A Cold Snap in the Visits Graph

A spike nobody planned

Open the visits graph for a furnace repair company and the year has a shape. Flat through the summer, then a sudden peak on the first cold night of the autumn.

A roofer's peaks arrive with the windstorms. A plumber's come with the first hard frost, when pipes in crawlspaces freeze. A snow clearing site barely registers until it snows.

Each peak falls back within days. Read months later, without the weather in mind, the graph tells another story: the new campaign worked, or December failed.

Early January, with a full year to look back on, is when most of those readings get made.

Why the weather moves the line

People search for a trade when the problem appears. A cold house, a burst pipe or shingles on the lawn sends a whole region searching on the same morning.

Nothing about your site has changed. Your position in the results may be exactly where it was. There are simply more people searching, and some of them find you.

When the weather settles, so do the searches, and the line drops back. Every business in the trade sees the same pattern at the same time.

Readings the weather can fake

• A campaign that worked, because it started the week of the first frost.
• A redesign that worked, because it launched just before a windstorm.
• A broken site, when December was simply mild.
• A good year, compared with a warm one before it.
• A bad month, compared with a month that had a storm in it.

The first is the most expensive mistake. A campaign credited with a cold snap gets more money, and the next mild month makes it look as though it has stopped working.

Annotations, made on the day

Universal Analytics lets you pin a note to any date on the graph. Click the small tab under the over-time chart, create a new annotation, and write a line: first frost overnight, windstorm, power out across the valley.

An annotation can be private or shared with all users of the view. Choose shared, so whoever reads the graph next year sees the same notes.

Note the other things that move the line too: campaigns starting and stopping, flyers delivered, price changes, the day a new page went live.

Write the note on the day. Three months later nobody remembers which week the storm came through.

Compare like with like

A cold week compared with a mild one proves nothing. Compare it with last year's cold spell instead, even if that fell in a different week.

The date picker in Analytics will set two periods side by side. Use weeks rather than months, since a single storm can make a whole month look unusually busy.

For a season, compare the whole of it: October to March against the October to March before, with both years' weather marked.

And compare enquiries as well as visits. A cold snap brings people who need someone today, and whether they called is what matters.

Split the sources before judging

In Analytics, the channels report separates visits from paid search, unpaid search, direct visits and other sites.

A cold snap lifts unpaid search and direct visits most, as people search for the service or look up a company they already know.

A paid campaign rises in a cold snap too, because more people search and more of them see the ads. So judge it on its own clicks and enquiries over a mix of weather, not on one busy week.

A worked example

A fireplace store in a suburban row of showrooms sells gas and wood inserts that fit into existing fireplaces, and installs them with two crews.

In late October the owner started an AdWords campaign on searches for gas inserts. The first cold weekend of the season arrived four days later.

Visits climbed sharply, the showroom filled, and the campaign got the credit. Its budget was raised for November.

November turned mild and wet. Visits fell back to their summer level, and the owner began to suspect the campaign, then the website.

Splitting the channels told a plainer story. Most of the rise had come from unpaid search and direct visits, many from people looking up the store by name; paid clicks had grown only as much as searches had.

The store now marks every cold snap in Analytics, compares cold weekends with last year's, and judges the campaign on a whole month of enquiries. Its budget went back to where it started.

Filling in last year from the records

Annotations can be added for past dates as well, so last season can be marked in an afternoon.

Historical weather records for most areas are published online, day by day: temperatures, rain, snow and wind. Note the first frost, each cold spell, each storm and any heavy snowfall.

Old emails, invoices and the job calendar fill in the rest: when the campaign started, when flyers went out, which weeks the crews were run off their feet.

What the searchers were looking for

The landing pages report shows which pages people arrived on. In a cold snap it is usually one or two: the emergency page, the no-heat page, the frozen pipes page.

The Search Queries report in Webmaster Tools shows the searches behind those visits, worded the way people actually type them when the house is cold.

Those pages are worth keeping current and quick to load, because the next cold snap will send people to them again.

The counter-case

Not every business moves with the weather. An accountant's or a dentist's graph has its own seasons, and an annotation for each storm adds little.

Weather also explains less than it seems to. A site can break during a cold snap too, and a busy week can hide a form that stopped sending.

And a campaign may genuinely work. The point is to judge it fairly, not to discount every rise.

For any trade that sells into bad weather, though, note the weather and compare like with like before deciding anything.

Before judging last year

1. Open last year's graph by week, not month.
2. Mark every cold snap and storm as an annotation.
3. Add campaigns and site changes on their dates.
4. Split paid, unpaid and direct visits before judging anything.
5. Compare cold weeks with last year's cold weeks.
6. Count enquiries, not just visits.
7. Annotate on the day from now on.

Step four is the one that separates the weather's work from the campaign's. Without it, every frost looks like good advertising and every mild month like a failure.

Who keeps those notes, and who reads the graph, is part of who decides what goes on the website this year.

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.