What should we measure before the redesign goes live?

The usual pattern we see is not that teams forget to measure. It is that the measurement changes at launch, so the before and after are not the same thing.

Tracking gets rebuilt with the new site. New event names, new tag setup, sometimes a new analytics property. The old conversion event fired on a thank-you page URL that no longer exists. The new one fires on a form submit. Nobody wrote down which was which. Six months later somebody puts the two numbers side by side in a board deck and the comparison is meaningless, but it still gets presented.

Two other things wreck baselines. Seasonality and ad spend move traffic far more than a design does, so a launch in a strong quarter looks like a triumph and the same site launched in a dead month looks like a failure. And when nobody agreed in advance what success meant, success gets decided afterwards by whoever is holding the dashboard.

Total sessions is almost always the one number that did get captured. It is also the one that proves the least.

Do not build a dashboard. Write down what you are claiming the redesign will do, in plain words, then attach one number to each claim. If you cannot name the claim, you do not need the number.

If the claim is “more people will request a demo”, the number is the conversion rate on that single action. If the claim is “the pricing page will stop confusing people”, the number is form starts versus form completions on the pricing enquiry form, tracked separately, because those two moving in opposite directions tells you something a single completion count cannot.

Three or four numbers is the working range. Under three and you have no way to tell a real change from noise in one metric. Over five and nobody maintains the definitions, which is how you end up back at total sessions.

These are the ones that survive a rebuild and still mean something afterwards.

• Conversion rate on your single primary action, defined as an event you can still fire on the new site. Check that before launch, not after.
• Form starts and form completions, recorded as two separate numbers. The gap between them is the thing a redesign actually moves.
• Organic entrances to the pages you are keeping. Not the whole site. If you are deleting 200 pages, whole-site organic traffic will fall and tell you nothing.
• Enquiry quality, judged by whoever handles enquiries. Ask sales to rate last quarter's enquiries as qualified or not, and write down the count. It is subjective and it is still evidence.

Bounce rate, if the tracking setup changes at launch. A different event configuration can halve or double it without a single visitor behaving differently.

Time on page, when the new navigation is faster and clearer. People finding what they need in less time reads as a decline. You will be defending a number that went the right way.

Total sessions, when marketing spend varies month to month. Unless spend is flat across both windows, this measures the budget, not the site.

Heatmap screenshots. They are useful for forming a hypothesis before design starts. They are not proof of anything afterwards, because you have no way to say whether the difference is real or the difference between two months of traffic.

Take a clean window, ideally two: the same season last year, and the most recent full quarter. Two windows let you separate a seasonal swing from a real one.

Export the raw numbers to a spreadsheet you own. Do not rely on a live dashboard. Dashboards get reconfigured, properties get replaced, and the historical view you were counting on quietly stops matching.

Then freeze the definitions in writing, in the same file. What counts as a conversion. Which pages are in scope. Which date range each number covers. One paragraph is enough, and it is the part that makes the comparison defensible when someone challenges it.

Record page speed and Core Web Vitals for your top ten pages before launch. Performance regresses quietly. A new site with heavier images and more scripts can be slower than the old one, and without a before number you cannot tell whether the site got worse or was always like that.

Screenshot the old site and save the full list of old URLs. Not for measurement, for the argument. When someone claims a section was removed or a page used to rank, the screenshot and the URL list end the discussion in a minute.

Decide now how long you will wait before judging anything, and say the number out loud to whoever will ask. Four to six weeks is a reasonable floor for most marketing sites, longer if your sales cycle is long. A dip in the first two weeks is normal and says nothing.

If a number drops sharply, check the tracking before you check the design. Fire the conversion event yourself and confirm it lands in analytics. In our experience most alarming post-launch drops in the first fortnight are an event that stopped firing, not visitors behaving differently.

Compare like pages against like pages. If the redesign merged twelve service pages into four, the site-wide average has changed shape and comparing it to the old average measures the merge, not the design.

If your site gets a few thousand visits a month and a handful of enquiries, a conversion rate will never move far enough to prove anything. Ten enquiries becoming thirteen is not a result. It is a normal month.

Say this before launch, not after. Then measure what you can actually read at your volume: the count of qualified enquiries, what sales says about their quality, and whether the questions people ask on calls have changed. “We stopped getting asked what we do” is weak as a statistic and strong as evidence.

You do not need a new analytics stack or a conversion optimisation retainer to run any of this. A spreadsheet, your existing analytics, and thirty minutes with whoever handles enquiries covers it.

An honest baseline makes it possible to show that the redesign did not work. That is the whole point, and it is also why the step quietly disappears from a lot of projects. Nobody refuses to measure. The window just gets missed, and by launch there is nothing to compare against, which is more comfortable for everyone being paid.

We would rather you had the numbers. A studio that only looks good when nobody is counting is not worth hiring.

This takes an afternoon and has to happen before design work starts, because once the new site is being built the clean pre-launch window is already shrinking.

1. Write down the two or three claims the redesign is making, in plain sentences.
2. Attach one number to each claim. Stop at four.
3. Export last quarter and the same quarter last year to a spreadsheet you own.
4. Write the definitions underneath: what counts as a conversion, which pages are in scope, which dates.
5. Record Core Web Vitals for your top ten pages.
6. Save a full list of current URLs and screenshots of the main pages.
7. Name one person who owns these numbers after launch and reports them at the agreed date.

One named person, not a team and not the studio. The person who will be asked in six months whether the redesign paid off should be the person who wrote the baseline down, because they are the only one who knows what the numbers meant when they were recorded.