ConsoleKit
Connect GSC
Back to blogConsoleKit / Blog / How-to

How-to

Is this drop real? How to tell noise from signal in Search Console

Clicks from Google move every day. Here's a free, spreadsheet-only way to work out how much movement is normal for your site, and when a drop deserves a closer look.

Open Search Console on any given morning and something will look different. Clicks are down 12% on yesterday. Impressions spiked on Tuesday. Position wobbled. Most of it means nothing, and some of it means a lot. The hard part is telling which is which before you've spent a day investigating a blip, or a week ignoring a real problem.

The answer isn't a rule like "only worry about drops over 20%". What counts as normal depends on your site. A site with 50 clicks a day can easily swing 40% from one Tuesday to the next. A site with 50,000 clicks a day rarely moves more than a few percent without a reason. So the useful question is: how much does my site normally move?

Why the numbers move when nothing has changed

Before measuring the noise, it helps to know where it comes from.

  • The week has a shape. Many sites get far fewer clicks at weekends, while some, like recipe and leisure sites, get more. Compare a Monday with a Saturday and you're mostly measuring the calendar.
  • Recent days aren't finished. Google describes the newest data as preliminary: it's still being collected and can change. A "drop" in the last day or two often fills in later.
  • Small numbers jump around. If a page gets 10 clicks a day, getting 6 one day is a 40% drop and completely ordinary.
  • Seasons and events. Bank holidays, school terms, paydays, weather and big news all move search demand, for you and for everyone else.
  • Google changes things. Core updates and other ranking changes are listed on the Google Search Status Dashboard. They can move rankings for weeks.

Only the last of these is likely to need action from you, and even then not always.

Step 1: compare like with like

Never compare a day with the day before. Compare it with the same weekday a week earlier, or compare whole weeks. In Search Console's Performance report, open the date filter, choose Compare, and compare the last 7 or 28 days with the period before. For seasonal businesses, compare with the same period last year instead.

Leave the most recent two or three days out of any comparison until they've settled.

Step 2: measure your normal range

This takes ten minutes in a spreadsheet and gives you a number to judge every future drop against.

  1. In Performance > Search results, set the date range to the last 6 months or more.
  2. Select Export at the top of the report. The download includes a Dates sheet with one row per day: clicks, impressions, CTR and position. Copy it into a new spreadsheet.
  3. Sort the sheet by date, oldest first, so clicks sit in column B from row 2.
  4. In C9, work out how much that day moved against the same weekday a week before:
=IF(B2=0, "", ABS(B9-B2)/B2)
  1. Fill that formula down to the last row.
  2. Find your typical and your large movements:
=MEDIAN(C9:C200)
=PERCENTILE(C9:C200, 0.9)

The median is your typical swing: on an ordinary day, clicks move about this much against the week before. The 90th percentile is your large swing: only one day in ten moves more than this.

Say your typical swing is 8% and your large swing is 22%. Then a single day that's 15% down is unremarkable. A day that's 30% down is unusual. Three or four days in a row beyond 22%, all in the same direction, is a signal worth chasing.

Prefer not to build the spreadsheet? Our free noise calculator does the same sums in your browser. Connect Search Console or paste the Dates sheet, and it shows your typical and large swings and flags any day in the last week that moved more than usual.

A couple of tips:

  • Use at least 90 days of data, so a single odd week doesn't skew the result.
  • If your site had a big change during the period, like a migration, start the range after it.
  • Run it separately for non-brand searches if you can (see how to split brand and non-brand), because brand demand has its own rhythm.

Step 3: work out what kind of drop it is

Once a drop clears your large swing, narrow it down before you start fixing things:

  • Is it the whole site or a few pages? Compare periods on the Pages tab and sort by the change in clicks. If three pages account for most of the drop, the cause is probably on those pages.
  • Is it impressions, CTR or position?
    • Fewer impressions with a steady position usually means fewer people searching. That's demand or seasonality, not you.
    • Same impressions but lower CTR means people see you and choose someone else. Check your titles, and what has appeared above you in the results.
    • A worse position means you're ranking lower. That's the one where your site, a competitor or a Google update is most likely involved.
  • Did anything change on the same dates? Releases, redirects, a CMS update, a Google update. Write these down as they happen. You won't remember them in three months.

Even when the timing lines up, a change that happened in the same week isn't proof of cause. It's the first place to look.

Where the spreadsheet runs out

The method above works. Its limits are practical:

  • It's a snapshot. Your normal range changes as the site grows, so it needs redoing every few months.
  • It's one property at a time. If you look after ten sites, that's ten exports.
  • It only covers the whole site. A section of the site can collapse while the total barely moves. To catch that, you'd repeat the exercise for every folder or topic you care about.
  • Someone has to look. The spreadsheet can't tell you that today's numbers crossed the line.

Doing it in ConsoleKit

ConsoleKit does this checking for you every day:

  • Overview opens with a one-sentence verdict on what happened over your dates, with the pages, searches and segments that moved most. Days Google hasn't finished are shaded, and if a period isn't complete you'll see No complete comparison rather than a misleading change.
  • Events writes a brief of what changed since you last looked. It only uses complete days, and when the evidence can't explain a move, it says so instead of guessing.
  • Select Show detected events under the Overview chart to see where ConsoleKit spotted something, lined up with Google updates and your own markers.

Sources

  1. Google Search Console Help: Performance report (Search results)
  2. Google: Search Status Dashboard
  3. Google Search Central blog: A deep dive into Search Console performance data filtering and limits

Get new articles by email

Research and practical guides on Search Console, about twice a month. Unsubscribe any time.

We only use your email for this newsletter. Privacy policy

Keep goingRead more articles