Inside Adriva's Hourly Budget Rebalancing Across Google and Meta

Budgets set on Monday are wrong by Wednesday. Demand moves by the hour, but most accounts are reviewed once a day at best. Adriva closes that gap: it watches Google and Meta continuously and moves money toward the outcome, not the dashboard.

Why static budgets go stale

A daily budget is a bet you place once and cannot adjust until tomorrow. It assumes the account you set up this morning is the account you have tonight. It never is. Search intent spikes and fades, auction pressure rises when competitors turn campaigns on, and conversion rates move with the clock. Within peak windows, specific hours can convert at two to three timesthe daily average, and off-peak hours can run 40–60% below it — documented dayparting patterns that most accounts leave on the table.

Reviewing budgets once a day means you are always acting on yesterday. You see that a Meta prospecting campaign printed money on Tuesday, so you raise its budget Wednesday morning — after the window that made it work has closed. The manual companion to this problem, deciding when to shift spend between Google and Meta, is a genuine skill. It is also impossible to do well by hand at an hourly cadence across a dozen campaigns and two platforms. That is the job Adriva was built to run.

Daily budgets optimize for the calendar. Demand optimizes for the hour. Adriva runs on the same clock as demand.

What Adriva watches every hour

Every hour, Adriva pulls fresh delivery and conversion data from the Google Ads and Meta APIs for every active campaign. It is not looking at vanity totals. It reads three signals that actually predict where the next dollar should go.

Budget pacing

Is a campaign on track to spend what it was given by end of day, or is it under-delivering and leaving reach unused? Pacing tells Adriva which campaigns have appetite and which are starved. A campaign pacing at 130% of plan by noon is either a demand spike worth funding or a leak worth capping — the other signals decide which.

Marginal ROAS and CPA

Lifetime averages lie to you. A campaign with a 6x blended ROAS can still be a bad place for your next dollar if it is saturated and the marginal return on additional spend has collapsed. Adriva estimates what the next increment of budget is likely to return — the slope, not the average — and ranks campaigns by that. Money flows toward the campaign where the next dollar works hardest, on either platform.

Intraday demand

The same audience behaves differently at 9am and 9pm. Adriva tracks how conversion rate and cost move through the day per campaign and per platform, so it can fund a channel during its strong hours and ease off during its weak ones instead of spreading budget flat across all twenty-four.

Reading demand this precisely depends on clean signal, which is why Adriva leans on the measurement layer it sets up when you connect an account. If you want the plumbing, see how Adriva connects Google Analytics and the pixels. Better conversion data in means sharper marginal-return estimates out.

The rebalancing loop

Rebalancing is not a single model deciding in a black box. It is a loop with checks at every step, and it runs the same way every hour.

1. Read. Pull the latest pacing, conversion, and cost data from both platforms for every active campaign.
2. Rank. Estimate marginal ROAS or CPA for each campaign and order them by where the next dollar is likely to work hardest right now.
3. Propose. Identify the smallest move that improves the portfolio — pull a small step from a saturated or off-peak campaign, add it to one with headroom and strong marginal return.
4. Check guardrails. Confirm the move stays inside caps, floors, step-size limits, and learning-phase thresholds before anything touches the account.
5. Apply or hold. If the move clears every check and the confidence is high enough, apply it through the platform API. If not, hold — most hours, holding is the right call.
6. Log. Record the change, the signals behind it, and the projected impact so you can see exactly what moved and why.

From signal to action

Each hour’s reading maps to a specific response. The logic is legible on purpose — an operator should be able to look at any decision and agree with the reasoning.

Guardrails and the learning phase

Moving budget hourly is only safe if it respects how the platforms learn. Both Google and Meta run a learning period during which their models gather enough conversions to deliver stably, and large edits reset it. Rebalancing that ignored this would trade a small allocation gain for a large delivery setback. Adriva is built to avoid exactly that.

On Meta, an ad set needs roughly 50 optimization events per week to stabilize, and disruptive edits push it back into learning. On Google, budget changes above about 20% are the ones that tend to trigger a reset. Adriva treats these as hard constraints:

• Small steps, not swings. Increments stay inside the thresholds each platform treats as safe, so no single hourly move restarts learning.
• Learning-aware holds. Campaigns and ad sets still in active learning are largely left alone until they stabilize; Adriva adds spend around them, not through them.
• No strategy churn. Rebalancing moves budget only. It does not switch bid strategies, rewrite conversion actions, or make the kind of structural edits that reset learning on their own.
• Caps and floors. You set the ceiling a campaign can climb to and the floor it cannot drop below. Adriva optimizes strictly inside that envelope.

This is also why hourly beats daily without beating up the account. The gain from hourly rebalancing does not come from big, frequent moves — it comes from many small, well-timed ones that a daily review would miss entirely. Frequency buys precision, and guardrails keep that precision from turning into churn. If you want the mechanics of the bidding layer underneath, we cover automated bidding strategies separately.

What the operator sees and controls

Automation you cannot inspect is just a black box with better marketing. Adriva is the opposite. Every rebalancing decision is transparent, and the operator keeps directional control.

You see a running log of what moved: which campaign gained budget, which gave it up, the signals that drove the call, and the projected impact — the same “shift this into prospecting, here is the expected return” framing Adriva uses across the product. Nothing happens that you cannot trace after the fact.

And you set the rules before the fact. Caps and floors bound every campaign. You can mark a campaign no-touch so Adriva reports on it but never moves its budget. You can leave rebalancing fully automated, or drop it into recommend-only mode, where Adriva surfaces the exact same moves for one-click approval instead of applying them. The operator decides how much rope to give the system, and can take it back in one click.

The goal is not to remove you from the loop. It is to remove the parts of the loop that never needed a human.

That is the whole thesis. Pacing, marginal-return math, and intraday timing are work — necessary, repetitive, unforgiving work that runs around the clock across two platforms. Adriva takes that load and hands it back as decisions you can see, bounded by rules you set. You keep the strategy. It keeps the clock. Connect your ad accounts and let Adriva run the rebalancing you were never going to do by hand every hour.