Testing on a new platform without historical data: how to do it right in 2026

Testing on a new advertising platform without historical data is one of the most challenging situations in online marketing. When no previous campaign data is available, the algorithm of Google Ads or Meta Ads has no foundation to optimise bids, refine audiences, or rank ad variations. The key to success lies in a structured launch methodology that makes the learning phase as short and cost-efficient as possible.

Key takeaways

• Without historical data, every new campaign enters a learning phase: the algorithm actively collects signals before Smart Bidding can optimise reliably.
• Correctly configured conversion tracking is the absolute foundation; without measurable conversions, the algorithm learns nothing.
• Start broad (broad match, wide audiences) and refine gradually, rather than applying overly tight targeting from the start.
• Sufficient budget during the launch phase significantly accelerates the learning phase; too low a daily budget needlessly prolongs it.
• AdBrains AI speeds up the launch trajectory via the Keyword Incubator, automated search term mining, and server-side signal enrichment, giving the algorithm faster access to usable signals.

What is the learning phase and why does it matter?

The learning phase is the period immediately after launching a campaign during which the Google Ads or Meta Ads algorithm actively collects data to calibrate its bids and targeting. Without historical data, the system has no reference point: it does not yet know which keywords convert, which audience segments respond, and what CPA or ROAS is realistic for your specific account.

According to Google Ads Help (2026), the learning phase typically lasts seven days after a campaign or bidding strategy has been significantly changed. In practice at AdBrains, we see that for brand-new accounts without any historical context, this can take longer, especially when the daily budget is too low or conversion tracking is not set up correctly. The learning phase is visible in the Google Ads dashboard via the "Learning" status next to the bidding strategy.

Ignoring the learning phase is a common mistake. Many advertisers adjust bids or settings while the campaign is still learning, causing the learning phase to restart every time. This results in a vicious cycle of suboptimal performance and constantly disrupted learning processes.

The right foundation: conversion tracking as your starting point

Correctly configured conversion tracking is the absolute prerequisite for success on a new platform. Without reliable conversion measurement, the algorithm places its bids based on click behaviour rather than actual value creation. This leads structurally to incorrect optimisations.

For e-commerce clients like ToetsJeKennis.nl or E-4motion.com, this means that purchases, cart additions, and product page visits must be set up as separate conversion goals, with the correct conversion values per action. For lead generation clients like Clima-Active.nl, it is equally important to measure form submissions, phone calls, and chat interactions as conversions, so the algorithm has a complete picture of the customer journey.

• Set up at least one primary conversion action (purchase or lead form)
• Add secondary conversions as an observation goal (not as a primary goal)
• Activate Enhanced Conversions for better matching of first-party data
• Consider server-side tagging via Google Tag Manager server-side for higher data quality
• Verify conversion tracking before campaign launch using Tag Assistant or Google Ads diagnostics
• Set realistic conversion values that reflect actual margins

Budget strategy in the launch phase

The daily budget directly influences how quickly the learning phase is completed, which is why choosing the right Google Ads starting budget deserves careful attention before launch. Google Ads advises (via Google Ads Help, 2026) setting the daily budget at a minimum of five to ten times the target CPA during the learning phase. This ensures sufficient click volume to draw statistically significant conclusions.

Example: suppose Clima-Active.nl uses a Target CPA of €40 per quote request. Then a daily budget of at least €200 is needed to complete the learning phase as quickly as possible. With a lower budget, the algorithm collects data too slowly, extending the learning phase from the standard seven days to sometimes several weeks.

For e-commerce clients like E-4motion.com, which sells new electric folding bikes with a higher average order value, the same principle applies. A tCPA or tROAS bidding strategy only works reliably once sufficient conversion data is available. In the launch phase, it is often wise to start with Maximize Conversions (without a Target CPA) and switch to Target CPA only once the recommended minimum of 30 to 50 conversions per month is reached, as described in Google Ads Help (2026).

Keyword strategy with zero historical data

The right keyword approach for a new account differs fundamentally from that of a mature account. Without historical performance data, you do not know which keywords convert and which only generate clicks without value. The solution: start broad and refine based on collected data.

Broad match combined with Smart Bidding is, according to Google Ads Help (2026), the most effective launch combination. Broad match gives the algorithm room to discover variations that would never have been thought of manually, while Smart Bidding simultaneously suppresses irrelevant impressions based on contextual signals.

Audience strategy: how to build audiences without historical data?

For audiences too, a new platform means starting with a blank slate. There are no remarketing lists, no lookalike audiences, and no conversion audiences to target. The strategy is to build usable audiences as quickly as possible while campaigns are running.

For Google Ads, this means creating remarketing lists from day one, even if they are initially too small to target. The lists grow organically with campaign traffic and can be used within weeks for RLSA (Remarketing Lists for Search Ads) or as Pmax audience signals to guide Performance Max. For Meta Ads, the same applies: a pixel or Conversions API must be active from the very start so that Custom Audiences fill up quickly.

In our practice at AdBrains, we see that advertisers who only activate audiences after campaigns have been running for weeks miss valuable remarketing data that cannot be reconstructed. Creating audiences is free and does not affect campaign performance, so there is no reason to delay this.

How AdBrains AI structurally accelerates the launch phase

The launch phase on a new platform is precisely the moment when AdBrains AI technology makes the greatest difference compared to manual management. While a manual manager adjusts settings weekly or even monthly, AdBrains AI systems work continuously in the background to accelerate the platform's learning process and minimise budget waste.

The most impactful system in the launch phase is the Keyword Incubator. New keywords are not placed directly into the production campaign, but first placed in a separate incubator campaign. Here they are safely tested for conversion potential. Only keywords that demonstrably contribute to conversions are promoted to the production campaign. This prevents unproven keywords from disrupting the Smart Bidding learning algorithm with irrelevant conversion signals.

In parallel, the automated search term mining system performs a daily analysis of all search terms generating clicks. Irrelevant search terms are automatically detected and added as negative keywords, without a human planner needing to check this daily. For a new campaign for HACCP-cursus.com, which offers online food safety courses, this system can already exclude dozens of irrelevant queries in the first week before they consume budget.

The server-side signal enrichment via AdBrains' own sGTM infrastructure plays a particularly significant role in the launch phase. When little data is available, every conversion signal carries more weight. By enriching conversions with first-party data and sending them more reliably to the Google Ads algorithm via server-side tracking, AdBrains gives the Smart Bidding engine more and better signals in less time. This approach is especially relevant for cookieless advertising, where first-party data increasingly determines signal quality. This noticeably accelerates the learning phase.

AdBrains' multi-agent verification system also ensures that no optimisation decision is executed without four independent AI agents having validated the decision. In the launch phase, this is especially valuable, because incorrect adjustments reset the learning phase and cause costly weeks of delay. The verification system catches these types of errors before they cause damage.

Finally, AdBrains' RSA improvement system analyses the Ad Strength scores of all ads in the launch phase. Ads with a POOR status are automatically rewritten with better headlines and descriptions, so that the Quality Score rises as quickly as possible. A higher Quality Score directly leads to lower CPC and better positions, which improves budget performance during the learning phase.

Common mistakes when launching without data

• Activating Target CPA or tROAS too early: without sufficient conversion volume, Smart Bidding underperforms compared to Maximize Conversions.
• Too tight keyword targeting at launch: exact match without data provides too little volume to complete the learning phase.
• Conversion tracking not tested before going live: a tag error keeps the learning phase from ever concluding.
• Adjusting bids or budgets during the learning phase: every significant change restarts the learning phase.
• No negative keywords at launch: leads to budget waste on irrelevant traffic.
• Not creating audiences at launch: remarketing data that is lost cannot be recovered.

Frequently asked questions

How long does the learning phase last on Google Ads for a new account?

The standard learning phase lasts seven days after a bidding strategy becomes active, according to Google Ads Help (2026). For brand-new accounts without any historical context, this can take longer, especially when the daily budget is low or conversion volume is limited. A new account in a niche market like HACCP-cursus.com may see less search volume in the first weeks than a broader product, extending the learning phase. Ensure sufficient budget and correct conversion tracking to complete the learning phase as quickly as possible.

Should I start with broad match or exact match for a new account?

For a new account without historical data, broad match combined with Smart Bidding is the recommended strategy, according to Google Ads Help (2026). Broad match gives the algorithm room to discover relevant search variations you would miss manually, while Smart Bidding suppresses irrelevant impressions. Exact match at the start limits volume too severely to complete the learning phase quickly. Do add a solid list of negative keywords to exclude irrelevant traffic.

When can I switch from Maximize Conversions to Target CPA or Target ROAS?

Google Ads advises (via Google Ads Help, 2026) having at least 30 conversions per month before switching to Target CPA, and at least 50 conversions per month for a reliable Target ROAS strategy. For lead generation clients like Clima-Active.nl or LeroyBrouwer.nl, which work with lower conversion volumes, this means the switch to tCPA sometimes only happens after six to eight weeks. Patience during this phase pays off: a Target CPA set too early leads to underperformance and higher cost per lead.

What is the biggest risk of testing on Meta Ads without historical data?

The biggest risk on Meta Ads without historical pixel data is that the algorithm stays in exploration mode too long, spending budget on a broad audience without a conversion focus. To prevent this, it is essential to set up the Meta Pixel or Conversions API correctly from day one and generate initial conversions as quickly as possible. Uploading an existing customer list as a Custom Audience helps Meta start with a direction for targeting, even if the pixel has not yet collected data.