Account Restructures That Actually Improve Performance (And the Ones That Just Reset Your Learning Phase)

I get pitched a 'restructure' almost every time I take over an account from another agency. New campaign structure, new naming convention, everything rebuilt from scratch. Sometimes that's genuinely the right call. Most of the time it's an agency demonstrating effort instead of judgment — and the client pays for it in a multi-week performance dip they were never warned about.

Any restructure that creates new campaigns, new ad groups, or moves keywords into different bidding contexts throws away the conversion history those algorithms were using to price bids. During that reset window, cost per conversion typically climbs and volume gets erratic — not because the new structure is wrong, but because the system is relearning from near-zero. If the restructure isn't fixing something that's actually costing you money, you're paying that price for nothing.

• Splitting campaigns by margin, not just by product category — so Smart Bidding can pursue a genuinely different target ROAS for a 60%-margin line versus a 15%-margin line instead of averaging them into one blended, wrong number for both.
• Consolidating fragmented ad groups that were split by irrelevant variables (color, size, minor keyword variants) into structures with enough conversion volume per ad group for Smart Bidding to actually learn a pattern.
• Migrating from Standard Shopping into Performance Max, or restructuring how Shopping and PMax coexist, when the current setup has them bidding against each other for the same queries — a problem I cover in detail in Standard Shopping vs. PMAX: How to Structure Both Without Them Fighting Each Other.
• Separating branded and non-branded spend when they've been mixed into one campaign and you can't tell how much of your 'performance' is just people searching your company name.

Each of these fixes a structural problem that caps performance no matter how much you optimize inside it. The dip is the cost of removing that ceiling. That's a trade worth making — and worth explaining to a client in plain terms before you make it.

• Rebuilding an account with a new naming convention and campaign shells but the same targeting, same bids, same ad groups underneath — busywork disguised as strategy.
• Splitting single-keyword ad groups into even more granular single-keyword ad groups in a Smart Bidding account, which just fragments conversion data further with no upside.
• Migrating to a new campaign type because it's newer, not because the current one is underperforming for a specific, identifiable reason.
• Recreating campaigns instead of pausing and rebuilding the parts that are actually broken, when a targeted fix would've left historical signal intact.

If you can't name the specific metric a restructure is supposed to move, it's not a strategy — it's an agency looking busy on your budget.

Before I approve any structural change on an account I run, I make the person proposing it answer three questions. What specific limitation in the current structure is capping performance, and can you point to the data showing it? What's the expected recovery timeline, and what's the plan for that window — do we pause spend, protect budget on unaffected campaigns, or just absorb the dip? And is there a narrower version of this change that fixes the same problem without touching campaigns that are already performing well. That last one matters most. A full rebuild and a targeted fix often solve the same problem, but only one of them resets your entire learning phase instead of one segment of it.

I'd rather split off the underperforming 20% of an account, restructure that, and leave the healthy 80% untouched than blow up everything for the sake of a clean slate. It's slower to explain in a proposal deck. It's also how you avoid a four-week revenue dip for a problem that only affected a fraction of spend.