New Product Launch Ads: A Three-Phase Strategy

New product launch ads should not run as a single campaign. They run in three phases: pre-launch (2-6 weeks of audience and signal building), launch week (budget front-loaded while conversion history is still thin), and scaling (gradual budget increases once the learning phase closes). The split is technical, not stylistic. With zero conversion history, bidding strategies such as target CPA or target ROAS have no distribution to model, so they guess. The first phase therefore collects cheaper intermediate signals, feeds the algorithm, and only then moves to value-based bidding.

Why is product launch marketing different from a normal campaign?

A product launch asks the ad account to create demand for something it has never seen data for. On an existing product the algorithm holds conversion history, lookalike signal and price-sensitivity data; on a new SKU it holds none of that. In practice the first 7-14 days are a teaching period, not a learning one: the goal is reliable signal volume, not day-one ROAS. Teams that skip this distinction kill launches early, judging performance that is not yet measurable.

• Zero conversion history: a new item ID carries no past behaviour on any platform, and similar-product signal only transfers indirectly through the catalogue and feed.
• Unknown price elasticity: you do not yet know how a discount level moves conversion rate, and changing price repeatedly during launch week contaminates the data.
• Creative uncertainty: which benefit actually sells is unverified, so running a creative testing framework before launch saves weeks.
• Time-bound demand: the first 14 days are usually the peak of the interest curve, and impressions missed there do not come back later.

What campaigns belong in the pre-launch phase?

The pre-launch phase starts 2-6 weeks before launch day and optimises for signal, not sales. Three campaigns are enough: broad video views or reach, a traffic or lead campaign for the waitlist, and a light remarketing layer for site visitors. Industry benchmarks report noticeably stronger day-one conversion for launches that open with a waitlist. The real asset, though, is a warm signal pool you can target the moment the product goes live. The metric that matters here is not ROAS but pool size and cost per pool member: how many thousand video viewers, how many hundred waitlist signups, and what each of them cost. Pre-launch usually takes 15-25% of the total launch budget; spending more means buying attention for something nobody can purchase yet.

1. Publish the product page 4-6 weeks before launch with a waitlist form; the page earns pixel signal and organic indexing time at the same time.
2. Run the video campaign on broad targeting and optimise for a 15-second or 50% view rather than 3 seconds, so the remarketing pool fills with better-qualified users.
3. Give the waitlist campaign its own budget and mark the signup as a conversion; during launch week that event is a valuable starting signal.
4. Build visitor and viewer pools at least 10 days before launch, because small pools cannot be scaled on launch day.

Which bidding strategy works with no conversion signal?

Target CPA and target ROAS are unreliable without conversion history, because both model bids from a past conversion distribution. Platform documentation and common agency practice suggest roughly 30 conversions in the last 30 days before moving to target CPA, and about 50 before target ROAS. Below that threshold, maximize conversions without a target produces usable data faster. In 2026 campaign setups Google already presents the target as an optional field, and leaving it empty for the first two weeks is a reasonable default.

How do you manage budget and creative during launch week?

Launch week is front-loaded: allocating roughly 40-50% of the total launch budget to the first 7 days is what closes the learning phase quickly. On Meta an ad set needs about 50 optimisation events within 7 days to exit learning, and a budget too small to produce that volume leaves the campaign permanently in a learning-limited state. Our learning phase guide covers the threshold in detail.

• Use fewer ad sets: consolidate the same budget into two ad sets instead of six, because the 50-event threshold applies per ad set.
• Cap creative at three angles: problem-solution, product demo and social proof. Anything beyond that will not separate statistically inside a single week.
• Keep brand search on: launch coverage and organic interest lift branded queries, and a small brand campaign captures that demand cheaply.
• Do not intervene in the first 72 hours: ad sets paused early are usually paused on differences that carry no statistical meaning yet.

When should you move to the scaling phase?

Scaling starts once the campaign has exited the learning phase and cost metrics have held steady for at least 7 days. Three conditions are worth checking: the ad set is not learning-limited, the last 7 days of CPA sit within roughly 20% of target, and a meaningful share of revenue comes from sources other than remarketing. When all three hold, raise daily budget in 20-30% steps every 3-4 days. Larger jumps reset delivery and reopen learning. The metric to watch while scaling is marginal CPA, not average CPA: the cost of the extra conversions that the last budget increase actually bought. If average CPA looks on target while marginal CPA has doubled, the campaign has grown rather than become profitable.

1. Change one variable per budget increase; adjusting budget and target CPA on the same day destroys any read on causality.
2. Move to target CPA or target ROAS only after the 30-day conversion threshold is met, and set the initial target close to the actual value of the last 14 days.
3. When widening audiences, open geography first and interest layers second; opening both at once hides which expansion worked.
4. Add the new creative batch alongside the winner rather than replacing it, because removing a winner reopens the learning phase.

How should you time a launch before Q4?

Q4 (October to December) is the most expensive advertising window of the year: industry benchmarks put CPMs 30-80% above the annual average, with sharper spikes during Black Friday week. That is why the learning phase of a product launch belongs before Q4, not inside it. A calendar that works in practice: pre-launch in late August, launch week in September, scaling and Q4 preparation in early October. The premium CPM you pay later then buys distribution for a campaign that already works.

The ROAS you measure in launch week is not the product's performance. It is the cost of the algorithm learning.Performance marketing industry practice, 2026

The hardest part of a three-phase launch is looking at the right metric in each phase: signal volume in pre-launch, learning status during launch week, marginal CPA while scaling. Assembling those three views by opening Meta, Google, TikTok and GA4 separately costs time and delays every phase transition. Ads Sensor merges that data in one dashboard, surfaces launch-campaign risks and opportunities as reasoned actions, and keeps automatic before/after tracking on the recommendations you apply. To map the phases onto funnel layers, read our marketing funnel guide, and join the dashboard through early beta signup.