New Product Launches Fail in Google Ads - Here's the Bidding Fix

The Cold Start Problem

Smart Bidding is a prediction engine. It predicts conversion probability based on historical signals - device, location, time of day, audience, and crucially, product performance history. A product with 500 conversions in the last 90 days gives the algorithm rich data to work with. A product with zero conversions gives it nothing.

Faced with uncertainty, Smart Bidding does what any rational algorithm would: it minimises risk. It bids conservatively on products with no history because it can't predict conversion probability with confidence. Meanwhile, it bids aggressively on established products because it has high-confidence predictions.

The result is a structural bias toward incumbents. Established products get more impressions, more clicks, more conversions - which gives the algorithm more data, which makes it more confident, which drives more spend. New products enter a negative feedback loop where low impressions lead to low data, which leads to low confidence, which leads to fewer impressions.

This isn't a bug in Smart Bidding. It's exactly how the system is designed to work - it maximises overall conversion value by allocating budget to the most predictable opportunities. But "most predictable" doesn't mean "most strategically important," and your new product launch might be critical to your commercial roadmap even if Smart Bidding can't yet predict its performance.

Smart Bidding Hates New Products

The bias against new products operates at multiple levels within Google's algorithm:

• • Quality Score deficit: New products start with no click-through rate history. Quality Score defaults to average or below-average, increasing your effective CPC relative to established competitors who have accumulated strong CTR data.
• • Product-level signals: In Shopping campaigns, Google uses historical product performance as a ranking signal. New products with no performance data rank lower than established products in the Shopping auction, even at equivalent bids.
• • Asset group competition: In Performance Max, new products compete for budget against established products within the same asset group. The algorithm always favours the known performer. If your new product is in an asset group with proven best-sellers, it will be systematically starved of impressions.
• • Missing social proof: Products without reviews, ratings, or significant sales history convert at lower rates - not because the product is inferior, but because customers default to proven options. This lower conversion rate becomes self-reinforcing in Smart Bidding.

The compound effect means new products can take 6-12 weeks to reach steady-state performance - if they survive that long in a Smart Bidding environment. Many don't. Budget gets silently redirected to established SKUs that the algorithm already trusts, and the new product never gets enough data to prove its potential.

The Data Starvation Spiral

Here's how the spiral typically unfolds for a new product:

• • Week 1-2: Product receives low impressions. Smart Bidding has no data, bids conservatively. The product gets 200 impressions and 8 clicks. Zero conversions. The algorithm's confidence remains at zero.
• • Week 3-4: Smart Bidding further reduces bids based on the low CTR and zero conversions. Impressions drop to 80. Clicks drop to 3. The product is effectively invisible.
• • Week 5-6: The marketing team notices the new product "isn't performing" and reduces its priority. Budget is explicitly redirected to proven products. The new product receives near-zero spend.
• • Week 7-8: Someone flags that the new product launch "failed." The product is removed from paid campaigns. Its potential is never realised because it was never given enough data to prove itself.

This spiral is especially damaging for seasonal launches. A new spring/summer product that enters the spiral in April will have missed its peak demand window by the time enough data accumulates in June. By then, the buying window is closing.

The spiral also creates a survivorship bias in your catalogue. The products that "work in Google Ads" are the products that happened to get enough early data - often because they benefited from external traffic, PR, or organic search. Products that relied solely on paid for discovery rarely survive the cold start unless you intervene.

Launch Bidding Strategies

Five approaches to overcome the cold start problem:

• • Dedicated launch campaigns: Create a separate campaign exclusively for new products. Set a manual CPC or maximise clicks strategy with a fixed daily budget. This prevents established products from cannibalising the new product's spend and forces Google to show the new product.
• • Staged bid strategy transitions: Start with manual CPC (weeks 1-3) → maximise conversions with no target (weeks 4-6) → target ROAS at 30% below your standard target (weeks 7-10) → standard target ROAS (week 11+). Each stage feeds data into the next.
• • Budget protection: Allocate 10-15% of your total Google Ads budget to new product launches, ring-fenced from optimisation. This budget should be treated as investment, not held to the same ROI standards as established products during the first 60 days.
• • Cross-channel data seeding: Drive initial traffic and conversions through email marketing, social ads, or influencer seeding before activating Google Ads. Even 20-30 early conversions from other channels give Smart Bidding a baseline to work from when you activate paid search.
• • Audience targeting for launch: Target new product ads to your existing customer base first. These audiences have higher conversion rates, which gives Smart Bidding positive signals faster. Expand to prospecting audiences only after the algorithm has enough conversion data to bid intelligently.

The overarching principle: treat new product launches as a separate discipline within your Google Ads strategy, not as products dropped into existing campaign structures. The campaign architecture that works for established products actively harms new ones.

Feed Optimisation for New SKUs

Feed quality matters even more for new products because they can't rely on performance history to overcome feed deficiencies:

• • Title precision: New product titles must be hyper-relevant to target search queries. Established products can rank for broad terms on historical performance; new products must win on relevance. Include size, colour, material, and brand - in that order.
• • Image quality: Use your best photography for new products. White background, high resolution, product fills 75%+ of the frame. New products without reviews rely entirely on visual appeal to generate clicks.
• • Product type granularity: Use the most specific product type classification possible. "Women's > Dresses > Midi Dresses > Cotton > Summer" performs better than "Women's > Dresses" because it helps Google match the product to specific search queries - critical when the algorithm has no conversion data to guide matching.
• • Custom labels for launches: Apply a custom label to all new products (e.g., custom_label_4 = "new_launch"). This enables campaign-level segmentation and bid adjustments without restructuring your entire feed.

Think of feed optimisation as compensating for the data gap. What established products achieve through performance history, new products must achieve through superior feed quality.

Measuring Launch Success

New products should be measured against launch-specific KPIs, not your standard account benchmarks:

• • Weeks 1-4: Measure impression share and CTR, not ROAS. The goal is visibility and click data. A new product achieving 30%+ impression share and 2%+ CTR in its target queries is on track.
• • Weeks 5-8: Measure conversion rate and CPA. Compare against category averages, not top-performing established products. A new product within 40% of category CPA is acceptable during learning.
• • Weeks 9-12: Transition to ROAS/POAS measurement. The product should be approaching (within 20%) your standard targets. If it's not, investigate feed quality, pricing, or product-market fit - not bidding.
• • Week 13+: Apply standard performance criteria. By this point, Smart Bidding has enough data to optimise effectively, and the product should be held to the same standards as established SKUs.

The most important metric across all phases is data accumulation rate. If the product isn't accumulating clicks and conversions fast enough, no bid strategy will fix the underlying visibility problem. Increase budgets, broaden targeting, or improve feed quality to accelerate the data flywheel.

Next Steps