Systems Atlas

Chapter 0.4

Search as a Business Lever

Why search is a profit center, not a cost center.


The Revenue Connection

43%

of e-commerce visitors go directly to search

2-3x

higher conversion for search users vs browsers

12%

revenue drop per 100ms of added latency

Case Study: The Zero Result Problem

Company

Mid-size e-commerce, 500K products

Problem

8% of searches return zero results

Analysis

  • • 8% zero result rate × 100K daily searches = 8,000 frustrated users/day
  • • If 20% of those would have converted at $50 AOV → $80K/day lost
  • • Annual impact: ~$29M

Fix

Added spell correction + synonym expansion

Result: Zero result rate dropped to 2%. Revenue up 6%.

The Competitive Moat

Search is hard to copy because of:

Data Flywheel

Better search → More users → More click data → Even better search

Behavioral Signals

Your click logs are proprietary. Competitors can't replicate.

Domain Knowledge

Synonyms, entity extraction, business rules encoded over years.

Talking to Leadership

❌ Wrong

"We need to upgrade Elasticsearch to version 8."

✓ Right

"Our zero-result rate is 8%, costing ~$80K/day. Upgrading enables spell correction, reducing this to 3%. ROI: 4-6 months."

Business Impact = (Query Volume) × (Conversion Rate Delta) × (AOV)

Prioritization Framework

ProblemImpactEffortPriority
Zero results on head queriesVery HighLowP0
Slow latency (P99 > 500ms)HighMediumP1
Bad ranking on long-tailMediumHighP2
PersonalizationMediumVery HighP3

Rule of thumb: Fix the head queries first (20% of queries = 80% of revenue). Then optimize the long tail for marginal gains.