In short: restructure inside the same account, never in a new one: what survives (lists, conversions, history) lives in the account; what gets recalibrated is Smart Bidding's learning and the visible QS. The protocol has seven phases — freeze and measure, design, duplicate, overlap, transfer in steps, switch off without deleting, and judge — and it is communicated before you start.
The module closes with the situation every previous case had to go through: changing the architecture of an account that is working. Rebuilding it in one go means losing months of learning and history; not touching it means living with the problem. This lesson's protocol sits in between.
What do you lose when restructuring an account?
| Element | Does it survive a restructure? |
|---|---|
| A keyword's QS history | Tied to the keyword, the ad and the URL within its ad group; an exact copy in another group starts with no visible history, though the account-level and URL-level history remain |
| Smart Bidding learning | Tied to the strategy/portfolio: moving campaigns into another portfolio partly resets it |
| Lists (audiences, negatives, exclusions) | They survive: they live in the account |
| Conversions and their data | They survive: they live in the account |
| Change history and performance history | It survives in the account; a new account starts from zero |
| All of the above when the account itself changes (the account belonged to the agency) | Nothing: the new account inherits nothing |
Hence the first rule: restructure inside the same account almost always; change account only if there is no alternative (ownership).
How do you restructure an account without losing history?
1. Freeze and measure (2 weeks)
No changes. Measurement verified (if it is broken, fix it first and exclude the data). Capture the BEFORE picture: per campaign, CPA/ROAS, conversions, IS, weighted QS. With no BEFORE you cannot judge the change.
2. Design the target structure on paper
Campaigns, portfolios, targets, reserves, listing groups; which current campaigns are kept, which are split, which are merged, which are switched off. An architecture document (principle 7).
3. Duplicate, do not move
Create the new campaigns as copies (the same keyword text, ads and URLs wherever possible: that preserves the history signals that depend on them), with their own budgets, paused.
4. Overlap
Switch the new ones on with a fraction of the budget (30%) while the old ones keep running at 70%; two weeks. The destination portfolios absorb the learning from the start. Keep watch with the judge in TEST mode.
5. Shift the budget in steps
Every week, move 20-30% of the budget from the old campaign to its new equivalent. With Smart Bidding, the new ones go into a portfolio with a realistic target (taken from the BEFORE picture). Four to six weeks in total.
6. Switch off, do not delete
Pause the old campaigns once the new ones match them; do not remove them (the history and any future queries stay accessible). Record the cut-off date in the log.
7. Judge the following month
AFTER against BEFORE with comparable windows and seasonality taken into account; and proof that the change added something, not merely that it did not break anything.
What about MCC, measurement or website changes?
- Changing manager account / agency: the account itself does not change; you unlink the outgoing manager account and link the incoming one. Scripts and lists from the old manager account disappear: they have to be rebuilt (or you keep your own at account level). Export the change history and the reports before unlinking.
- Changing measurement platform (GTM, CMP, server-side): a dual-measurement phase with a new action in observation; once the numbers reconcile, swap primary and secondary; a data exclusion if there was a gap.
- Changing website / domain: the URLs change → the visible URL and landing page history is lost; prepare redirects, update final URLs in bulk (bulk editor), and accept a few weeks of QS recalibration. Warn Smart Bidding with a data exclusion if conversion tracking breaks during the switch.
- A forced new account (ownership): migrate audience lists (they do not transfer; they are rebuilt), negatives (export/import) and the structure (Editor), and accept 2-3 months of learning. Do it in the low season.
What should you communicate before restructuring?
To the client (or to management): what is changing, why, the phased timeline, which metrics will get temporarily worse (conversions in learning, the visible QS of the copies) and when it will be judged. Restructures fail on expectations more often than on technique.
💡 Ninja trick: the scripts' TEST mode is the ideal tool for a restructure: during the overlap the judge records its verdict on both the old and the new campaigns without touching them, Guardian watches that total spend does not drift while the budget shifts, and the change history is documented automatically. Afterwards, the scripts are switched on over the new structure without reinstalling anything: they identify by ID, and the new IDs are configured in their sheet.
What you should remember
- Restructure inside the account; whatever survives lives in the account (lists, conversions, history); Smart Bidding learning and the visible QS recalibrate.
- The protocol: freeze and measure → design → duplicate → overlap → shift in steps → switch off without deleting → judge.
- Manager account, measurement and website changes each have their own variant; a new account only for ownership reasons, in the low season.
- Communicate the timeline and which metrics will temporarily get worse.
End of Module 9. Module 10 is the living module: permanent news — how to keep up with Google Ads every month without getting lost.