agentleFS
scumunnascumunna/programmatic-skills/skills/dv360-pacing-and-optimization/SKILL.md

dv360-pacing-and-optimization

Skill1 starsChanged 5 months agoMITView on GitHub

Diagnose and fix pacing, and run the optimization loop in DV360. Use when the user asks about pacing, under-delivery, over-delivery, spend is behind or ahead, a campaign that will not spend its budget, impression loss, even vs ASAP vs ahead pacing, daily vs flight pacing, how to optimize a campaign, why a line item is underpacing, or how to reallocate budget. Covers the pacing math (even daily target, expected spend to date, pacing index) and the ordered fix trees for under and over-delivery.

The file

We split this skill into the rules it gives an agent

8 rules. About 0 are common, 6 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use when the user wants to:

    line 14
  2. Distinctive:

    Read the impression loss breakdown and decide what to fix.

    line 19
  3. Distinctive:

    Run an optimization cadence over a flight.

    line 20
  4. Very specific:

    If a fix is "loosen the frequency cap" or "relax brand safety or viewability," make the change with dv360-frequency-and-brand-safety, which …

    line 26
  5. Very specific:

    Run the numbers before adjusting anything.

    line 54
  6. Distinctive:

    Recovering a behind flight: prefer raising the bid or daily cap and loosening the single binding filter over switching to ASAP, because …

    line 96
  7. Distinctive:

    Fix the named loss category, not a guess: If loss is dominated by below-minimum-bid, more budget will not help.

    line 98
  8. Distinctive:

    Use the Optimization view: https://support.google.com/displayvideo/answer/7563784 (as of June 2026)

    line 130

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

DV360 pacing and optimization

Keep spend on track to the flight and steer it toward the KPI. This skill encodes pacing settings, the pacing math, ordered fix trees for under-delivery and over-delivery, and the daily optimization loop. A pacing problem is almost always a delivery constraint upstream, so the discipline is to read the impression loss breakdown before touching bids.

For shared definitions (CPM, win rate, the KPI math) see the programmatic-foundations skill. This skill ships a pacing_calculator.py helper for the arithmetic.

When to use this skill

Use when the user wants to:

  • Diagnose under-delivery (spend behind) or over-delivery (spend ahead, or budget gone too fast).
  • Choose or change a pacing setting (ASAP, even, ahead of schedule; daily vs flight; amount vs impressions).
  • Compute even daily target, expected spend to date, or a pacing index.
  • Read the impression loss breakdown and decide what to fix.
  • Run an optimization cadence over a flight.

Boundary with sibling skills:

  • For pulling and scheduling the underlying reports and metric definitions, use dv360-reporting. This skill reads the numbers; that skill produces them.
  • For deep no-delivery, low-win-rate, viewability, and creative-disapproval playbooks, hand off to dv360-troubleshooting. This skill orders the likely causes; that skill is the exhaustive runbook.
  • If a fix is "loosen the frequency cap" or "relax brand safety or viewability," make the change with dv360-frequency-and-brand-safety, which owns those controls.

Quick reference

SymptomFirst checkLikely fix
Spend behind flightImpression loss breakdownRaise bid or budget, widen targeting, loosen the binding filter
Spend gone too earlyPacing setting and bidSwitch ASAP to even, lower bid, tighten audience or cap
Pacing index < 1.0Expected vs actual spendDiagnose the loss category, then act
Lost to auctions / below min bidBid vs floor, win rateRaise bid or bid strategy target
Lost to budget / pacingDaily cap, flight budgetRaise daily cap or reallocate budget
Lost to frequencyFrequency capLoosen cap (in dv360-frequency-and-brand-safety)
Lost to viewability / brand safetyViewability threshold, exclusionsRelax the control to the KPI level
Lost to no eligible creativeCreative status and specsFix or add a compatible creative

Pacing settings

Set pacing at the insertion order and the line item. The IO budget caps the combined line item spend regardless of individual line item budgets, the same way it caps frequency.

  • Pacing type:
    • Even spends a steady amount each day, scaling to daily inventory availability. The default for most flighted delivery.
    • ASAP spends as fast as inventory allows (roughly an order of magnitude faster than even). Use to capture limited or time-sensitive inventory, or to recover a badly behind flight, knowing it front-loads spend.
    • Ahead of schedule targets up to about 120% of the even daily amount to finish the budget before inventory or competition tightens late in the flight.
  • Pacing period: daily (spend the daily amount each day) or flight (spend the whole budget across the flight). Daily gives smoother control; flight gives the system more room.
  • Budget unit: amount or impressions. An IO and its line items must use the same unit.

Pacing math

Run the numbers before adjusting anything; "behind" is only meaningful against an expected position. The pacing_calculator.py script computes all of these from a budget, flight dates, an as-of date, and spend to date.

  • Even daily target = remaining budget / remaining days. At flight start this equals total budget / total flight days. Recomputing on remaining budget and days is what lands the flight exactly on budget after a slow or fast stretch.
  • Expected spend to date = even daily target at flight start times days elapsed. Where even pacing would have you today.
  • Pacing index = actual spend / expected spend to date. 1.0 is on pace, below 1.0 is behind (under-delivery), above 1.0 is ahead (over-delivery). Treat roughly 0.95 to 1.05 as on pace and investigate outside that band.
  • Projected end-of-flight spend = current daily run rate times total flight days. If this lands below budget, you will under-deliver unless the daily target rises.

Count flight days inclusive of both the start and end date, because that is how the flight is booked.

Under-delivery: ordered causes and fixes

Work top down. Each cause is roughly ordered from most common and cheapest to fix, to least. Confirm the cause in the impression loss breakdown before acting, so you fix the binding constraint and not a symptom.

  1. Budget or daily cap too low. A daily cap below the even daily target throttles delivery. Raise the daily cap or reallocate budget from an underperforming line item.
  2. Bid too low / losing auctions. Lost-to-auction or below-minimum-bid loss means you are priced out. Raise the bid or the bid strategy target (see dv360-bid-strategy).
  3. Targeting too narrow. A small audience or tight geo/device/contextual filter starves the line item. Broaden the most restrictive dimension.
  4. Frequency cap too tight. Frequency-limited loss means the cap is exhausting available users. Loosen it in dv360-frequency-and-brand-safety.
  5. Viewability or brand-safety filters too strict. High predicted-viewability targeting or aggressive exclusions cut the eligible pool. Relax the control to the KPI level, not below it.
  6. Deal not delivering. A PMP or Programmatic Guaranteed deal can under-supply. Check deal status and inventory in dv360-deals-and-inventory.
  7. Creative disapproved or missing. No-eligible-creative loss means the line item cannot serve. Fix the disapproval or add a compatible creative.

Over-delivery: causes and fixes

  1. ASAP pacing front-loading spend. Switch to even (or ahead) so the budget lasts the flight.
  2. Bid too high. Winning expensive impressions burns budget fast. Lower the bid or bid strategy target.
  3. Audience too broad. A wide pool plus a healthy bid over-delivers. Tighten targeting.
  4. Frequency cap too loose. Spending repeatedly on the same users. Tighten the cap.

Optimization workflow

  1. Monitor pacing vs flight and performance vs KPI together. Pacing is the rate of spend toward budget; performance is the value from impressions bought. They are different questions; a line item can pace perfectly and miss the KPI, or hit the KPI while underpacing. Read both.
  2. Compute the pacing index with the script. If outside 0.95 to 1.05, treat it as a delivery problem to diagnose, not a number to force.
  3. Read the impression loss breakdown to see where impressions are lost: auctions, below minimum bid, budget or pacing, frequency, viewability, brand safety, or no eligible creative. This names the binding constraint.
  4. Segment by dimension (exchange, device, app or URL, geo, creative) to find where the loss or the underperformance concentrates, rather than adjusting the whole line item blindly.
  5. Act on one lever at a time: bid, budget reallocation, targeting, or creative. Changing several at once makes the next read uninterpretable.
  6. Re-check after the system settles. Optimization-view data can lag (it can take up to about three days to populate after setup), and bid strategies relearn after a change, so do not stack changes faster than you can read their effect.

Cadence: daily in week one while delivery stabilizes, then a few times a week through the middle of the flight, then watch more closely near the end when the remaining-days math makes each day's target swing harder.

Decision rules and thresholds

  • Pacing index bands: below 0.95 is behind, above 1.05 is ahead, between is on pace. The further from 1.0 and the later in the flight, the more urgent, because fewer remaining days mean a larger daily correction.
  • Recovering a behind flight: prefer raising the bid or daily cap and loosening the single binding filter over switching to ASAP, because ASAP front-loads and can over-correct. Use ASAP only when inventory is genuinely scarce or time-boxed.
  • One change per read. Adjust a single lever, let it settle, re-read. This keeps cause and effect legible and protects bid-strategy learning.
  • Fix the named loss category, not a guess. If loss is dominated by below-minimum-bid, more budget will not help; the bid is the constraint. If loss is budget or pacing, a higher bid wastes money; raise the cap or reallocate.

Reference material

  • scripts/pacing_calculator.py: computes even daily spend, expected spend to date, pacing index, projected end-of-flight spend, and the recommended new daily spend to finish exactly on budget. Run it with --help for usage. Use it whenever the user gives a budget, flight dates, and spend to date.

Templates and examples

Underpacing line item, mid-flight. A $30,000 line item runs 2026-06-01 to 2026-06-30. On 2026-06-10, $8,000 is spent. Run:

python3 scripts/pacing_calculator.py --budget 30000 --start 2026-06-01 --end 2026-06-30 --as-of 2026-06-10 --spend 8000

Output: even daily target $1,000, expected to date $10,000, pacing index 0.80 (behind), projected end-of-flight $24,000, recommended new daily $1,100 for the remaining 20 days. The line item is behind and would leave $6,000 unspent. Next step: read the impression loss breakdown. If loss is dominated by lost-to-auction, raise the bid; if by budget or pacing, raise the daily cap toward the new $1,100 target; if by frequency or viewability, loosen that control in dv360-frequency-and-brand-safety.

Over-delivering insertion order. Budget is spending at a 1.30 pacing index four days in, set to ASAP. The deal it targets is not scarce. Fix: switch the IO to even pacing so the budget lasts the flight, and lower the bid if the index stays high after the change. Re-read after the system settles before any further change.

Even-paced but underperforming. Pacing index 1.00 but cost per action is above the KPI. Pacing is fine, so do not touch budget. Segment by exchange and creative, find where the bad outcomes concentrate, and act on bid or creative there. This is a performance problem, handed in part to dv360-bid-strategy.

Common pitfalls

  • Reacting to "behind" without computing expected spend. Early in a flight a low absolute spend can still be on pace. Run the index first.
  • Reading pacing and ignoring performance, or the reverse. They answer different questions; check both.
  • Changing bid, budget, and targeting at once, then being unable to tell what worked. One lever per read.
  • Throwing budget at a line item whose loss is below-minimum-bid. Wrong constraint; raise the bid.
  • Switching to ASAP to recover a behind flight and then over-delivering early. Prefer bid, cap, and a single filter loosen first.
  • Re-tuning faster than data populates or a bid strategy relearns, churning without signal.

Sources

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.