How to Analyse SEO Trends Correctly
This is slow work measured with fast tools. The reporting updates daily while the underlying change takes months. That mismatch is where most misjudgement in this field comes from.
The Timescale Mismatch
The work changes things over months. The reporting refreshes every day. That gap means somebody is always looking at a figure far more precise and far more recent than anything it is describing, which produces a great many confident decisions made on noise.
Why the tools are built that way. They serve everybody.
The same reporting supports advertising, where daily movement genuinely matters. Nothing about it was designed for a subject that changes across quarters.
What the mismatch produces. Reaction.
A figure dips, somebody alters a page, the figure recovers on its own and the alteration takes credit. That wrong lesson then guides the next year of work.
The second thing it produces. Abandonment.
Work stopped during a flat period that was entirely normal. Block four covers why the flat periods happen and why they are the most dangerous moment.
Who feels it most. Whoever is paying.
Monthly invoices create monthly expectations. A month is frequently too short a window to contain anything meaningful.
What this page gives you. A way to read it.
Which window to use, what genuine progress looks like from a distance and how to avoid the three commonest misreadings.
What Window To Actually Use
A fortnight shows nothing. A month shows very little. A quarter set against the same quarter a year earlier shows something real. That is a principle rather than a rule, because how much data a site produces changes how quickly a pattern becomes visible.
Why a fortnight is useless. All noise.
Ordinary variation is larger than anything the work produces in that time. Whatever you see is movement with no cause anybody could name.
Why a month is barely better. Too short to separate.
The signal and the noise are roughly the same size, so any conclusion is a coin toss dressed up as analysis.
Why a quarter works. Signal exceeds noise.
Enough time for the work to affect something and enough data for random movement to average out. That is the point where a direction becomes readable.
Why against the same quarter last year. It removes the season.
Comparing against the previous quarter compares a busy period against a quiet one. The same quarter a year back holds the seasonal position steady.
What changes the answer. Volume of data.
A site with a great deal of traffic shows patterns sooner. A small local business needs longer, because a handful of visits either way moves everything.
What this is not. A promise about results.
These are windows for reading data rather than statements about when the work starts producing. We give no timescale for that anywhere in this series.
What Genuine Growth Looks Like
Real growth in search is untidy. It is not the smooth rising line that appears in proposals. Knowing what it actually looks like prevents people from concluding that untidy means broken.
The overall shape. Upward and rough.
A general direction visible only across a long window, with substantial movement in both directions inside it. Zoom in far enough and any healthy site looks chaotic.
What it is not. A straight line.
Any chart showing smooth consistent growth has either been drawn rather than measured. Or it covers a period long enough to hide everything interesting.
The falls inside it. Expected.
A rising year contains falling months. That is not a contradiction, it is what growth looks like when it is not being smoothed for presentation.
How to see the direction. Step back.
Compare the same quarter across two or three years. Detail disappears and the direction becomes obvious, which is exactly what you wanted.
What accelerates over time. The base.
A site that has accumulated coverage tends to gain more readily than one starting out, because new pages benefit from what is already there.
The pattern inside the shape. Steps.
The roughness is not random. It has a characteristic form, which is block four and the most useful thing on this page.
The Step Pattern
Pages frequently sit unchanged for weeks and then move suddenly. That produces a staircase rather than a slope. The flat parts of the staircase are where people conclude the work has failed and stop paying for it.
What the flat part is. Not nothing.
During it, pages are being assessed, links are being found and the site is being understood. None of that is visible in a traffic figure until it resolves.
What the step is. A reassessment.
A page moves from one broad standing to another, sometimes considerably. It happens over days after weeks of apparent stillness.
Why this is so costly. The timing.
A flat period is exactly when a business asks whether to continue. Stopping then abandons the work at the point immediately before it resolves.
How to tell a flat part from a failure. Look underneath.
During a healthy flat period, the range of searches you appear for keeps widening even while visits hold steady. That widening is the work continuing.
What genuinely stalled looks like. Flat everywhere.
No new searches, no new pages appearing, nothing moving underneath. That is a different picture and it deserves the question.
What to do during a flat part. Keep going.
Publishing continues, the plan continues and nothing gets rewritten in a panic. The discipline is the same as during a drop.
What to tell a client. In advance.
Explain the staircase before it happens. A business expecting flat periods does not treat the first one as a crisis.
Direction Beats Position
Where something currently sits matters less than which way it is moving and how quickly. That is true generally and it matters most early on, when every absolute figure is small enough to be discouraging.
Why position misleads early. Everything starts low.
A new or neglected site has poor figures by definition. Reporting them as the headline tells the business what it already knew and nothing about whether anything is changing.
What direction tells you. Whether the approach works.
A modest figure moving steadily upward is better evidence than a larger figure sitting still. One of those is a trajectory and the other is a state.
How to report it. Both, with direction first.
Where you are, then where you were, then how fast the gap is closing. That order answers the question the business is actually asking.
The comparison that gives direction meaning. A starting point.
Direction can only be measured from somewhere. Our benchmarking guide covers recording one before work begins.
When position takes over. Later.
Once a site is established, absolute standing on the searches that matter becomes the more useful measure. The switch happens gradually.
The caution attached. Direction can flatter.
A large percentage rise from a very small base is arithmetic rather than achievement, which is why the underlying figures belong beside it.
Separate The Trend From The Event
One large movement inside a period distorts everything read across it. Identifying that movement and setting it aside is necessary before any trend can be read. Almost nobody does it.
What an event looks like. A spike or a hole.
One week wildly unlike its neighbours. A piece of local news, a mention somewhere, a technical fault or a day the site was unreachable.
Why it distorts so much. Averages absorb it.
A single extreme value pulls a whole period with it, which makes the surrounding weeks look worse or better than they were.
How to handle it. Read the period twice.
Once with the event and once without. If the direction is the same both ways, the event was noise. If it changes, the event was the whole story.
The event people forget to exclude. Their own launch.
A new page or a campaign produces an initial burst that settles. Reading a trend across that burst overstates everything after it.
The seasonal version. A different animal.
A recurring peak is not an event, it is part of the shape. Our seasonality guide covers telling them apart.
What to record. The events themselves.
A dated note of anything unusual. Six months later nobody remembers why one week looked strange. The note settles it instantly.
Segment Before You Conclude
A flat total frequently hides one section growing while another falls. That cancellation is the commonest way a report misses something important. It survives every other check on this page, because nothing about the headline figure looks wrong.
What cancellation looks like. Nothing.
The total holds steady month after month. Underneath, one service is gaining and another is losing at roughly the same rate. Neither is visible.
Why it matters commercially. The two are rarely equal.
If the falling half is your profitable work and the rising half is not, a flat total conceals a business getting worse.
The splits worth making. Four.
By service or section, by page, by whether the search included your name and by device. Each catches a different kind of hidden movement.
The one that reveals most. By section.
Grouping pages by what they are about maps directly onto how the business makes money, which makes any movement immediately meaningful.
When to segment. Before concluding, always.
Not only when something looks wrong. A flat total is precisely the situation where the split is most valuable and least likely to be done.
What to do with the finding. Two conversations.
A section growing and a section falling need separate discussions. Merging them into one figure guarantees neither gets addressed.
Do Not Compare To Last Month
Adjacent months are the default comparison in almost every report and they are the least informative one available. The habit comes from invoicing rather than from analysis. It introduces problems that have nothing to do with performance.
The three problems it introduces. All calendar.
Different lengths, different numbers of trading days and different positions in the seasonal cycle. Any one of those moves a figure without the business changing.
What it is genuinely good for. Spotting a break.
A sudden fault shows up immediately in a monthly comparison. That is a real use and it is a narrow one.
What it cannot do. Show a trend.
A trend needs more periods than two. Two adjacent ones are also the least comparable pair you could choose.
The better comparison. Same month, previous year.
Same length, same holidays, same seasonal position. One change removes all three calendar problems at once.
The better one still. Rolling periods.
A rolling span against the equivalent span, which removes month length entirely. Rarely done and genuinely superior.
Where the full argument sits. Its own guide.
Month on month against year on year covers which question each answers.
Where everything sits. The hub.
All nineteen guides are on the measuring performance guide.
People abandon
the work during
the flat part.
Pages sit unchanged for weeks and then move suddenly, which produces a staircase rather than a slope. During the flat part the site is still being assessed and understood. None of that shows in a traffic figure until it resolves. A flat period is also exactly when a business asks whether to continue, so stopping then abandons the work immediately before it pays.
How we read a trend:
During a healthy flat period the range of searches you appear for keeps widening while visits hold steady. That widening is the work continuing.
Every guide.
One question.
What each number actually means, how to read the data without being misled, what belongs in a report and what does not, then how to connect any of it to the enquiries the business is actually paying for.
