Recommendations#

Vector's core value proposition: turning raw analytics data into prioritized, actionable recommendations.

How It Works#

  1. Data Collection: Vector pulls metrics from your connected integrations
  2. Pattern Analysis: AI identifies anomalies, trends, and opportunities
  3. Impact Scoring: Each insight is scored on impact, confidence, and ease
  4. Ranking: Recommendations are sorted by overall priority score
  5. Specification: Generate LLM-ready specs for engineering handoff

Impact Scoring#

Every recommendation includes three scores on a 1-10 scale:

Impact Score#

What it measures: Potential business value if implemented

Factors:

  • Revenue impact (direct or indirect)
  • User experience improvement
  • Conversion rate potential
  • Traffic/engagement lift

Example: A checkout optimization affecting 45% of users with $12k/mo potential lift = Impact 8.5

Confidence Score#

What it measures: How certain Vector is about this recommendation

Factors:

  • Data volume (more data = higher confidence)
  • Statistical significance
  • Pattern consistency over time
  • Multiple corroborating signals

Example: Clear drop-off at a specific step with 10,000+ data points = Confidence 9

Ease Score#

What it measures: Implementation complexity (higher = easier)

Factors:

  • Technical complexity
  • Design changes required
  • Dependencies and risks
  • Estimated effort

Example: Adding social proof (copy + testimonials) = Ease 9 Example: Full checkout redesign = Ease 5

Priority Score#

The overall priority score combines all three:

Priority = (Impact × 0.4) + (Confidence × 0.3) + (Ease × 0.3)

This formula prioritizes high-impact, high-confidence, easy wins—the low-hanging fruit that delivers ROI fastest.

Recommendation Categories#

Recommendations are tagged by category:

| Tag | Focus Area | |-----|------------| | conversion | Funnel optimization, checkout, signups | | engagement | User retention, feature adoption | | ux | Navigation, usability, accessibility | | performance | Speed, load times, technical debt | | growth | Acquisition, virality, referrals |

Recommendation Lifecycle#

Pending#

New recommendations awaiting review. These appear in your main feed.

Actions:

  • Accept: Move to your implementation queue
  • Dismiss: Remove from active view (can be recovered)
  • Generate Spec: Create engineering handoff document

Accepted#

Recommendations you've decided to implement.

Track these through implementation and measure results.

Dismissed#

Recommendations you've decided not to pursue.

Reasons might include:

  • Already planned or in progress
  • Not aligned with current priorities
  • Technical constraints
  • Business reasons

Dismissed items can be reviewed and reconsidered later.

Generating New Recommendations#

Click Generate New to trigger a fresh analysis. This will:

  1. Pull latest data from all connected integrations
  2. Run AI analysis across all metrics
  3. Generate new insights based on recent patterns
  4. Add new recommendations to your feed

Note: Generation uses AI credits. We recommend running this weekly or after significant product changes.

Best Practices#

  1. Review weekly: Set a recurring time to review new recommendations
  2. Accept strategically: Don't accept everything—prioritize what aligns with current goals
  3. Track outcomes: After implementing, note the actual impact
  4. Provide feedback: Dismissing with reasons helps improve future recommendations