Skip to content
allanninal.dev/build
Collaborate

Part 4 of 7 · Search rank reporter series ~5 min read

How a ranking change is judged

There are two ways to be wrong about a search report and most reports manage both. One is treating a movement on tiny volume as meaningful. The other is treating a fall in clicks as a ranking problem when the ranking did not move at all and fewer people were searching.

Key takeaways

  • Impressions gate everything. Below a floor, nothing is reported however dramatic.
  • Compare against the same weeks last year, not just last week, wherever there is a year of data.
  • Position moving and clicks moving are different findings and are reported separately.
  • A theme can lose clicks while ranking identically. That is demand, not ranking.
  • A theme can gain impressions while losing position, which is usually a good thing.

Volume gates everything

How a change in a search theme is classifiedA vertical chain of five steps entered by a box labelled A theme, this week, against the baseline. Step one asks whether there are enough impressions against a floor you set; if not it exits to Not reported, however dramatic the change looks. Step two asks whether the weighted position moved against the last eight weeks; if so it exits to A ranking finding, naming the pages. Step three asks whether impressions moved while position stayed steady; if so it exits to A demand finding, which is not your doing. Step four asks whether only clicks moved while position and impressions stayed steady; if so it exits to A snippet finding, concerning the title or description. Step five is Nothing to report, which is most themes in most weeks. A note says these are three different findings with three different responses, and most reports conflate them.AWS ACCOUNTA theme, this weekagainst the baselineEnough impressions?a floor you setNot reportedhowever dramaticnoPosition moved?weighted, vs 8 weeksA ranking findingname the pagesyesImpressions moved?position steadyA demand findingnot your doingyesClicks moved only?position and impressions steadyA snippet findingtitle or descriptionyesNothing to reportmost themes, most weeksThree different findings with three different responses. Most reports conflate them.
Fig 1. How a theme’s movement is classified. The three branches lead to completely different actions, and a report that reduces them to one number cannot tell you which you are looking at.
  • App integration
  • Machine learning
  • Security & identity
  • Analytics

The three findings

What movedWhat it meansWhat to do
PositionYour ranking changedLook at the page and what else ranks now
Impressions, position steadyDemand changedNothing, or plan for a season
Clicks onlyYour snippet is less appealingRewrite the title or description

The third one is the most actionable and the least reported. A theme whose position and impressions are unchanged but whose clicks have fallen twenty per cent is telling you that the same people are seeing you in the same place and choosing something else. That is a title tag, and it is a twenty-minute fix with a measurable result.

Seasonality

Week-on-week comparison is fooled by seasonality with complete reliability. A plumbing firm’s emergency callout theme triples in a cold snap and halves in July, and neither has anything to do with the site.

How three different comparison windows produce three different answersA horizontal row of five boxes. Against last week: down forty per cent. Against eight weeks: down thirty-five per cent. Against the same week last year: up four per cent. Position: unchanged. Verdict: it is July. A note says only the third comparison is any use for a seasonal business and most reports omit it.THREE COMPARISONS, THREE ANSWERSvs last weekdown 40%vs 8 weeksdown 35%vs same week last yearup 4%PositionunchangedVerdictit is JulyOnly the third one is any use for a seasonal business, and most reports omit it.
Fig 2. Why a single comparison window misleads. The same data supports three different conclusions depending on what it is compared against.
  • Machine learning
  • Management
  • Analytics

So where a year of history exists, the year-on-year comparison is the headline and the week-on-week is context. Where it does not, the report says so — “no comparable period last year” — rather than presenting a week-on-week figure as though it meant the same thing.

Position falling while impressions rise

The comparison that looks alarming and usually is not. A theme that starts ranking for a much wider set of queries will show more impressions and a worse average position, because the new queries are ones you rank badly for and did not appear for at all before.

That is normally good: more impressions and more clicks with a worse average position means you are reaching more people. The report says so explicitly when it sees that pattern, because the naive reading is that something broke.

Next: what the report says.

All posts