Install campaign stops spending after a creative swap
PulseFit, a subscription fitness app, runs two Android app campaigns. The acquisition campaign bids Target CPI $2.50 and drives installs; a second campaign bids Target CPA $9.00 for trial starts. Daily budget is $8,000, split roughly 60/40. The account's historical CPI is $3.10. Two weeks ago the team added a new asset group and paused the three videos that had produced most impressions. Since then the acquisition campaign spends only about 40% of its budget, and delivery collapses by early afternoon instead of running all day. Installs fell from 1,900 to 700 per day. The trial campaign is unaffected and its CPA is steady at $8.40. Conversion tracking was last validated six weeks ago; the MMP dashboard still shows installs arriving. The account manager proposes raising the tCPI to $4.00 and doubling the budget today.
1. What is your diagnosis of the delivery drop, and what would you verify before changing any bid or budget?
Read the worked answer
The pattern matters: spend reaching 40% and then collapsing mid-afternoon is not what a purely low bid looks like. A tCPI below the clearing price usually throttles volume evenly all day. A mid-day cliff points to an eligibility or delivery constraint that appeared around the change. Verify in this order:
- Change history and asset group completeness. Pausing the three top-impression videos removed most of the video inventory. App campaigns need enough creative across formats and aspect ratios to compete in every auction. Check ad strength and whether images and videos now sit below recommended counts.
- Policy and review status at campaign, asset group, and asset level. One disapproved asset will not stop delivery, but a destination or store-listing problem can.
- Conversion action health. Installs appearing in the MMP does not prove Google Ads is recording them. Check the linked account, whether the action shows recent conversions, and whether postbacks are being deduplicated.
- Delivery signals: impression share lost to budget versus lost to rank, plus auction insights. If rank loss dominates, the bid is the constraint; if budget loss dominates, it is not.
- Bid strategy status and learning reset triggered by the structural change.
Only if checks 3 and 4 come back clean does the tCPI hypothesis become primary. Changing bid and budget simultaneously would destroy the ability to attribute any recovery to either lever.
2. If the checks come back clean and the tCPI now sits below the auction clearing price, how would you restore volume without wrecking efficiency? Give a concrete sequence.
Read the worked answer
Treat the fix as a controlled experiment, not a single leap:
- Restore creative inventory first. Reinstating the paused videos is cheaper than paying more per install, and it is the most plausible trigger for lost eligibility. Hold the tCPI at $2.50 for 7 days and see whether delivery returns on its own.
- Then step the bid in 10–15% increments ($2.50 → $2.80 → $3.10), waiting at least 7–14 days between steps. A jump to $4.00 is 60% above target and well above the $3.10 historical CPI; the system will overspend while relearning and the result will not be readable.
- Do not raise budget in the same window. A campaign spending 40% of budget has no budget constraint; adding budget only makes the eventual ramp harder to control.
- Set a stop rule before you start. Compare marginal CPI against the value implied by the trial campaign: a $8.40 trial CPA multiplied by the trial-to-paid rate gives a ceiling. Stop stepping when marginal CPI approaches it.
- Only consider maximize conversions if you can absorb CPI variance. It removes the ceiling along with the cost control, so it suits a launch or a hard volume deadline, not a steady-state account with a payback threshold.
Recommended path: restore assets, hold bid for a week, then step to $2.80. Going straight to $4.00 is a last resort, not a first move.
3. A stakeholder now demands 60,000 installs next month. Using the numbers in the setup, is that achievable, and how do you set expectations?
Read the worked answer
Do the budget math before agreeing to anything. The acquisition campaign receives about 60% of $8,000, or $4,800 per day. At the account's historical $3.10 CPI that buys roughly 1,550 installs per day, or about 46,500 in a 30-day month - with perfect delivery. Hitting 60,000 would require either about $6,200 per day at $3.10 or a CPI near $2.40 at the current budget. Neither is supported by the data, especially right after a two-week disruption and an impending learning period.
Also surface the hidden baseline the stakeholder is using. The pre-drop 1,900 installs per day equals 57,000 per month, but that run rate was achieved with video creative that has since been paused, and it was sustained against a $3.10 market price while bidding $2.50 - it was probably never reproducible at that bid.
Set a realistic plan: restore delivery to 1,300–1,600 installs per day (39,000–48,000 per month) at a CPI of $3.00–$3.30 while creative is reinstated. Present 60,000 as conditional on raising acquisition budget to roughly $6,000–$6,500 per day, and quantify the trade: an extra $45,000–$50,000 of spend for about 12,000–14,000 additional installs at $3.10. Then ask the business question: do those incremental installs clear the trial-to-paid economics? If the answer is no, the target is the problem, not the campaign.