The Warmup Period

List view
Understanding Nobi
Getting Started
Implementing Nobi
Knowledge Base
Merchandising
Reporting
Custom Actions
Query Overrides
Plans And Billing
Developers Guide
Beta Products
References

The Warmup Period


When you first install Nobi, Nobi starts learning about your data and what converts. Search results over the first few hours won’t be perfect, but this is normal. Nobi's search quality improves significantly during the first 1-2 days as the system learns from real customer interactions. Here's what to expect and why it happens.

What Is the Warmup Period?

The warmup period is the initial day or two after your product catalog is first indexed. During this time, Nobi warms up based on your customers’ searches and behaviors.
What you might notice:
  • Some searches return less-than-ideal results
  • Product rankings may not match your expectations
  • Results feel "generic" rather than tailored to your store
What happens next:
  • After a few hundred searches, results improve noticeably
  • By 24 hours, search quality is typically much better
  • Results continue to improve over time as more data is collected
This is completely normal and expected. Don't panic if day-one results aren't perfect.

Why Does This Happen?

Caching Builds Over Time

Nobi caches the results of frequently searched queries. This means:
On day one:
  • Every search is computed fresh
  • No popular search patterns are established yet
After warmup:
  • Common searches are cached and optimized
  • Popular queries return refined results
  • Response times improve for frequent searches

Conversion Data Improves Rankings

Nobi tracks which products customers actually buy after searching. This conversion data directly improves results:
Initially:
  • All products are ranked by relevance alone
  • No purchase behavior data exists yet
  • Rankings are based purely on product data
After warmup:
  • Products that convert well get ranked higher
  • Results that lead to purchases are prioritized
  • Rankings reflect real customer preferences
This means search results become increasingly aligned with what actually sells, not just what seems relevant on paper.

What to Expect: Timeline

Hours 0-12: Building Up Data

  • Search works but results may not be fully optimized
  • Caches are being built for common queries
  • Some queries may return unexpected products
  • This is normal—don't make configuration changes yet

Hours 12-24: Quality Stabilizes

  • Search quality is typically excellent
  • Most common queries are cached
  • Early conversion signals are being incorporated

Day 2 and Beyond: Continuous Improvement

  • Search continues to improve gradually
  • More conversion data refines rankings further
  • Popular products that sell well rise in rankings

What You Should Do

During the Warmup Period

Do:
  • Let real customers use the search naturally
  • Give it at least 24 hours before evaluating
  • Note any specific problem queries to check later
Don't:
  • Make hasty configuration changes based on day-one results
  • Assume something is broken
  • Manually test dozens of queries and worry about each one

After the Warmup Period

If results still seem off after 24+ hours:
  1. Test the specific queries that concern you
  1. Check that your product data is complete and accurate
  1. Review your knowledge base for relevant content
  1. Contact support if issues persist
Most "search quality" concerns resolve themselves after the warmup period.

Factors That Speed Up Warmup

The warmup period depends on search volume. More searches = faster cache building and more conversion data.
Faster warmup (high traffic):
  • Stores with thousands of daily visitors
  • Product launches or promotional events
  • Active marketing driving traffic to search
Slower warmup (lower traffic):
  • New stores with limited traffic
  • Niche products with fewer searches
  • Recently launched sites
If you have lower traffic, the warmup period may extend slightly beyond 24 hours. This is fine—quality will still improve, just more gradually.

Tips for Best Results

Ensure Good Product Data

Nobi's search quality depends on your product data. Better data = better results, even during warmup.
Helpful product attributes:
  • Descriptive titles
  • Complete descriptions
  • Accurate categories and tags
  • Size, color, and variant information

Index Your Knowledge Base

If you have product guides, size charts, or FAQ content, add it to your knowledge base. This gives Nobi additional context for understanding customer queries.

Be Patient with Edge Cases

Unusual or highly specific searches may take longer to optimize. Common searches improve first, followed by long-tail queries over time.

Common Questions

Should I delay launching Nobi until warmup is complete?

No. The warmup period requires real customer interactions. Synthetic testing doesn't help—you need actual searches. Launch Nobi and let the warmup happen naturally.

Can I speed up the warmup period?

Not artificially. The warmup depends on genuine customer search behavior to build caches and collect conversion data. More real traffic means faster warmup, but fake searches don't help.

What if results are still bad after 24 hours?

If specific searches are still problematic after 24 hours:
  1. Check that relevant products exist in your catalog
  1. Verify product titles and descriptions are accurate
  1. Contact support with specific examples
Most issues resolve naturally, but some may indicate data quality problems.

Does warmup happen again if I update my catalog?

Minor updates (adding products, changing prices) don't restart warmup. Your existing caches and conversion data remain intact. Major catalog overhauls may require caches to be rebuilt for new products, but existing data is preserved.

Will search quality ever decrease?

Search quality generally only improves over time as more conversion data is collected. If you significantly change your catalog (removing popular products, major restructuring), caches for those products will need to be rebuilt.

Summary

Timeframe
What's Happening
What You'll See
0-12 hours
Building caches, collecting data
Results work but may not be fully optimized
12-24 hours
Caches established, conversion data incorporated
Excellent results
Day 2+
Continuous refinement
Results keep improving
Key takeaway: Don't evaluate search quality on day one. Give Nobi 24 hours of real customer activity, and you'll see the results you expect.

Getting Help

If you have concerns about search quality after the warmup period:
Include:
  • Specific search queries that concern you
  • What results you expected vs. what you saw
  • How long Nobi has been installed
We're happy to investigate and help optimize your search experience.