The most common structure I inherit on national accounts is one campaign per state.

‍

It is usually there for a good reason. The sales team needs more volume in Texas this month and needs to pull back in Florida because capacity is full, and a separate campaign per state is the most obvious way to give them that control. I have built accounts this way myself.

‍

The trouble is what it costs on the bidding side, and that cost has grown every year as more of the work moved from the bid to the algorithm.

‍

Let me walk through the math, then the structure we use instead, and the two places it genuinely does not work.

‍

What fragmentation does to the signal

Smart Bidding needs conversion volume to predict who is going to convert. Split the volume and you split the prediction.

Here is an illustration rather than a case study, but the shape of it is right. Take an account doing 2,000 conversions a month nationally.

‍

Consolidated, the bidding model sees all 2,000. It learns that people who search a certain way and behave a certain way are worth more, and it applies that everywhere.

‍

Split across 50 state campaigns, the Texas campaign sees 100, and the Vermont campaign sees 4. The model cannot see that someone in Austin behaves like someone in Miami, because as far as the structure is concerned they are in different campaigns. You are paying to teach Google the same lesson 50 times.

What you get out of it is thin data in every bucket, more volatility month to month, and campaigns that never leave the part of the curve where the bidding is guessing.

‍

National learning, local steering

You do not need a separate campaign to control state volume. You need values.

‍

Group campaigns by what the person wants rather than where they live. One set per product or intent theme, competitors in their own campaign, national geo targeting on all of them. That gives the bidding the whole dataset.

‍

Then handle regional pacing with conversion value rules. A value rule applies a multiplier to conversions from a given region, so instead of fencing Texas into its own budget you are telling the algorithm what a lead from Texas is worth to you right now. Dials rather than fences.

There is a prerequisite that trips people up. Value rules only feed into bidding on campaigns running Maximize conversion value or tROAS. If your campaigns are on Maximize conversions or tCPA, a value rule will change the conversion value column in your reports and have zero effect on what you bid. Value-based bidding comes first, or none of this works.

‍

The workflow

1. Consolidate by intent. Build campaigns around what the person is looking for. National targeting, maximum data in each one.

‍

2. Get real values on the conversions. Assign a value to each conversion event based on national averages, using the step that actually books revenue rather than the top-of-funnel one. This takes the longest and everything else depends on it.

‍

3. Calibrate monthly against the goals. At the start of the month, look at where each region sits against its target.

Behind goal in a state, the old move is to raise that campaign's budget, which usually just buys more expensive volume out of the same pool. The value rule version is a positive multiplier on that state, say +30%. Google sees a higher predicted value and bids more aggressively into that inventory on its own.

Ahead of capacity in a state, the old move is to cap the budget, which throws away efficient leads to stop the flow. A negative multiplier, say -20%, pulls bids back on those leads while the campaign keeps serving everywhere else.

4. Give it a conversion cycle or two before reading it. Changing a value rule changes your value scale, and Google's guidance on value scale changes is 1 to 2 conversion cycles to recalibrate. That is not a learning phase reset, and the difference is worth being precise about, since the idea that any change resets learning gets repeated far more than Google actually supports it. Google says outright that changing a tCPA or tROAS target does not reset anything the bidding has learned. What the recalibration does mean is that a value rule set on the 1st is not something to judge on the 5th.

‍

Where this does not work

Two real limits, and I would rather name them than have someone find out in month two.

‍

Value rules are account-wide on their condition type. Every value rule in the account has to use the same primary condition and the same secondary condition. Build the whole system on location, then decide later that audience is the more useful lever, and you delete every rule and start again. Pick the dimension deliberately.

‍

The second one is an argument against this whole approach, and it comes from Google's own documentation: Smart Bidding already uses geography as a bidding signal. If leads from Texas convert better and that shows up in your data, the algorithm is already bidding them up. A geo value rule only does work when the business wants a region weighted differently than its conversion rate justifies, because capacity is full, or the market is strategic, or the deals there are bigger in a way the conversion tracking does not capture. Those are common and legitimate reasons. They are also different from the algorithm being wrong, and if you cannot say which of the two you are solving for, the rule is guesswork with a percentage attached.

‍

Reporting is the thing people underestimate. Sixty campaigns gave every regional stakeholder their own row in the platform. Consolidation takes that away, and geo reports plus a dashboard are not the same experience for someone used to opening the campaign named after their state. Sort that out before the restructure, not after.

‍

If you are running 60 campaigns to control geography, the structure is doing the pacing job and paying for it in bidding accuracy. Moving the pacing into values gives the bidding everything and keeps the control. It is more work up front, almost all of it in getting real conversion values in place, and that is the part to get right before touching anything else.

‍