A Year of Analytics, Read Once

Why once a year is the right frequency

For most small businesses, monthly analytics produce more anxiety than insight.

The numbers are small enough to be noisy, so a quiet month looks like a problem and a busy one looks like success, and neither is usually either.

Twelve months at once smooths that out and answers questions the monthly view genuinely cannot: whether the direction is up, which pages consistently do the work, and where your visitors actually come from.

An hour in December, done properly, is worth more than twelve fifteen-minute glances that produced nothing you acted on.

The five questions

• Where did people come from, by channel.
• Which pages received the traffic.
• Which pages produced enquiries.
• What is the trend against last year.
• What changed, and when.

Those five cover almost everything a small business can act on. Anything beyond them is usually interesting rather than useful.

Traffic and enquiries are different questions

The distinction that makes this worth doing at all.

The pages receiving the most traffic and the pages producing the most enquiries are frequently different pages, and the gap between the two lists is the finding.

A page with high traffic and no enquiries is either attracting the wrong people or failing to ask.

A page with few visitors and a high conversion rate is worth promoting, because you have something that works and nobody is seeing it.

Most businesses have both, and neither is visible from a traffic report alone, which is why the enquiry side has to be counted on your own system.

Compare against last year, not last month

The comparison that removes seasonality.

December against November tells you about Christmas. December against last December tells you about your business.

Look at the full twelve months against the previous twelve as a single figure, then at each quarter against the same quarter.

That produces a trend rather than a series of movements, and it is the only comparison that survives a seasonal business.

Where you do not have last year's data, this is the year to start, and next December will be considerably more useful than this one.

A worked example

A firm looked at a full year for the first time, having previously glanced at monthly figures.

Total traffic was up about a fifth, which they had assumed from the monthly view.

The channel breakdown was the surprise: almost all the growth was from one article about a technical question, which brought a large number of visitors and had produced no enquiries in twelve months.

Meanwhile their three service pages had flat traffic and had produced almost every enquiry they received.

So the growth was real and commercially irrelevant, and the pages that mattered had received no attention all year.

The following year they wrote for the service pages rather than for the traffic, and their traffic grew more slowly while enquiries rose.

Note what changed and when

The part that turns data into learning.

Mark on the year what you did: a new page, a rebuild, a campaign starting or stopping, a price change, a season.

Then look at whether anything in the numbers responded.

Most of the time nothing did, which is itself useful and stops the same effort being repeated.

Occasionally something did, and knowing which change produced it is worth more than the change itself.

Businesses that never record what they did cannot learn from a year of data, because the data has nothing to be compared against.

Write down four numbers

The step that makes next year's review possible.

Total organic visitors, total enquiries, enquiries from the top three sources, and your best guess at the cost per enquiry from anything paid.

Four numbers, a date, and three sentences saying what you think happened.

Keep it in the same place every year, so that in three years you have a record that no analytics tool will give you, because tools change and data gets lost when accounts move.

That file is worth more than the reports it summarises, and it takes ten minutes to produce.

What a year cannot tell you

Worth stating so the review is not over-read.

It cannot tell you why anything happened, only that it did, and the temptation to attribute causes to changes you made is strong and frequently wrong.

It cannot tell you about the people who never arrived.

And it cannot connect visits to revenue unless you have been recording enquiries and their outcomes, which most businesses have not.

Pair the analytics with your own record of enquiries and where they came from, which is the part that makes the whole exercise commercial rather than descriptive.

Check what is not being counted

An hour spent on the year is wasted if the measurement itself has been broken for months, which happens more than people expect.

Confirm the tracking is present on every page, including any built later in the year, since a page added in June with no tracking is invisible in the review.

Check that your own visits are excluded, because a small site can have a meaningful share of its traffic coming from the people who run it.

Look for obvious bot or referral noise inflating a channel, which is common on smaller sites and distorts exactly the breakdown you are about to read.

Five minutes on that before the review is the difference between reading your business and reading an artefact.

The counter-case

Once a year is too infrequent for some businesses.

Anybody running paid advertising needs to look far more often, since a campaign performing badly for three months is money gone rather than a data point.

A site that is actively being changed also needs checking after each change, because the point of the change was to move something.

And a business where the site is the shop, taking orders daily, has enough volume for weekly figures to mean something.

The annual review suits a business whose site produces a modest steady flow of enquiries, which is most small businesses, and it should sit alongside rather than replace whatever the advertising requires.

The hour

1. Set the range to twelve months, and the year before.
2. Look at channels first.
3. List the top ten pages by traffic.
4. List the pages that produced enquiries.
5. Compare the two lists.
6. Mark what you did and when.
7. Write four numbers and three sentences.

Step five is where the useful finding almost always is, and it takes two minutes once the lists exist.

Working out what the site is worth is covered in the return on a website, calculated honestly.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.