Skip to main content

AI CSAT Score

Understanding how your AI CSAT Score is calculated and what it means

J
Written by James Caples
Updated over a week ago

Kaizan’s AI analyzes your client communications weekly to deliver a precise Client Satisfaction (CSAT) score. Here's exactly how we transform emails, calls, and chats with all client stakeholders into actionable insights about your business relationships.

Your CSAT score is a leading indicator — it changes before contracts get renewed or cancelled. A drop from 7.8 to 6.9 means you have time to course-correct. A jump from 7.1 to 8.3 signals opportunity to expand.

By analyzing what clients actually say (not what we think they mean), and then combined with quantitative engagement stats and risk phrase analysis, AI CSAT gives you the clearest possible view of relationship health — updated weekly, attributed accurately, and focused entirely on what drives your business forward.

Two-Layer Analysis: Context + Details

Layer 1: Relationship Temperature (70% of score)

We classify the overall health of each client relationship based on the narrative in your communications:

Band

Score Range

What We Detect

Promoter

8.6–10

"Exceeded all KPIs" • Case study requests • Budget increases without pushback

Advocate

7.5–8.5

Clear ROI evidence • Unsolicited referrals • "Love working with you"

Satisfied

6.6–7.4

KPIs on track • Professional tone • Routine renewal and next project discussions

Cautious

5.6–6.5

"Concerned about timeline" • ROI justification requests • Mixed feedback

At Risk

4.0–5.5

"Underperforming campaign" • "Let's pause and reassess" • Budget scrutiny • Breakdown and misalignment in expectations

Critical

0–3.9

Contract termination mentions • Payment issues • Legal escalations

Layer 2: Moment-by-Moment Analysis (30% of score)

Every business comment gets tagged and weighted by impact:

  • 2× Weight — Campaign/project results, deliverable quality

  • 1.5× Weight — Meeting deadlines, solving problems proactively

  • 1.3× Weight — Budget efficiency, ROI discussions

  • 1× Weight — Strategic advice, communication style

How Layers Combine: A Real Example

Starting Point: Your communications show "satisfied" patterns → Baseline: 7.0

Adjustments from specific moments:

  • ✓ "Campaign exceeded targets by 30%" (+0.3)

  • ✓ "Quick turnaround saved our launch" (+0.2)

  • ✗ "Timeline slipped on the Asia report" (-0.1)

  • ✗ "Need more strategic thinking" (-0.1)

Final Score: 7.3 (Baseline 7.0 + Net adjustment +0.3)

The Commercial Impact Multiplier

When clients mention specific business consequences, those comments get double weight in the adjustment calculation:

  • "Pause the £200k media spend" → -0.4 instead of -0.2

  • "Board approved additional budget" → +0.4 instead of +0.2

These high-stakes moments reveal true satisfaction better than routine feedback.

What Makes Scores Move

Signals That Push Scores Up:

  • "Results beat forecast by 40%"

  • "Can you take on our European markets too?"

  • "The CEO loved your presentation"

  • Faster response times from clients

  • Including you in strategic planning

Signals That Pull Scores Down:

  • "This isn't what we discussed"

  • “ROI performance isn’t great”

  • "We need to explore other options"

  • Delayed responses to your messages

  • Semantics of "Disappointed" appearing multiple times

  • Questions about contract terms

Reading Your Weekly Report

1. Your Score (0.0–10.0)

Precise to one decimal place — a shift from 7.2 to 7.5 means something changed.

2. Evidence-Based Insights

  • What Went Well: Direct quotes showing satisfaction

  • Areas to Improve: Specific concerns raised

3. Stakeholder Attribution

When possible, we identify who said what:

  • "The CMO mentioned budget concerns..."

  • "Your day-to-day contact praised the quick turnaround..."

The Integrity Layer

What Kaizan Promises:

  • Only Real Quotes: No interpolation or guessing

  • Business Focus: Personal chat filtered out automatically

  • Transparent Logic: Every score component traceable to actual communications

What We Don't Do:

  • Analyze work quality directly (we analyze how clients talk about it, you)

  • Make assumptions during quiet periods

  • Include internal team communications

Did this answer your question?