From a number to a decision

The gap between measuring and improving is where most analytics effort disappears. The numbers get collected, discussed, and nothing is changed, usually because no single number ever makes the case on its own.

Go from number to cause to action

A figure never tells you what to do. It tells you where to look, and the looking is a separate step that people skip in favour of a general improvement.

1. 01Notice the movement, against the same period last year rather than last month.
2. 02Narrow it. Which page, which source, which device, which country. Almost every total that moved is one segment moving, and the total view hides which.
3. 03Form one explanation you could be wrong about. Not traffic is down. Rather: search traffic to one page fell after we changed its title.
4. 04Check it against something other than the same tool. Search Console, your own records, the page itself.
5. 05Make one change and write down what you expect, then look again in a fixed period.

The fourth step is the one that separates this from guessing, and the fifth is what turns a change into something you can learn from.

The traps on the way

• Acting on a week. Most weekly movements are noise and acting on them produces a site that is constantly changing and never improving.
• Changing several things at once, which guarantees you cannot attribute whatever happens next.
• Explaining a movement with whatever you did recently, which is the most available explanation rather than the most likely.
• Ignoring the outside world. A competitor launched, a platform changed, a season turned. Not everything is about you.
• Treating a tool's number as the thing itself rather than as a measurement of it.

What usually turns out to be worth doing

After the narrowing, the useful actions on a small site are a short and repetitive list, which is reassuring rather than disappointing.

• Rewrite a title and description for a page with impressions and no clicks.
• Add the missing section to a page that nearly ranks for a query it only mentions.
• Remove a step or a field from the path to the outcome.
• Say the price, the delivery cost or the timescale earlier.
• Stop spending on a channel whose cost per customer exceeds what a customer is worth.
• Update the three pages that bring the most traffic, which is the cheapest work available on an established site.

None of it is dramatic. The businesses that improve steadily are doing this short list repeatedly rather than finding insights, and the main thing analytics provides is knowing which item to do next.

When to do nothing

Underrated and genuinely difficult, because a report that prompts no action feels like a wasted report. Four situations where the correct response is to leave it alone.

• The movement is within the normal range. If the last eight weeks have swung by a fifth either way, this week swinging by a fifth is not information.
• You changed something recently and are still waiting. Changing again now means you learn nothing from either change.
• The cause is outside the business. A season, a holiday, a competitor's campaign, a platform update. Reacting to these produces a business that chases other people's events.
• The number is bad but the cause is known and already being worked on. Adding a second response does not make the first one faster.

The discipline is to write down which of these applies, rather than simply not acting. A recorded decision to wait is a decision; an unrecorded one is indistinguishable from not having looked.

This matters more than it sounds for a one-person business, where the real constraint is attention. Every change made in response to noise is a week not spent on the one item from the list above that would have mattered.