Setting a Baseline Before You Change Anything

Why this gets skipped

The work is exciting and the measurement is not. A redesign, a new campaign, or a batch of new pages arrives with momentum, and nobody stops to write down where things stood.

Six months later somebody asks whether it worked. The honest answer is that nobody knows, because the figures before the change are no longer at hand and the comparison period was never defined.

What usually happens instead is that everyone forms an impression, and the impression matches whatever they expected. A redesign somebody championed is remembered as a success regardless of the numbers.

What to record

Fewer things than you might think, and each one with its definition written down.

• Enquiries, by source, for the last twelve months. The number that matters.
• Sessions, total and from search, for the same period.
• Conversion rate, with the denominator stated explicitly.
• Top ten landing pages by sessions, with their individual figures.
• Search impressions and clicks for the site and for the main pages.
• Average position for the handful of terms you actually care about.
• Speed measurements for the homepage and one service page.
• Review count and rating, which change slowly and are easy to forget.

Twelve months rather than one, because seasonality will otherwise be mistaken for effect. A roofing business comparing October against July learns nothing about the redesign.

Write down the definitions too

The part that makes the comparison valid, and the part always omitted.

Which conversions are counted, whether the rate is against sessions or users, what date range was used, whether internal traffic was filtered, and which platform version produced the numbers.

Without those, the later comparison is against figures calculated differently, which produces a difference that means nothing. Platforms also change their definitions, so a note of what a metric meant at the time is genuinely useful a year on.

Capture it in a form that survives

Not a link to a dashboard, which will show different data later and may not exist at all.

Screenshots, or an exported spreadsheet, or a document with the numbers typed in. Something that is fixed and cannot be reinterpreted afterwards.

Stored somewhere findable in a year, with the date on it. A file named baseline with the month in it, alongside the project notes, is enough.

Decide what would count as success

Before the change, not after, because deciding afterwards produces a definition that fits the result.

State it specifically: enquiries up by a certain proportion within three months, or conversion rate improved from one figure to another, or speed reduced below a threshold.

A stated target is falsifiable, which is the point. It also surfaces disagreement early, since a redesign expected to double enquiries and one expected to modernise the appearance are different projects with different budgets.

The comparison window

Longer than people want it to be.

A change affecting search visibility needs three to six months before the comparison means anything. A conversion change on an existing page can be assessed sooner, in weeks rather than months, provided the traffic is sufficient.

Comparing the fortnight after a launch against the fortnight before is measuring novelty and disruption, not effect. Anything that looks dramatic in the first two weeks should be treated as noise.

What will confound it

Worth noting in advance so the eventual reading is honest.

• Seasonality, which is why the twelve-month figure matters.
• Anything else that changed at the same time, such as advertising starting or a competitor closing.
• Tracking changes, where a redesign moves or breaks the measurement itself.
• Platform changes, including how metrics are defined.
• The temporary dip that follows most significant site changes before recovery.

The third one deserves attention. A redesign that loses the conversion tracking makes enquiries appear to collapse, which is a measurement failure that looks like a business failure. Confirming tracking still works on launch day is part of the baseline discipline rather than separate from it.

The version that takes twenty minutes

If nothing else, three things.

A screenshot of twelve months of enquiries by source. A screenshot of twelve months of search performance. And a note of the date and what is about to change.

That is enough to answer the question honestly later, which is more than most businesses can do, and it costs less time than the first meeting about the project.

It also has a second use: the baseline is what tells you which pages were already working, so they can be preserved rather than replaced, which is where most redesign losses come from. That reading depends on knowing what the figures actually mean, which is covered in sessions, users, and pageviews explained.

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.