Ad action decisions
Which ads to kill, pause, scale, or reduce spend on — and why. The daily action list a Meta specialist spends most of the day on.
Planned
The question: "Which ads should I kill, pause, reduce, or scale today — and when should I start scaling down?"
This is the performance team's daily work. Ad platforms have automatic rules, but they're noisy enough that people still make these calls by hand. Tattva gives the specialist a ranked, reasoned list to act on.
Who asks
- The performance team lead, running the daily review
- Each performance executive, for their own portfolio of ads
Frequency
Daily for active campaigns; more often during sale weeks.
Data you need
adswith daily performance: spend, revenue, clicks, impressions, CTR, conversion ratead_creativeswith hook rate and hold rateads.first_live_date, to compute days since launch (decay context)- Optionally
ads.scale_band, bucketed spend levels. Without it, ask for spend percentiles instead.
How to ask it
- "For the active Meta ads, recommend kill / pause / reduce / hold / scale for each one, with reasoning."
- "Which 10 ads should I scale tonight? Which 10 should I kill? Show the reasoning per recommendation."
- "Top spenders this week with NRoAS below 1.2 — are they worth killing, or is the dip temporary?"
What you'll get back
A ranked action list, one row per ad (figures made up):
| Ad | Spend (7d) | NRoAS (7d) | Trend | Days live | Recommendation | Why |
|---|---|---|---|---|---|---|
| ad-101 | 2,100 | 0.8 | ↓ 28% | 18 | Kill | Below 1.0 with no recovery in 14 days; hook rate falling |
| ad-102 | 3,400 | 2.9 | ↑ 12% | 7 | Scale | Above 2.5, hook rate improving, conversion steady |
| ad-103 | 900 | 1.4 | flat | 32 | Hold | Mature, declining at the expected rate |
| ad-104 | 1,700 | 1.1 | ↓ 8% | 5 | Reduce 30% | Early decline; protect budget, don't kill yet |
Open the Trail to check the figures behind each row: the Compute card shows the periods, filters and SQL.
The logic tree
For each active ad:
├── Is spend below a minimum (too little data)?
│ └── Skip — needs more data
├── Is NRoAS below the kill threshold AND trending down over 7 days?
│ └── Kill
├── Is NRoAS above the scale threshold AND trending up over 7 days?
│ ├── Is spend already at the top of its band?
│ │ └── Hold (already maxed)
│ └── Scale (raise spend by your usual step)
├── Is NRoAS fine but hook rate decaying past the usual period?
│ └── Pause (creative fatigue)
├── Is conversion dropping while NRoAS holds?
│ └── Investigate landing page / payment (don't touch the ad)
└── Otherwise: HoldPut your thresholds in the question ("kill below 1.0, scale above 2.5, at least 7 days live"), or ask your data analyst to record them as conventions so every answer uses the same ones.
Rolling it out
Tattva recommends; your team acts in the ad platform. Tattva doesn't change your ads.
- Weeks 1–4: read the list each morning, apply your own judgement, act. Note which recommendations you accept.
- Weeks 5–8: where the team agrees with a recommendation type most of the time (say, Kill), act on it with less review.
- After that: keep reviewing the types where you still disagree, and tell Tattva why with Not satisfied under the answer. Your analytics team reviews each comment and fixes the ontology, so later answers improve for everyone.
How to make it recurring
- Click Save as tracker on the answer. The tracker is created with a daily cadence.
- On the Delivery tab, add the team's Slack channel so the list arrives with each run.
- Link it to your marketing-efficiency goal when you create the goal; see Marketing efficiency.
Trackers have no time-of-day setting. The next-run time is shown on each question; use Run now before your stand-up if you want the freshest list.
Pitfalls
- Killing too early. A 5-day-old ad with one bad day is noise, not a kill. Set a minimum days-live in the question.
- Scaling on a thin sample. Three good days on low spend is a test-scale, not a full scale.
- Ignoring decay. Mature ads should decline; flagging them as kills wastes attention.
- Treating all products alike. SKU-A's ads may decay faster than SKU-B's. Ask per product.
Related
- NRoAS drop RCA
- Creator pattern trends
- Custom tools: bring in data from another service over HTTP
NRoAS drop RCA
The weekly performance-marketing question, "Why is my acquisition NRoAS not holding up?", answered with the right tables, the right breakdown, and a trail you can check.
Audience overlap diagnosis
Are two ad sets eating each other's new-customer reach? The most common deep dive within an NRoAS drop.