand the problem back to you
All of this is descriptive. The reporting names which strings brought people, which pages received them, how the picture differs by device and country, and how this month compares to the last. It cannot say whether a customer from a given query was worth having, whether building a page for it fits the week you have, or whether the competitor above you is defending that position or merely occupying it.
Those are judgments, and where the competition is good they are the only durable advantage left. The data narrows the field of plausible decisions, sometimes dramatically; it does not make one. The generative research views — the competitiveness score with its market circle, intent classification, content-gap analysis — narrow it further, then stop at the same wall.
A first pass takes an afternoon. Open the query breakdown for ninety days, sort by impressions with clicks near zero, and read rows rather than totals. Then sort by clicks and look at the longest strings there. The first list shows where you are passed over; the second shows what you are already good at without meaning to be. Between them sits most of the work worth doing, and the market research and analytics screens keep that pass from taking a week.
If the site is one of several, or your category vocabulary has moved twice since the last review, it is worth seeing the data assembled rather than collected — open the panel and connect a verified property to find the queries nobody has claimed. What that surfaces here, more than in most cities, is not a ranking problem. It is a list of specific things people ask for that nobody in your category has answered plainly. Which of them you take on is a question about your own capacity, and that part is not in the report. If you would rather have the pass run for you, our engagement outlines cover it, and the export and report builder makes the result something you can hand on.
