Budget Pacer

Two questions, answered together: are we on track to spend what we planned, and is the money sitting in the right places. Pacing without reallocation just means hitting a number badly; reallocation without pacing means running out of money on the 22nd.

Everything this skill produces is a draft for approval. Budget changes move real money and reset algorithmic learning, so the output is a change list someone reads and approves, never an applied action.

Getting the data

1. Connected account — pull campaign/ad set level spend and performance for the current period, plus daily spend so you can see the pacing curve rather than just the total.
2. File or paste — any export with campaign name, spend, conversions (or revenue), and ideally a date breakdown.
3. Ask for what data can't tell you: the budget for the period, the period boundaries (calendar month? campaign flight?), and the target CPA or ROAS. Pacing is meaningless without a target to pace to, and no export contains it.

Also ask, or infer and state your assumption: is the goal to spend the full budget (typical for brand/committed spend) or to spend only what performs (typical for performance accounts with efficiency targets)? These produce opposite recommendations when performance is weak.

Step 1 — Pacing

Compute where spend should be versus where it is:

elapsed        = days elapsed ÷ days in period
expected spend = budget × elapsed
pace           = actual spend ÷ expected spend
projected      = actual spend ÷ elapsed                     (simple run-rate to period end)
required daily = (budget − actual spend) ÷ days remaining   (to land exactly on budget)

Read it as: 0.95–1.05 = on pace, under 0.9 = underspending, over 1.1 = overspending.

Then say what it means in money, not ratios: "You're 19 days into a 30-day month with $29,830 spent — $1,570/day. At that run rate you'll finish at $47,100 against a $40,000 budget, $7,100 over. To land on budget, the remaining $10,170 has to cover 11 days: daily spend needs to drop from $1,570 to $925."

Always compute that corrective figure from the remaining budget, never by trimming the current daily by a fixed percentage — the 20–25% step-size rule below governs per-change increments to individual campaigns, not period-level pacing arithmetic. If the required daily is a steep enough drop to hurt delivery, say so and show the trade-off rather than quietly recommending a gentler number that still misses the budget.

Two refinements worth applying:

• Weekday/weekend shape. Most accounts spend unevenly across the week. A simple run rate on a Monday-heavy account will mislead; if you have daily data, project using the same day-of-week pattern rather than a flat average.
• Underspend is often a delivery problem, not a budget decision. If spend won't rise to meet budget, the cause is usually a limited audience, too-low bids, a restrictive schedule, or disapproved ads — say which, rather than just recommending "raise budgets".

Step 2 — Where the money should go

Classify every campaign or ad set against the target, using the same discipline as any good audit: require enough volume before acting (a floor of roughly one target-CPA's worth of spend, and 3+ conversions before calling something a winner).

• Scale — beating target with real volume, and not budget-constrained by something else.
• Hold — at or near target.
• Trim — meaningfully over target with enough volume to be sure.
• Pause — well past the kill line (2× target CPA with no conversions, or sustained heavy loss).
• Too early — under the volume floor. Explicitly do not recommend changes here; say it needs more time.

Step 3 — Draft the changes safely

Three rules keep the recommendations from doing harm:

Step size: ~20–25% per change, not more. Large budget jumps push ad sets back into the learning phase, where performance is unstable and often worse. Doubling a winner's budget is the single most common way people kill a winner. State the current budget, the proposed budget, and the increment.

Wait between steps. Recommend one step, then 3–4 days (or ~50 conversions) of stability before the next. Give the date to re-check, not a vague "monitor".

Fund from somewhere. By default make the plan net-neutral against the budget: money moved into winners comes out of the pauses and trims. If the plan needs new money, say so explicitly and show the new total.

Also flag anything mid-learning and recommend leaving it alone — changing budget during learning restarts it.

What to produce

BUDGET PACING — [account], [period]
Budget [X] · spent [Y] ([%] elapsed) · pace [Z] · projected [P] → [over/under] by [amount]

DRAFT CHANGES (for your approval — nothing has been applied)

PAUSE
  [name]  $[spend], [conv] conv, CPA [x] vs target [y]   frees $[daily]/day

SCALE  (+20% steps)
  [name]  $[current]/day → $[proposed]/day   CPA [x], [conv] conv
  Re-check [date] before the next step

TRIM
  [name]  $[current]/day → $[proposed]/day

LEAVE ALONE
  [name]  in learning / under volume floor — [reason]

Net effect: $[±]/day, projected period spend $[new projection] vs budget $[budget]

Then a two-line summary of what happens if they approve everything, and what to watch.

Judgment worth applying

Pacing is not the goal; outcomes are. If performance is poor, "spend the remaining budget" is often the wrong recommendation. Say clearly when hitting the budget number would cost more than the results are worth — and note when the reverse applies, because some budgets are committed or use-it-or-lose-it.

Don't recommend scaling into a wall. A campaign beating target on a small, saturated audience won't hold its CPA at triple the budget. Check frequency and audience size before recommending aggressive scaling, and say what the realistic ceiling looks like.

Month-end cliffs are self-inflicted. Slamming budgets down on the 28th to hit a number damages learning right before the new period. Prefer smaller adjustments spread across the remaining days.

Watch for shared budgets. With campaign budget optimization or portfolio bidding, ad-set-level budget changes may not do what someone expects. Note when the structure means the lever they're reaching for isn't the real one.

Never apply budget changes automatically. Present the draft, get explicit approval per change, and if you do have write access, apply only the specific items approved — then confirm exactly what was changed.