ACoS on a campaign went from 22 percent to 31 percent over two weeks. Was it the bid change on the 14th? The competitor who started advertising on the 16th? The negative keyword list that got updated? The budget increase before the weekend? The price change on the listing?
Without a record, this question takes an afternoon and ends in a guess. With one, it takes thirty seconds. That is the whole case for a change log, and it gets stronger the moment more than one person or tool is touching the account.
What a change log is
A record of every change made to the ad account, in order, with enough detail to understand each one later. Not a report of results; a list of actions. Date, what changed, from what to what, who did it, and why.
It can be a spreadsheet. It can be a shared document. For changes made by software, it should be something the software writes itself. The format matters less than the habit, and the habit matters less than the "why" column.
Why the console is not enough
The console records some history, in some views, with limited search and no export worth relying on. It will tell you a bid is $1.20 now; it is much harder to see that it was $0.90 until the 14th, and it will never tell you why it changed. For an account with one person making a few changes a month, memory covers the gap. For anything larger, it does not.
What to record
Date and time. To the hour if changes are frequent, to the day if not.
Scope. The campaign, ad group, keyword, target, or account-level setting. Specific enough that someone else could find it.
Before and after. The old value and the new one. "Raised bid" is not a record; "bid $0.90 to $1.20" is.
Who. A person's name or a tool's name. When a rule makes a change, the rule's name.
Why. The reason in a sentence. "ACoS above target for 7 days." "Prime Day ramp." "Competitor launched, defending top of search." This is the field that turns a log into a tool for learning, because it lets you check later whether the reasoning held.
Review date, optionally. When to look at whether the change worked. Changes with a review date get reviewed; changes without one get forgotten.
The lightest format that works
One row per change, six columns: date, scope, before, after, who, why. A spreadsheet tab per account. Sorted newest first. That is it.
Resist the urge to add more. The log that gets kept is the one that takes fifteen seconds per entry. A log with twenty fields gets abandoned in a month.
When it earns its keep
Diagnosing a performance change. The example at the top. Sort the log to the week in question, read the entries, and the candidates are in front of you.
Handing off an account. A new person, an agency, a consultant. The log tells them what was tried and why, which is most of what they need and almost never what they get.
Reversing a change. A bid change that did not work needs reversing to the old value. If the old value is in the log, it is a two-minute task. If it is not, the old value is a guess.
Learning what works. Six months of logged changes with reasons and outcomes is a record of which kinds of decisions paid off in this account. That is worth more than any general guide, including this one.
Separating the tool from the market. When a rule or a bid model is running, and performance moves, the first question is whether the tool did it. A tool that logs its changes makes that answerable. One that does not makes it a matter of faith.
Software and the log
This is where the log stops being a habit and becomes a requirement. Any software that changes bids, budgets or campaign state on your behalf is making entries in your account that you did not make. If it does not write them down, you are running an account with an unlogged participant, and every diagnosis starts with "maybe the tool did something."
Ask any tool you evaluate: show me every change you made to a real account last week, with the before, the after and the rule that made it. The ones that can produce that list are the ones you can run with confidence. The reason Off Hours exists is partly this: rules you can read, and a record of every time they fired.
Starting one today
Open a spreadsheet. Six columns. Enter the last three changes you remember making to the account, with the reasons. Then make the rule that nothing changes in the account without a row, and tell everyone who touches it. The first month will feel like overhead. The first time you answer "what happened on the 14th" in thirty seconds, it will not. Budget rules and dayparting schedules belong in the log too, as the one entry that explains a recurring pattern.
Frequently asked questions
Does the Amazon Advertising Console keep a change history?
It keeps a history of some changes in some views, and it is not easy to search or export. For practical purposes, sellers who want to know what changed on a given date and why need to keep their own log or use a tool that records every change it makes.
What should a PPC change log include?
Date and time, what was changed (campaign, ad group, keyword or setting), the before and after values, who or what made the change, and why. The why is the field most often skipped and the one that makes the log useful six months later.
How do I keep a change log when software is making changes?
Choose software that writes one. A tool that changes bids or budgets without a readable record of each change is a tool whose effect you cannot separate from everything else happening in the account. The log should be part of the product, not something you reconstruct.
Off Hours writes a log entry for every change a rule makes: the campaign, the before and after, the rule, and the time. Start a free 14-day trial.