Measuring a Site With Fifty Visitors a Day

Small numbers behave differently

Analytics interfaces are built for sites with a great deal of traffic, where a small percentage movement represents thousands of people and means something.

At fifty visits a day, the same percentage movement represents ten people, and ten people is weather.

One person sharing a link, a supplier checking something repeatedly, a slow news week, or half-term can each move a small site's figures by a quarter without anything having changed.

So the ordinary habit of looking at a weekly figure and reacting to the direction produces confident conclusions about noise.

What the numbers look like at this size

• Fifty a day is roughly 350 a week.
• A twenty percent week is seventy people.
• Which could be one shared link.
• A page with ten visits tells you almost nothing.
• A conversion rate from twelve enquiries is not a rate.

The last is the one that produces the worst decisions, since a rate calculated from a handful of events looks like a measurement and behaves like a coin toss.

Use longer periods

The single change that makes small-site analytics usable.

Look at quarters rather than weeks, and months only when comparing against the same month a year earlier.

Three months at fifty a day is about four and a half thousand visits, which is enough for a movement to mean something.

That also fits how a small business actually operates, since almost nothing you do produces a result within a week anyway.

Set your default date range accordingly, since the interface will show you seven days and invite you to react to it.

Anybody looking at a small site's figures weekly is reading noise on a schedule.

Percentages are the trap

Worth naming, because they are how small numbers are made to sound significant.

Traffic up forty percent sounds substantial and may be twenty additional people.

Conversion rate down by half may be three enquiries instead of six, which is entirely ordinary month-to-month variation.

The habit worth adopting is reading the underlying count first and the percentage second, and refusing to quote a percentage without the count beside it.

Where somebody presents a report full of percentages with no counts, ask for the counts, since that presentation is either careless or is making small numbers look like results.

Count enquiries instead

The measurement that works at any size, which visits do not.

Twelve enquiries in a month is a real number about a real thing, and it can be compared against the eight last month meaningfully because each one is a person who contacted you.

Visits are a proxy for interest that becomes reliable only in volume. Enquiries are the thing itself.

Record them yourself, in a spreadsheet, with the date and where they said they came from, which is a minute a week.

A year of that is worth more to a small business than any analytics configuration, and it survives every migration and every product change.

It also answers the question the analytics cannot, which is how many became customers.

A worked example

A business with around sixty visits a day published a blog post and saw traffic rise by roughly a third the following week.

They concluded the post had worked and planned more like it.

The rise was about a hundred and thirty extra visits over the week, of which the new post accounted for eleven.

Most of the remainder came from a local group sharing their homepage after a community event, which had nothing to do with the post.

Enquiries that month were four, the same as the month before and the month after.

Reading the quarter rather than the week would have shown a flat line with a bump in it, and the post would have been judged on its eleven visits, which is the honest figure.

What is still worth reading

Small volume does not make analytics useless, and it does change which parts are informative.

The proportional split of traffic sources is stable enough to be useful over a quarter, and it answers a real question about where people come from.

Which pages people land on is meaningful cumulatively, even if any individual week is thin.

Device split is reliable quickly, because it does not depend on volume in the same way.

And any page with zero visits over a quarter is telling you something definite rather than something noisy.

Those four are readable at fifty a day. Rates, week-on-week movements, and per-page conversion figures are not.

Wait longer before concluding

The discipline that follows from all of the above.

After making a change, resist judging it for a full quarter, and preferably against the same quarter last year.

That feels impossibly slow and it is the only way to distinguish an effect from variation at this size.

Where you cannot wait, accept that you are making a judgement rather than a measurement, and say so.

The alternative is the common pattern of changing something, seeing a movement, changing something else, and never learning anything from a year of activity.

The counter-case

Waiting a quarter is not always right.

Anything spending money needs watching more often, since a campaign performing badly for three months is money gone rather than a data point, and cost per enquiry is worth checking weekly even on small numbers.

Obvious breakages also announce themselves immediately: traffic going to near zero is not noise, and a form that stops working shows up as an absence of enquiries within days.

And a site growing quickly has genuinely different numbers each month, where a quarterly view would hide a real trend.

Watch spending and breakages closely. Read everything else in quarters.

What to do

1. Set your default range to a quarter.
2. Read counts before percentages.
3. Refuse to quote a rate from a handful of events.
4. Compare against the same period last year.
5. Record enquiries yourself, weekly.
6. Judge changes after a quarter, not a week.
7. Watch spending and breakages more often.

Step five is the one that makes measurement possible at this size, since enquiries are countable and visits are not, at fifty a day.

Recording where you started is covered in setting a baseline you can compare to.

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.