How to Recover an Ecommerce Website From a Google Penalty
Almost nobody who searches this has a penalty. The most useful thing this page can do is help you check, because acting on the wrong diagnosis costs months. If you do have one, blocks five and six are what to do about it.
What A Penalty Actually Is
The word penalty covers two entirely different situations. One is a decision made about your site by a person. The other is not a penalty at all. Confusing them sends people down the wrong path immediately.
A manual action. A reviewer has looked at your site and decided something breaches the guidelines.
This is a deliberate decision about your site specifically. It is recorded, it is notified. There is a defined process for asking it to be reconsidered. Block five covers that.
An algorithmic change. Nobody decided anything about your site.
The way sites are assessed changed, so yours is now assessed differently. There is no notification, no record and nothing to appeal, because no decision was made about you. Block six covers what that means.
Why the distinction matters more than anything else here. The responses share almost nothing.
One has a submission process and a defined outcome. The other has neither, so treating an algorithmic change as something to appeal means waiting for a reply that will never come while the actual work goes undone.
What neither of them usually is. What is actually happening to your shop.
Block two is the important one on this page. It is deliberately placed second.
You Probably Do Not Have One
Of the shops that come to us convinced they have been penalised, very few have. That is worth saying before anything else, because the belief itself is expensive.
Why people reach this conclusion. It is the explanation that fits the feeling.
Traffic falls sharply and nothing obvious changed. The drop feels like something done to you rather than something that happened, so penalty is the word available for it.
What acting on it costs. Months, plus damage.
A shop that believes it has been penalised starts removing things. Links get disavowed, pages get deleted, content gets rewritten wholesale. None of that addresses the actual cause. Several of those actions cause harm of their own that then needs undoing.
The specific pattern we see. A repair that becomes the problem.
A shop with a technical fault deletes hundreds of pages in response, which loses whatever those pages earned. Six months later there are two problems where there was one. The original fault is still there.
What to do instead. Check, before acting.
Block three is a sequence that takes less than an afternoon and gives you a definite answer. Nothing on this page below block four should be acted on until you have done it.
Why we lead with this. Because it is the most useful thing we can tell most readers.
Talking somebody out of an expensive diagnosis they arrived with is worth more than any recovery advice we could offer them.
How To Tell Which You Have
Four checks, in this order. The first is definitive and takes minutes, which is why so much worry is avoidable.
One. Check for a manual action directly. This is the only question with a certain answer.
A manual action is recorded and shown to the site owner in the search console for the property. If there is nothing there, you do not have one. That is not an inference, it is the record. Most people asking this question have never looked.
Two. Establish the shape of the decline. Sudden or gradual, everywhere or somewhere.
A manual action produces a sharp drop affecting whatever it covers. An algorithmic change is usually sudden and broad. A competitor improving is gradual and specific. A technical fault can look like any of them, which is why check three exists.
Three. Check whether your pages are still eligible to be found. Pages that exist against pages indexed.
A large drop in indexed pages is a technical fault rather than a penalty. It is the single most common cause of a decline that felt like one.
Four. Establish what changed and when. Take the date the decline started and find out what happened that week.
Including anything a developer, an application or a supplier did without telling anybody. Per our diagnostics guide, self inflicted declines are the most common kind.
What the sequence produces. A definite answer rather than a suspicion.
If check one is clear and check three shows a change, you have a technical problem. If everything is clear and the drop was broad and sudden, block six. If check one shows something, block five.
The Far More Likely Causes
Six explanations that between them cover almost every decline we investigate. Our diagnostics guide covers telling them apart in full.
A migration done badly. The most damaging and the most common.
Addresses changed without redirects, so everything the shop earned was discarded at once. Our migration guides cover both prevention and repair.
A redesign. Where the addresses survived and the content did not.
Category pages lose the text that was earning their position, while everybody agrees the site looks better.
Indexing broken. A directive, a setting or an application excluding pages nobody meant to exclude.
A competitor improving. Your position fell because theirs rose.
Gradual, specific to certain terms, invisible unless you look at what is now above you.
An algorithmic update. Real, covered in block six.
Seasonality. Not a decline at all.
Per our seasonal guide, comparing to last month in a seasonal business produces a collapse the calendar already decided.
What most of these have in common. They are repairable, several of them quickly.
Which is the practical reason for checking before assuming. Most causes on this list are better news than the one you arrived with.
If It Is A Manual Action
A manual action tells you what it covers and why. That is more information than any other situation on this page gives you. It is the starting point rather than the problem.
What the process involves. Three stages, described rather than promised.
Understand what the notice says is wrong. Genuinely fix it. Then submit a request for reconsideration explaining what was wrong, what you did and what will prevent it recurring.
Why the middle stage is the whole thing. Nothing else substitutes for it.
The reviewer is establishing whether the problem still exists. A request submitted before the work is done fails. The shop has lost the time it took to be reviewed for nothing.
What genuinely fixed means. More than most shops assume.
Partial remediation reads as an attempt to do the minimum. Where the issue is content, that means dealing with the pattern rather than the examples cited. Where it is links, it means addressing what was built rather than a sample of it.
What a good request contains. Evidence rather than assurance.
What was wrong, what specifically was done about it, what remains and why, then what has changed so it does not happen again. Documented rather than described.
What we will not tell you. That this will work. Or how long it takes.
Nobody outside the review process controls either. Any supplier telling you otherwise is describing something they cannot deliver, which block seven returns to.
If It Is Algorithmic
There is no submission, no reconsideration and nobody to write to. Understanding that properly changes what a shop does with the following six months.
Why there is nothing to appeal. Because no decision was made about you.
The assessment changed. Your site is being judged by criteria that now weigh things differently. It is one of a great many sites in that position.
What the route actually is. Making the site better on whatever is now weighed more heavily.
That is slower and less satisfying than a submission. It is the only thing available. It is also, usefully, the same work the rest of this cluster describes.
What we will not say. That recovery follows.
It is not guaranteed. Some sites recover, some recover partially, some never return to where they were even after genuine improvement. Anybody telling you otherwise is selling certainty that does not exist.
What that means practically. Judge the work on its own merits.
Do the things that make the shop better for the people using it, because those are worth doing whatever happens next. A programme justified only by a recovery that may not come is a programme built on a promise nobody can make.
The one thing worth knowing. Assessment changes are revisited.
A site that improves genuinely is assessed again in due course. That is a reason for patience rather than a prediction. It is as far as anybody can responsibly go.
What Not To Do
Four responses that are common, understandable and actively harmful. Each is the sort of thing a frightened shop does in the first fortnight.
Disavowing links in a panic. Telling search engines to ignore links pointing at your site.
Where a shop has not built links artificially, most of what points at it is ordinary and some of it is valuable. Disavowing broadly discards that. Our backlinks guides cover why this tool is used far more often than it should be.
Deleting large numbers of pages. Removing anything that might be the problem.
Per block two, this is the response that becomes the problem. Every deleted page takes what it earned with it. A shop cannot tell afterwards which loss was the original decline and which it caused.
Changing everything at once. Structure, content, links and configuration together.
Even where every change is an improvement, nobody can tell afterwards which one mattered. If things get worse, there is no way back to the version that was merely bad.
Hiring somebody promising guaranteed recovery. The most expensive of the four.
Nobody controls whether a site recovers, so a guarantee is a claim about something outside the supplier's control. It is the clearest signal available that whoever is offering it either does not understand the situation or is relying on you not to.
What to do instead. One change at a time, recorded, with time between them.
How Long, Realistically
We will not give you a date. Anybody who does is guessing. What can be given is the shape of each situation and the reasoning behind it.
A technical fault. The fastest, because nothing is being rebuilt.
Where pages were excluded and are restored, the position was earned already. Movement can begin within weeks, though on a large catalogue it takes longer simply because the shop takes longer to be worked through.
A migration repaired. Weeks to months, depending on how much was lost and how completely it is put back.
A manual action. The review itself takes as long as it takes.
The variable nobody controls is the queue. The variable you do control is whether the work was genuinely done, since an inadequate request means starting again.
An algorithmic change. The longest. The least predictable.
Improvement has to be made, then seen, then weighed. Per block six there is no guarantee at the end of it, which is why we describe this in seasons rather than weeks.
Why we refuse to give a date. Because we would be wrong.
None of these has a timetable anybody outside can see. A supplier quoting one has either not understood that or has decided a confident answer is better for winning the work than an accurate one.
What evidence would show it is working. Watch these rather than rankings.
Indexed pages returning. Non-brand impressions rising even before positions move. And the affected section recovering ahead of the rest, which tells you the diagnosis was right. Our measurement guide covers what to record.
When To Accept A Rebuild
Sometimes the site itself is the problem and repairing it costs more than replacing it. That is rare. A page that never says it is not being useful.
When it is genuinely the answer. Three situations.
Where the platform cannot support what has to change. If the fault is structural and the platform will not allow the structure to be different, repair has a ceiling below what the business needs.
Where the shop was built on something that has to go. A catalogue of generated pages nobody wanted. A structure assembled from a supplier feed with nothing beneath it.
Where the repair estimate approaches the rebuild estimate. At which point you are paying similar money for the worse outcome.
What a rebuild does not fix. The thing shops hope it will.
A new site does not clear a manual action, does not reset an assessment and does not remove a history. Rebuilding to escape a problem usually carries the problem across, then adds a migration to it.
The condition we attach. Only with the search work leading the build.
A rebuild that repeats the decisions in our ecommerce web design guides produces the same shop with different colours, plus the loss that any address change involves.
How to decide. Get the repair scoped properly first.
Most shops considering a rebuild have never had the repair costed. Once it is, the decision usually makes itself. It usually goes the other way.
Check before
you act.
A manual action is recorded and shown to you. If nothing is there, you do not have one. Most people asking this question have never looked. The check takes minutes. Acting on the wrong diagnosis costs months.
What is included every month:
£350 per month, one target area. No setup fee, nothing billed separately.
Twenty-two guides.
One subject.
This guide covers penalties and recovery. The rest of the series covers the sequence, structure, category and product pages, technical health, measurement and everything a store owner has to decide.
