Setting Limits and Goals
Last updated
English
Last updated
After entering the basic identification details of your new clusterization in the next key step is defining limits and goals. These settings determine the boundaries within which the Karsa algorithm will work, and help ensure that the resulting clusters will be not only statistically optimal but also practically usable for your Google Ads campaigns.
This setting allows you to determine the range of how many resulting clusters (which usually corresponds to the number of future campaigns or segments) you wish to create.
Minimum number of clusters/campaigns:
What it is: The lowest number of clusters into which your products will be divided.
How it works: Even if you set a low minimum, the Karsa Labelizer machine learning algorithm always tries to find the optimal number of clusters with respect to your data distribution, chosen parameters, and the requirement for high internal consistency of each cluster and meeting minimum conversion requirements.
Maximum number of clusters/campaigns:
What it is: The highest possible number of clusters that the algorithm should create.
Important: Even if you set a high maximum (e.g., 20 or more), such a number of clusters may not necessarily be created. The optimal number always depends on the actual distribution of your product values and the number of parameters chosen for clusterization. Machine learning looks for the number of clusters that maximizes the predictability of future campaigns.
Recommendations:
For smaller e-commerce stores or specific segments, a lower number of clusters may be more effective (e.g., 3-7).
For large e-commerce stores with diverse assortment and large volumes of data, it may be appropriate to experiment with a higher number (e.g., 10-20 or more), if the data distribution supports it.
This is one of the most important settings for ensuring the stability and effectiveness of your future campaigns.
Minimum number of conversions:
What it is: Defines the minimum required sum of conversions (or orders, see below) over the last 30 days for all products that will be assigned to one cluster.
Why it's important: Google Ads algorithms (especially Smart Bidding strategies like tROAS) need a sufficient volume of conversion data to effectively "learn" and optimize bids. Campaigns (clusters) with too few conversions:
May have unstable performance.
May lead to incorrect predictions and suboptimal spending.
May have trouble maintaining budget or achieving goals.
Recommendations for minimum number of conversions:
Google Ads generally recommends:
At least 15-20 conversions over the last 30 days for a campaign before you start using automated bidding strategies.
For stable Target ROAS (tROAS), it's ideal to have 30-50 conversions over 30 days.
For advanced prediction and optimization, 100+ conversions monthly is recommended.
Karsa Labelizer recommendation:
While Google recommends at least 50 orders (conversions) per campaign, for better consideration of seasonality and product availability, it's often better to aim for 60-70 conversions over 30 days within one cluster.
Our advice: Start with a value of at least 30-50. If you have very large volumes of data, you can experiment with higher values. If, on the contrary, you have little data, consider a lower value with awareness of possible risks, or focus on a smaller number of resulting clusters.
Parameter for the minimum number of orders:
What it is: You can choose whether the minimum limit should apply to:
Conversions: Value downloaded directly from your Google Ads account. This is the most common choice.
Start with a reasonable range (e.g., min 3, max 10) and based on the results (especially the Consistency metric in the report), you can adjust the limits for further tests.
Proper setting of these limits and goals is essential for creating meaningful, stable, and high-performing clusters. Always consider the specifics of your assortment, the volume of available data, and your business goals. In the next section, we will focus on the equally important .
