Since we introduced expectation ranges for all relevant search KPIs to detect and analyze significant short-term fluctuations, it became evident that we also needed a seasonal expectation range. This would incorporate long-term data to help distinguish temporary variations from genuine long-term trends.

Implementing this was a complex challenge, but we are excited to announce the launch of the Seasonal Expected Range view for your data!

We analyze historical data from previous years, remove extreme outliers, and smooth the data based on extracted seasonal effects. This smoothing process is performed using the Whittaker-Eilers method.

We then calculate confidence intervals and overlay them with the current year's data (the time frame of interest).

By doing so, we effectively filter out short-term outliers if they align with expected seasonal patterns. The following chart illustrates this approach using sample search CTR data.

Three years ago, we introduced our Query Interpretation Service, designed to detect true shopper intent by breaking queries down into distinct concepts.

Since then, we’ve been hard at work integrating a taxonomy classification mechanism capable of mapping shopper intent to our shopping taxonomy with over 95% precision.

After processing billions of queries and interactions, we’re thrilled to announce the launch of our Automatic Query Taxonomy Classifier, now available through the Keyword Interpretation view.

In the coming weeks and months, we’ll roll out several new features that leverage this powerful capability. For now, we’re excited to make it accessible to our users. Stay tuned!

In any e-commerce search system, analytics and insights are essential for driving continuous improvement. Historically, while SearchInsights has focused on analyzing search paths, we have placed too much emphasis on the final stage of the funnel—conversion. However, our data highlights the importance of focusing on the preceding step: add-to-basket.

Add-to-basket data provides a powerful signal of user intent, revealing which products resonate most with shoppers after being discovered through search. By analyzing this data, we can uncover behavioral patterns, prioritize high-performing products, and enhance the relevance of future search results. Furthermore, comparing add-to-basket rates with checkout rates offers valuable insights into whether issues in the final stage of the funnel require attention, even if they are not directly tied to search performance. This approach ensures a more holistic understanding of the customer journey and identifies opportunities for optimization at every stage.

Ranking also plays a pivotal role in determining the order in which products are displayed, balancing factors like relevance, popularity, and business objectives such as inventory clearance or promoting high-margin items. Significant effort is invested in improving ranking performance in search. However, many commonly used metrics for evaluating ranking quality in e-commerce search, such as MRR and NDCG, are heavily rooted in traditional web search use cases and may not fully capture the unique needs of e-commerce.

To address this gap, we set out to provide our customers with a clearer, user-focused perspective on their ranking quality—ultimately judged by their shoppers. This led to the introduction of EngagementRank. EngagementRank measures how far your current ranking deviates from the optimal ranking, while accounting for biases such as position and presentation effects. A score of 100% represents a perfect ranking, while lower values highlight areas for improvement. This metric ensures a more accurate and actionable evaluation of ranking performance tailored to the e-commerce context.

We are excited to announce an update to the data table CSV export functionality: category affinity raw data is now included! This enhancement provides valuable insights into category preferences, distributed as frequencies across various taxonomy levels. The specific taxonomy levels included in the export depend on your tenant configuration, ensuring the data aligns with your setup.

With this addition, you can now leverage detailed category affinity data to analyze user behavior and refine strategies for product categorization, personalization, and search optimization. This update makes it easier to extract actionable insights and improve decision-making based on user preferences.

Our performance charts have become the go-to tool for search managers to identify micro- and macro-trends and benchmark search performance effectively. However, search is complex. Internal changes, external factors, deployments, and unexpected hiccups can all impact the charts, often leading to spikes, dips, or other notable patterns.

To help provide clarity, we’ve introduced annotations to our performance charts. This new functionality allows you to add additional context—either systematically or manually—to document what might have caused these changes.

There are two types of annotations:

With annotations, you can now better understand and explain the dynamics behind your performance data.

We received feedback suggesting that it would be helpful to display average values and trend indicators for specified time period in the "Performance Charts" section. In response, we have updated the layout of all graphs to incorporate these features.

With the recent addition of deeper and more complex comparison modes for your data, we have also updated the Compare Breadcrumb to better reflect the various available modalities and make their usage even more intuitive.

We have also enhanced the Status Message Menu to provide you with a more comprehensive overview of recent updates and events.

Intent-based redirects are often one of the most effective ways to enhance the search experience, particularly for short- to mid-tail searches. However, due to the frequent changes made by various teams across a webshop, some updates may go unnoticed, which can cause existing redirects to fail. To help address this, search managers need timely notifications about these "broken redirects" so they can resolve issues promptly.

Starting now, searchHub will regularly check all active redirects and highlight any issues directly within the redirects view. This makes it quick and efficient for search managers to monitor and manage the status of their active searchHub redirects.

We currently cover four cases, each represented by a traffic light icon:

: The only reliable way to verify if a URL is working is by requesting it through crawling. We have developed a highly reliable method to make this work, to facilitate this process for both you and us, please ensure that our bot/crawler is whitelisted within your environment.

Enhancements to Query Suggestions for a Better Search Experience

Suggestions are essential for enhancing user search experiences, and our data reveals that for customers using our smartSuggest feature, nearly 60% of search queries are influenced by these suggestions.

We identified two key areas where our query suggestions could be further refined:

Previously, like many auto-suggest solutions, ours was limited in leveraging word segmentation—an essential feature for languages such as Norwegian, Swedish, Danish, German, and Dutch.

This limitation sometimes prevented users from feeling fully understood during query formulation. For instance, a user typing “tisch” (meaning “table” in German) might not receive an ideal range of prefix and suffix suggestions, such as "tischdecke," "gartentisch," and "klapptisch." Our improved implementation now fully addresses this, delivering more accurate and engaging suggestions.

In 2023, we introduced inspirational AI-Suggestions (https://searchhubio.canny.io/changelog/smartsuggest-new-ai-suggestions-type) that aim to inspire users before they even start typing by identifying and amplifying micro-trends in search behavior.

Through extensive data analysis, we found that these suggestions indeed impact user searches, but we saw potential to enhance them further. The updated implementation now automatically filters underperforming queries, giving higher-impact queries a greater opportunity to drive traffic.

From the start, it has been crucial for SearchHub to place search KPIs in a broader context to provide deeper insights into the trajectory of your search quality. We often received questions about whether a particular KPI was within a "good" range, but our response was typically, "it depends." Now, with enough customers and data, we can finally offer qualitative feedback in the form of an Industry Benchmark.

As always, data on its own is just data. However, when viewed in a broader context, it can lead to valuable insights—and even actionable steps. That's why we're excited to introduce the new Industry Benchmark comparison, which provides visual insights into how your KPIs measure up against peers and market trends. We firmly believe that comparing your eCommerce search KPIs with industry benchmarks is essential for objectively evaluating performance and for setting realistic performance goals.

Following our recent roll-out of confidence intervals, we decided to maintain the same method of calculating them across a large, diverse sample of customers representing various domains. Additionally, we've introduced gradient color coding, ranging from red to green, to enhance the visual clarity of the results.

We are excited to announce that following our recent technology migration, we are now launching a new initiative to provide deeper insights into your search data than ever before.

KPIs naturally fluctuate over time, but the key question is whether these changes are simply random noise or indicate a systematic improvement or issue. That’s where the new Confidence Intervals come in. By factoring in historical data, they predict the expected upper and lower bounds for your KPIs.

And this is just the beginning. Over the next few weeks, we’ll be rolling out several exciting and value-adding improvements!