Short answer
Your baseline is the twenty-eight days before the drop began, agreed and recorded in writing before any work starts. Measuring from the bottom of a decline flatters every result. If Search Console no longer holds that period, say so openly and use the earliest clean window available instead.
Every argument about whether a recovery worked is really an argument about the baseline.
Set it once, in writing, before anyone has a result to defend. It takes ten minutes and prevents months of dispute.
Which period to use
- Twenty-eight days, not thirty. It covers exactly four weeks and removes day-of-week distortion.
- Ending the day before the break date, not the day you noticed.
- A normal period. Not your best month ever, and not a promotional spike.
Find the break date with the drop analyzer rather than estimating it.
Why the break date matters more than the period
Get this wrong and everything downstream is wrong.
The date you noticed is almost never the date it started. Most people notice a decline one to three weeks late, because weekly reporting smooths the first few days out of view.
If you place your baseline using the date you noticed, the first two weeks of the decline end up inside your baseline. That drags the comparison figure down and makes any later recovery look better than it was.
- Use daily data, never weekly or monthly. The break is invisible at weekly resolution.
- Look for the step, not the lowest point. Declines often continue for weeks after they begin.
- Check whether there are two steps. Compounding declines are common and each needs its own baseline.
Exclude these periods
A baseline containing any of these is not a normal period.
- A viral post or news spike. One page can double a month and never repeat.
- Black Friday, Christmas, or your annual peak. You will be measured against an outlier for the whole project.
- A campaign period. Paid activity lifts branded organic search too.
- A previous outage. An artificially low month makes recovery look easy.
- The weeks around a site migration. Those figures are already unstable.
If your only clean window sits further back, use it and label the gap rather than accepting a distorted one.
What to record
| Record | Why |
|---|---|
| Clicks, impressions, CTR, average position | All four. Clicks alone hides displacement |
| The same four for the affected page group | Site totals hide partial recoveries |
| Top 50 queries with positions | Shows breadth returning before volume does |
| Indexed page count | Separates ranking loss from indexing loss |
| The same period one year earlier | Removes seasonality from the comparison |
Export it now
Search Console keeps sixteen months. After that your baseline is gone permanently.
Export the full sixteen months today, whatever stage you are at. It costs nothing and it is unrecoverable once it ages out.
If the data is already gone
Common on sites down since 2023. Say it plainly rather than working around it.
- Use the earliest clean window Search Console still holds, and label it as such.
- Check analytics separately. It often retains longer than Search Console.
- Accept a lower target. You are measuring against a diminished baseline, not the original one.
One baseline is not enough
Set three, not one. They answer different questions.
| Baseline | Answers |
|---|---|
| Site-wide | What the business lost overall |
| Affected page group | Whether the work is actually doing anything |
| Unaffected page group | Your control. Did the rest of the site move too? |
The third is the one almost nobody sets, and it is what turns a claim into evidence. If the affected group rose while the control stayed flat, the work caused it. If both rose together, something else did.
Write it down where others can see it
A baseline that lives in one person’s spreadsheet is not agreed.
Put it somewhere the client, the manager and whoever replaces you can find it. Recovery projects routinely outlast the people who started them.
- The exact dates of the baseline period.
- The break date and how you established it.
- The four figures, site-wide and per group.
- The URL pattern defining the affected group.
- A note on anything unusual in the period, if you had no choice.
Presenting this to stakeholders is covered in how to explain a traffic drop to your boss.
Three ways baselines get gamed
Usually without anyone intending it.
- Starting from the bottom. Measuring from the lowest point makes any movement look like success.
- Switching baselines mid-project. If the comparison changes, the result means nothing.
- Counting new pages. Content published during the work is new traffic, not recovery.
Verification itself is covered in how to know if the recovery actually worked.
A worked example
Same site, two baselines, opposite conclusions.
- Drop began 12 March. Noticed 2 April.
- Baseline set from the noticed date: 5 March to 1 April = 41,000 clicks. Three weeks of decline are inside it.
- Baseline set from the break date: 12 February to 11 March = 58,000 clicks.
- Six months later the site reaches 46,000 clicks.
Against the first baseline that is a 12 percent gain and a success. Against the correct one it is 21 percent still missing.
Same work, same result, opposite verdicts. That is why the date is worth twenty minutes before anything else starts.
Not sure when your drop actually began?
The date you noticed is rarely the date it started, and the wrong break date gives you the wrong baseline. Send your Search Console access and within 48 hours you will have the real one.
Get free recovery audit