Two Weeks After a Rebuild

The temptation to read too early

A rebuild is the largest single investment many small businesses make in their marketing, and the desire to know whether it worked is immediate.

The data arrives immediately too, which makes the temptation worse.

Almost none of it means anything in the first fortnight. The site is being recrawled, redirects are being followed, and the mix of visitors is temporarily unrepresentative because your own team, your builder, and everybody you told are all looking at it.

Reading that as performance leads to changes made against noise, which is the main way the fortnight after a launch causes damage.

The four things worth watching

• Not-found errors, which reveal gaps in the redirect map.
• Index coverage, which shows the new addresses being taken up.
• Form submissions, which prove enquiries are still arriving.
• Enquiry volume, counted by hand if necessary.

All four are diagnostic rather than evaluative. They tell you whether something is broken, which is the only question the first fortnight can actually answer.

Not-found errors are the highest value

Check them daily, in both your search console and your server logs, and treat the two as complementary rather than duplicates.

The console shows addresses a search engine was aware of. The logs show every address anybody actually requested, including ones from emails, bookmarks, and printed material that no crawler ever saw.

Anything appearing with meaningful traffic behind it is a line the map missed, and adding it in week one recovers nearly all the value.

This is the single most productive thing to do with the fortnight, and it is finite: the list shrinks each day and by week three is mostly noise from automated scanning.

Index coverage tells you the switch took

You are watching for a pattern rather than a number.

Old addresses should be dropping out gradually while new ones are picked up, and the total should end up roughly where it started, allowing for pages you deliberately removed.

What you do not want is the new addresses not appearing at all, which usually means either the blocking instruction was left in place or the sitemap was never submitted.

Expect this to be untidy for a while. Coverage reports lag by days and show a mixture of old and new states, so a confusing picture in week one is normal.

Enquiries are the number that matters

Count them directly, by hand if you have to, and compare against the same fortnight last year rather than against last month.

This is the only number in the first two weeks that carries real information, because it is close to the outcome and hard to misread.

A sharp drop means something is broken, and the something is nearly always a form rather than anything to do with search.

Steady enquiries with a visible dip in traffic is a normal, healthy migration and needs no action at all.

What to ignore

Daily ranking positions, which fluctuate constantly for reasons unrelated to you and are genuinely uninterpretable during a recrawl.

Bounce rate, which will move simply because the page structure changed and which measures something different on a redesigned site.

Time on page, for the same reason.

And day-to-day traffic compared against the previous week, which is dominated at this point by your own people, your builder, and everybody who was told about the launch.

A worked example

A firm launched a rebuild and watched traffic fall about eighteen percent in the first week.

The instinct was that the migration had failed and something should be undone.

Enquiries, counted by hand, were unchanged: eleven in the fortnight against ten the previous year.

The not-found logs showed thirty-one addresses receiving real traffic that had no mapping, mostly documents linked from an old newsletter.

Those were mapped in week one. Traffic recovered to roughly the previous level by week five and passed it by week nine.

Had they judged on the week-one number they would have concluded the rebuild was a failure. The number that mattered had not moved at all.

Give it a quarter before judging

The honest timeframe for assessing whether a rebuild worked is about three months, and six for anything involving search.

That is uncomfortable after a large invoice, and it is the reality of how the underlying systems work.

What you can assess in a fortnight is whether anything is broken. What you cannot assess is whether the site is better.

Setting that expectation before launch, with whoever paid for it, prevents a difficult conversation in week two based on numbers that were never going to mean anything yet.

Write down the baseline before you need it

Worth doing before launch and almost always forgotten.

Record twelve months of traffic, the top twenty pages by visits, enquiry counts by month, and your position for a handful of terms you care about.

Keep it outside the analytics account, in a document, because comparing later against a system whose configuration has changed is unreliable.

Without it, the argument in three months becomes a disagreement about what the numbers used to be, which nobody can settle.

Terms, not positions

There is one search number worth a look in the fortnight, and it is not where you rank.

Your search console lists the terms people used before arriving, and that list is far more stable than the positions attached to it.

What you are checking is whether the same kinds of terms are still bringing people in. If a rebuild reworded the site heavily, it is possible to rank perfectly well and rank for different things, which the position data will not tell you.

A term that disappears entirely usually points at a specific page: either it was dropped, merged into something broader, or rewritten in language nobody searches for.

Compare the fortnight after against the same fortnight a year earlier rather than the month before, and look only at the shape of the list. A term moving from position four to seven is noise; a cluster of terms vanishing is a page to go and look at.

The counter-case

There is one situation where acting immediately is right, and it overrides everything above.

If enquiries stop, or if the site is returning errors, or if the not-found list is enormous, that is not noise and it does not need a quarter of data.

The distinction is between a signal that something is broken, which is actionable on day one, and a signal about whether the site performs better, which is not actionable for months.

Broken is urgent and usually obvious. Better is slow and requires patience, and confusing the two in either direction is the error.

The fortnight

1. Check not-found reports daily, in console and logs.
2. Map anything with real traffic behind it.
3. Watch index coverage for the new addresses appearing.
4. Count enquiries by hand and compare year on year.
5. Ignore rankings, bounce rate, and daily traffic.
6. Diarise a proper review at three months.

Item five is the hardest and the most valuable, because those are the numbers that move most and mean least.

Numbers that move without meaning anything are covered in numbers that move without meaning anything.

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.