We relaunched our own website nine days ago. Here is what the data actually says

Our new site went live on 3 October. In the first full week, search impressions went from 13,829 to 48,274 and clicks went from 30 to 63. Both roughly what you would hope for.

In the same week our average position went from 14.7 to 20.8, click-through rate halved, and our own analytics stopped being able to answer the only question that matters.

All four of those things are the same event. This is what a relaunch actually looks like in week one, and the reason to publish it is that nine days is far too early for anyone to be claiming a result, including us.

Search Console, the first full week after launch against the week before it.

If I wanted a LinkedIn post out of this I would use the first two rows and stop. Impressions up 249%, clicks up 110%, nine days after launch. Both figures are true.

The third and fourth rows are why I am not doing that.

The old site had around 127 blog posts and a modest set of service pages. The new one puts roughly 600 URLs in the sitemap, because the services are now broken out properly: technical SEO separate from schema markup separate from JavaScript SEO, and so on down.

So Google suddenly has several hundred more pages of ours to consider, and it is showing them. That is what an impressions jump after a relaunch mostly is. It is a measure of how many doors now exist, not of how many people are walking through them.

You can watch it settle in the daily figures. Impressions peaked at 9,449 on 7 October, then 7,364, then 5,518, then 3,935 by the 10th. Google tries the new pages, most of them do not earn their place yet, and the number comes back towards something real.

Average position moved the same way and for the same reason. 14.7 to 20.8 is not a decline on the pages that were already working. It is an average across a far larger set of queries, most of them new and most of them ranking badly, which is exactly what you would expect from pages that are nine days old.

The honest headline is not "visibility up 249%". It is "we published a lot of pages and Google has started looking at them".

Underneath the average, the pages that matter are fine.

"seo agency bristol" sits at position 4.9. "seo agency in bristol" at 6.2. "seo company in bristol" at 6.9. "seo bristol" at 7.7. Those are the four terms this business is actually bought from, and they held through a full platform migration, which is the thing that usually goes wrong. Our migration guide exists because we have watched other people lose exactly this.

The homepage took 25 of the 63 clicks at an average position of 5.5.

And the glossary pages have started earning in week one. The page explaining header tags took 457 impressions at position 9.3. Ad rank, sessions versus users, all picking up impressions from a standing start. That matches what HubSpot's AI Search Lab found when they tested glossary pages, which we wrote up in what HubSpot tested on AI search, and it is the cheapest thing on any content plan.

Click-through rate halved, from 0.2% to 0.1%.

Even the higher number is terrible, and I want to be straight about that rather than let it sit in a table. A site taking 48,274 impressions and 63 clicks is being shown constantly and visited almost never.

Some of that is the long tail of new pages appearing at position 30 where nobody clicks anything. But we have written about this pattern before using one of our own glossary pages: it ranks 3.3 for its main term, and across all its queries it took 2,695 impressions last month and no clicks at all. That piece is here.

Impressions holding while clicks do not follow is the single most common shape in search right now, and a relaunch makes it look dramatic rather than causing it.

Here is the part that is embarrassing and worth publishing anyway.

Last week I wrote a piece arguing that if you cannot measure what a website brings in, you cannot defend what it cost. The DeskLodge case note sits under it, where tracking every booking step is the reason that project can show £16,000 of bookings rather than a feeling.

Our own new site has recorded zero conversions since it launched.

Two things happened at once. The site now has a consent manager, which is correct and which we wanted, and nothing non-essential loads until a visitor agrees. That means our analytics numbers before and after launch are not comparable, and anyone comparing them is measuring consent rates rather than traffic. Fine, as long as you know.

The zero is separate. Nine days, no conversions logged. Enquiries have come in. The site is not recording them.

So the first honest claim I can make about this relaunch is that search visibility held through the migration. The claim I actually care about, whether the new site turns more visitors into enquiries, I cannot make at all, because I broke the instrument while rebuilding the room. We are fixing it this week.

If you take one thing from this: set up conversion tracking before launch, test it on the live site on day one, and do not take "the tag is on the page" as evidence that anything is being recorded.

That the relaunch worked. We do not know yet.

Nine days is shorter than the window we tell clients not to draw conclusions from, and we have written down why. The long tail has not settled. Seasonality is unaccounted for. The conversion data does not exist. Anyone publishing a success story off this week's numbers is publishing a chart, not a result.

What we will do instead is come back in January with 90 days of data, October to December against the same months last year, with the comparison defined now rather than chosen later to suit whatever happened. If it is worse, that goes up too.

1. 01Expect impressions to spike and position to worsen. Both are arithmetic, not performance. Judge the pages that were already earning, separately, by name.
2. 02Watch your commercial terms specifically. Four or five queries decide whether a migration went well. Write them down before you launch so you are not choosing them afterwards.
3. 03Test conversion tracking on the live site on launch day. Not the staging site. Not the tag manager preview. Submit a real form and watch it arrive.
4. 04If you add consent management at the same time, your analytics history ends there. That is a fair trade and it is not a reason to avoid it, but note the date, because every comparison across it is wrong.
5. 05Wait 90 days before you say anything. Then say it with the comparison you defined on day one.