The instinct when you take over a messy Google Ads account is to start fresh. Build clean campaigns, proper ad groups, tight keywords, and leave the old wreckage behind.
Sometimes that is right. But the old account is not only a mess. It is also a record of everything that has already been tested with real money, and most rebuilds throw that away without reading it.
This chapter is about what to read first, and about being honest when the answer is that there is nothing worth keeping.
The Account Has Already Paid for the Answers
Every account with any history contains experiments somebody already funded.
Which searches produced calls. Which cities produced customers. Which keywords spent for months and never converted. Which campaign was paused, and whether anybody wrote down why.
You cannot buy that from a keyword tool. It is the only data that describes this business, in this market, at these prices, rather than the category in general. Rebuilding without reading it means paying a second time for answers you already own.
Quality Score Is Attached to the Keyword, and It Took Months to Earn
Here is the finding most people do not expect, and it is the strongest argument for restraint.
Quality score accumulates. A keyword that has been running for a long time with relevant ads and a decent page has built up a history, and that history is what makes its clicks cheaper than a competitor's. It is a real asset sitting in the account, and it does not transfer to a brand new keyword in a brand new campaign just because you typed the same words.
On one account we reviewed, the recommendation was explicitly to build inside the existing structure rather than start over, in order to preserve the quality scores that had built up on some of the keywords. The structure needed fixing. The history did not need destroying, and the two are separable.
So the question is not "rebuild or leave it." It is which parts of this account are carrying earned value, and how do I fix everything else without resetting those.
In practice that usually means restructuring within the account rather than starting a new one. New ad groups can be built around keywords that already have a history. Ads can be rewritten without touching the keyword. Match types can be tightened. What you avoid is the clean-slate instinct that deletes a keyword with two years of standing and retypes it into a campaign created this morning.
Pull All Time, Not Ninety Days
The default reporting window is the enemy here.
Ninety days tells you about the current state of the account, which is often the state after somebody already broke it. All-time tells you what the account did when it was working, and whether it ever was.
We look at both deliberately. Recent windows for current performance, and the full history for structural questions: what has this account ever spent, what did it look like when conversions were highest, when did the shape change.
Then sort the whole history by conversions rather than by cost. Cost tells you where the money went. Conversions tell you what worked, and on a long history the list of terms that ever produced a customer is usually shorter and more interesting than anyone expects.
Two things in particular are worth extracting before you touch anything:
The converting search terms, which are your customers' own words and the best keyword list you will ever be handed.
The converting locations. On one rebuild we pulled years of prior results into a spreadsheet and let the places that had genuinely produced customers define the new target area. The map for that campaign came out of what had happened rather than out of an assumption about the service area.
Paused Is Not the Same as Failed
Paused campaigns are the most misread thing in an inherited account.
A pause usually means somebody stopped, not that somebody decided. Budget moved, an agency changed, a person left, a season ended. The campaign is sitting there looking like a verdict when it is actually an interruption.
So read paused campaigns as evidence rather than as conclusions. A paused campaign with a decent conversion history is a strong signal about what works. A paused campaign with heavy spend and nothing to show is worth understanding before you rebuild something that resembles it.
The same goes for old conversion actions, old naming conventions and old campaign types. They tell you when someone last made a deliberate decision in there, which is useful context for how much of the account to trust.
Sometimes the History Is Not Evidence
Now the honest counterweight, because the argument above can be taken too far.
We reviewed an account with years of accumulated data where the correct call was to largely disregard it. The business had changed substantially: different company, different positioning, different offer. The data was real and it described someone who no longer existed.
We looked at another where the entire history was a starter campaign that had spent about forty-five dollars in total. There was nothing to preserve. Starting fresh was not laziness, it was the only sensible reading.
Three questions settle it:
Is this the same business? Different offer, different market or different pricing means the old results describe a different experiment.
Is there enough volume to mean anything? A few dozen dollars of history is not a finding, it is noise, and the same sample size discipline applies here as anywhere else.
Was conversion tracking working at the time? This is the one people skip. If the tracking was broken during the period you are mining, the history is not a record of what worked. It is a record of what happened to be measured.
Today's Numbers May Belong to Last Quarter
One last thing worth knowing when you inherit an account mid-flight.
Changes have tails. On an account where we had recently switched off an entire state, the following month's figures still carried a residual effect from what had been running before. The numbers on screen were partly describing decisions that were no longer in force.
This matters because the natural move on taking over an account is to judge last month and act on it. If somebody made a significant change three weeks ago, last month is a blend of two different accounts, and reacting to the blend produces a third one nobody designed.
Before you conclude anything from a recent window, find out what changed and when. The change history is in the account, and reading it takes minutes.
There is a version of this that applies even to accounts nobody is inheriting. If you have been running the same account for years, you are the person with the history, and the same exercise works. Most owners have never sorted their own all-time data by conversions, which means the account has been telling them what their best searches are for years and nobody has read it back.
The Twenty Minute Version
- Set the date range to all time before you form any opinion about the account.
- Sort by conversions, not cost. You are looking for what worked, not where the money went.
- Export the converting search terms. That is your keyword list, written by customers.
- Export performance by location and find the places that actually produced.
- Read the paused campaigns as interruptions rather than verdicts, and find out why each one stopped.
- Check the change history so you know which recent numbers describe decisions still in force.
- Confirm tracking was working during the period you are mining. If it was not, the history is not evidence.
- Ask the three questions before preserving anything: same business, meaningful volume, working measurement.
- Preserve earned quality score where you can. Fix structure inside the existing account rather than resetting keywords that have a good history.
Starting fresh feels decisive, and on a genuinely dead account it is the right call. On an account with real history it is an expensive way to relearn things the business has already paid to find out.
We read the history on every Google Ads account we take over, before proposing a single change, because the rebuild plan should be shaped by what already worked rather than by what looks tidy. Hand it over and we will do that reading. Or open your own account, set the range to all time, sort by conversions, and look at the top ten rows. Those are the things your market has already told you it wants.
