Setting a Baseline You Can Compare To

The argument that happens in month six

Somebody says the new site is doing better. Somebody else says enquiries are down.

Both are looking at real data and reaching opposite conclusions, because nobody wrote down what the position was beforehand.

What follows is a discussion about what the numbers used to be, conducted from memory, in which the person with the strongest interest in the answer usually has the clearest recollection.

That conversation is entirely preventable by half an hour of work before anything changes, and it is the single most skipped step in website projects.

Why nobody does it

Because it produces nothing visible.

Everything else at the start of a project makes something: a design, a plan, a page. A baseline produces a document nobody will look at for six months.

It also happens at the point of maximum enthusiasm, when the interesting part is about to begin and recording the current state feels like administration.

And there is a quieter reason: a baseline makes failure legible. A project with no recorded starting point can always be described as an improvement.

What to record

• Twelve months of traffic, monthly totals.
• Top twenty pages by visits, with their numbers.
• Enquiries by month, counted from your own records.
• Positions for ten search terms you care about.
• Search clicks and impressions, monthly.
• Speed figures for the homepage and one service page.
• Screenshots of the main pages as they are now.

The last is worth more than it sounds. In a year nobody will remember what the old site looked like, and the screenshots settle a surprising number of later questions.

Keep it outside the analytics account

The detail that makes the difference between a baseline and a notion.

If the record lives inside the tool, then anything that changes the tool changes the record: reconfiguration, a new property, a filter added, a switch to a different product.

Comparing against a system whose settings have moved is exactly the problem you were trying to avoid.

Export it. A spreadsheet, a document, a folder of screenshots, dated, stored where the business keeps things rather than where the agency does.

The point is that it is a fixed record, unaffected by anything done afterwards.

Enquiries are the line that matters

If only one thing gets recorded, record this.

Count them by hand from your own inbox, call log and job records rather than taking a conversion figure from analytics, because analytics conversions are frequently reconfigured during a project and become incomparable.

A hand count from source records survives any technical change, which is exactly the property a baseline needs.

Twelve months of monthly enquiry counts, on a single line, is the most useful baseline a small business can hold.

A worked example

A firm rebuilt their site and, six months later, were in dispute with the agency about whether it had worked.

Traffic was down, which the agency attributed to more accurate bot filtering on the new setup.

Nobody could establish whether that was true, because the old configuration no longer existed to check.

The only figure that resolved anything was the enquiry count, which the office manager had kept in a notebook throughout for her own reasons, unrelated to the project.

Enquiries were up by about a fifth.

The notebook settled a dispute that fourteen months of analytics could not, because it was the only record that had not changed underneath everybody.

Note the conditions too

A baseline of figures alone loses the context that explains them.

Write down what was running at the time: advertising spend, any campaign, a trade show, a seasonal peak, staff changes.

Six months later these are forgotten completely, and a change in results gets attributed to the website when it was actually a campaign that ended.

Three or four lines of narrative alongside the numbers prevents a substantial category of wrong conclusion, and costs almost nothing to write while it is still obvious.

Set the review date now

The other half of the baseline, and it belongs in the same document.

Decide when the comparison will happen, and what will be compared, before you have any results.

Three months for anything operational, six months for search, and a year for anything structural is a reasonable default.

Writing the date down prevents the two failure modes: judging at four weeks, when nothing is legible yet, and never judging at all, which is more common.

Take it again at the end

The step that turns a one-off into something cumulative, and almost nobody does it.

When the project finishes, record the same seven items again, dated, in the same document.

That closing record is the opening baseline for whatever happens next, and there is always a next thing: a campaign, a new section, a change of supplier.

It also captures the position while the project is fresh, rather than leaving it to be reconstructed a year later when somebody asks what the rebuild actually achieved.

A business that does this consistently ends up with a dated record of its own trajectory, taken at consistent points, which is worth considerably more than any individual comparison.

Keep them all in one file rather than one per project, so the sequence is visible in a single view. Three or four such entries is enough to see a direction that no single before-and-after ever shows.

The counter-case

A baseline can be taken too far and become its own project.

Recording forty metrics because they are available produces a document nobody will read at review time, and the comparison then gets made on whichever two lines somebody happens to look at.

Seven items, as above, is enough for almost any small business.

There is also a case where the baseline barely matters: if the current site is plainly broken, or there is no site at all, you are not evaluating a marginal improvement and the comparison is not in doubt.

Even then, record the enquiry count. It costs two minutes and it is the number somebody will ask about later.

The half hour

1. Export twelve months of traffic and top pages.
2. Count enquiries by hand from your own records.
3. Note ten search positions and current search clicks.
4. Run a speed test and save the result.
5. Screenshot the main pages.
6. Write three lines on what else was running.
7. Store it outside the tools, dated, with a review date.

Nothing on that list is difficult and all of it is worthless if done afterwards.

Reading the numbers after a change is covered in two weeks after a rebuild.

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.