Quick Index — Business & Product Metrics
Activation Rate · Active Users · Adoption Rate · ARR · Availability · AOV · ARPU · Bounce Rate · Burn Rate · CAC Payback Period · Change Failure Rate · CTR · Conversion Rate · CPA · CPC · CPM · CAC · Customer Churn Rate · CES · LTV/CLV · Customer Retention Rate · CSAT · DAU · DAU/MAU Ratio · Defect Density · Deployment Frequency · Engagement Rate · Expansion MRR · Feature Adoption Rate · Feature Usage Frequency · Gross Margin · GRR · Lead Time for Changes · LTV:CAC Ratio · Magic Number · Market Share · MTTD · MTTR · MAU · MRR · NPS · NRR · North Star Metric · Payback Period · PMF Score · Quick Ratio · Retention Cohort · ROAS · Revenue Churn · ROI · Rule of 40 · Runway · Session Duration · Session Frequency · SLA Compliance · TTFV · TTV · Trial-to-Paid Conversion · WAU/MAU Ratio · WAU · Win Rate
Quick Index — AI-Specific Metrics
AI Adoption Rate · AI Cost Savings · AI ROI · Automation Rate · BLEU Score · Context Relevance / RAG · Drift Rate · Explainability Score · F1 Score · False Negative Rate · False Positive Rate · GPU Utilization · Groundedness Score · Guardrail Trigger Rate · Hallucination Rate · Human Override Rate · Inference Cost · Latency · Log Loss · MAE · Model Accuracy · Perplexity · Precision · Prompt Injection Success Rate · Prompt Success Rate · Recall · RMSE · ROC-AUC · ROUGE Score · Throughput · TTFT · Token Cost · Toxicity Rate · Training Cost
Business & Product Metrics
| Metric | Definition | Formula | Example |
|---|---|---|---|
| Activation Rate | Users completing first key action | Activated Users ÷ Signups × 100 | 700/1000 = 70% |
| Active Users | Users performing meaningful activity | Count | 75,000 users |
| Adoption Rate | New users adopting product | New Users ÷ Target Users | 35% |
| Annual Recurring Revenue (ARR) | Yearly subscription revenue | MRR × 12 | ₹2.4 Crore |
| Availability | System uptime | Uptime ÷ Total Time × 100 | 99.95% |
| Average Order Value (AOV) (new) | Average value of a transaction | Revenue ÷ Number of Orders | ₹1,200 |
| Average Revenue Per User (ARPU) | Revenue per customer | Revenue ÷ Customers | ₹450/month |
| Bounce Rate | Visitors leaving immediately | Single Page Visits ÷ Total Visits × 100 | 40% |
| Burn Rate | Monthly cash spent | Cash Outflow per Month | ₹12 Lakhs/month |
| CAC Payback Period (new) | Months needed to recover CAC through revenue | CAC ÷ (ARPU × Gross Margin) | 14 months |
| Change Failure Rate | Failed deployments | Failed Deployments ÷ Total Deployments × 100 | 8% |
| Click Through Rate (CTR) | Percentage clicking an ad | Clicks ÷ Impressions × 100 | 4% |
| Conversion Rate | Visitors converting to customers | Conversions ÷ Visitors × 100 | 500/10,000 = 5% |
| Cost per Acquisition (CPA) | Marketing cost per conversion | Marketing Cost ÷ Conversions | ₹600 |
| Cost Per Click (CPC) | Cost for each click | Ad Spend ÷ Clicks | ₹12 |
| Cost Per Mille (CPM) | Cost per thousand impressions | Cost ÷ (Impressions/1000) | ₹180 |
| Customer Acquisition Cost (CAC) | Cost to acquire one customer | Sales & Marketing Cost ÷ New Customers | ₹10,00,000 ÷ 500 = ₹2,000 |
| Customer Churn Rate | Percentage of customers lost during a period | (Customers Lost ÷ Customers at Start) × 100 | Lost 50 of 1,000 customers = 5% |
| Customer Effort Score (CES) | Ease of completing a task | Survey Average | Average score 6.2/7 |
| Customer Lifetime Value (CLV/LTV) | Total revenue expected from a customer | Average Revenue × Gross Margin × Customer Lifetime | ₹2,000/month × 24 months = ₹48,000 |
| Customer Retention Rate | Percentage of customers retained | ((Ending Customers − New Customers) ÷ Starting Customers) × 100 | Started 1000, ended 1100, gained 200 → 90% |
| Customer Satisfaction (CSAT) | Satisfaction after interaction | Positive Responses ÷ Total Responses × 100 | 450/500 = 90% |
| Daily Active Users (DAU) | Unique users active daily | Count of daily users | 80,000 users |
| DAU/MAU Ratio | User engagement ("stickiness") | DAU ÷ MAU × 100 | 80K ÷ 200K = 40% |
| Defect Density | Bugs per software size | Bugs ÷ KLOC | 0.8 bugs/KLOC |
| Deployment Frequency | Production deployments | Deployments per Day/Week | 15/day |
| Engagement Rate | Meaningful interactions | Interactions ÷ Users | 8 actions/user/day |
| Expansion MRR (new) | Additional recurring revenue from existing customers (upsell/cross-sell) | Upsell MRR + Cross-sell MRR | ₹2 Lakhs/month |
| Feature Adoption Rate | Users using a feature | Feature Users ÷ Active Users × 100 | 4000/10000 = 40% |
| Feature Usage Frequency | Average feature usage | Total Feature Uses ÷ Users | 5 uses/user/week |
| Gross Margin | Profit after production costs | (Revenue − COGS)/Revenue × 100 | 72% |
| Gross Revenue Retention (GRR) | Revenue retained excluding expansion | (Starting MRR − Lost MRR)/Starting MRR | 90% |
| Lead Time for Changes | Time from commit to production | Deployment − Commit Time | 2 hours |
| LTV:CAC Ratio | Measures profitability of customer acquisition | LTV ÷ CAC | ₹48,000 ÷ ₹2,000 = 24:1 |
| Magic Number (new) | SaaS sales efficiency indicator | (Current Qtr ARR − Prior Qtr ARR) × 4 ÷ Prior Qtr S&M Spend | 0.8 |
| Market Share | Company's market portion | Company Sales ÷ Market Sales × 100 | 18% |
| Mean Time To Detect (MTTD) | Average detection time | Total Detection Time ÷ Incidents | 12 min |
| Mean Time To Repair (MTTR) | Average recovery time | Repair Time ÷ Incidents | 45 min |
| Monthly Active Users (MAU) | Unique users active monthly | Count of unique monthly users | 1.2 million users |
| Monthly Recurring Revenue (MRR) | Monthly subscription revenue | Sum of Monthly Subscriptions | ₹20 Lakhs |
| Net Promoter Score (NPS) | Customer loyalty score | %Promoters − %Detractors | 60% − 20% = 40 |
| Net Revenue Retention (NRR) | Revenue retained including expansion | (Starting MRR − Lost + Expansion)/Starting MRR | 118% |
| North Star Metric (new) | Single metric capturing the core value delivered to customers | Varies by product | e.g., "Weekly Active Trips" for a rides app |
| Payback Period | Time to recover investment | Investment ÷ Annual Savings | 2 years |
| Product-Market Fit Score | Customer willingness to miss product | Survey | 45% "Very disappointed" |
| Quick Ratio (SaaS) (new) | Growth efficiency: new + expansion revenue vs. losses | (New MRR + Expansion MRR) ÷ (Churned MRR + Contraction MRR) | 4:1 |
| Retention Cohort | User retention by signup cohort | Cohort Analysis | 60% retained after Month 3 |
| Return on Ad Spend (ROAS) | Revenue from ads | Revenue ÷ Ad Spend | 6× |
| Revenue Churn | Revenue lost from existing customers | Lost MRR ÷ Starting MRR × 100 | ₹50K/₹10L = 5% |
| ROI | Return on investment | (Gain − Cost)/Cost × 100 | 150% |
| Rule of 40 (new) | Combined growth + profitability health check for SaaS | Revenue Growth % + Profit Margin % | 25% + 18% = 43 (healthy) |
| Runway | Months before cash runs out | Cash Available ÷ Burn Rate | ₹2.4 Cr ÷ 12L = 20 months |
| Session Duration | Average time spent per session | Total Session Time ÷ Sessions | 12 minutes |
| Session Frequency | Average sessions per user | Sessions ÷ Users | 3 sessions/day |
| SLA Compliance | Service agreement adherence | Met SLAs ÷ Total SLAs | 98% |
| Time to First Value (TTFV) | Time until first meaningful outcome | First Success − Signup | 20 minutes |
| Time to Value (TTV) | Time until customer gains value | Activation Date − Signup Date | 3 days |
| Trial-to-Paid Conversion | Free trials becoming paid customers | Paid Users ÷ Trial Users × 100 | 250/1000 = 25% |
| WAU/MAU Ratio (new) | Weekly engagement stickiness | WAU ÷ MAU × 100 | 33% |
| Weekly Active Users (WAU) (new) | Unique users active in a week | Count of weekly users | 400,000 users |
| Win Rate (new) | Percentage of sales deals won | Deals Won ÷ Total Deals × 100 | 28% |
AI-Specific Metrics
| Metric | Definition | Formula | Example |
|---|---|---|---|
| AI Adoption Rate | Employees using AI tools | Active AI Users ÷ Employees | 74% |
| AI Cost Savings | Savings due to AI | Previous Cost − Current Cost | ₹3 Crore/year |
| AI ROI | Return on AI investment | (Benefit − AI Cost) ÷ AI Cost | 240% |
| Automation Rate | Tasks automated by AI | Automated Tasks ÷ Total Tasks | 65% |
| BLEU Score (new) | Measures generated text overlap with reference text (translation/generation quality) | N-gram precision score | 0.42 |
| Context Relevance / Retrieval Precision (RAG) (new) | Relevance of retrieved context in RAG pipelines | Relevant Chunks ÷ Retrieved Chunks × 100 | 88% |
| Drift Rate | Performance degradation over time | Current Accuracy − Baseline Accuracy | Accuracy dropped from 96% to 91% |
| Explainability Score (new) | How interpretable a model's decisions are to humans | Audit/survey-based rating | 7.5/10 |
| F1 Score | Harmonic mean of Precision & Recall | 2 × Precision × Recall ÷ (Precision + Recall) | 90% |
| False Negative Rate | Missed positive predictions | FN ÷ (FN+TP) | 6% |
| False Positive Rate | Incorrect positive predictions | FP ÷ (FP+TN) | 4% |
| GPU Utilization (new) | Compute efficiency during training/inference | Used GPU Capacity ÷ Available Capacity × 100 | 78% |
| Groundedness Score | Responses supported by source data | Supported Answers ÷ Total | 97% |
| Guardrail Trigger Rate (new) | How often safety guardrails intervene | Triggers ÷ Total Requests × 100 | 2% |
| Hallucination Rate | Incorrect AI-generated responses | Hallucinations ÷ Responses | 3% |
| Human Override Rate | AI decisions overridden by humans | Overrides ÷ AI Decisions | 4% |
| Inference Cost | Cost per prediction | Compute Cost ÷ Predictions | ₹0.002 |
| Latency | Time to return prediction | Average Response Time | 120 ms |
| Log Loss | Prediction error | Cross-Entropy Loss | 0.18 |
| MAE | Mean Absolute Error | Σ|Error| ÷ n | 1.8 |
| Model Accuracy | Percentage of correct predictions | Correct Predictions ÷ Total Predictions | 96% |
| Perplexity (new) | Measures how well a language model predicts text (lower is better) | 2^(cross-entropy loss) | 12.4 |
| Precision | Correct positive predictions | TP ÷ (TP+FP) | 92% |
| Prompt Injection Success Rate | Successful prompt injection attacks | Successful Attacks ÷ Attempts | 0.2% |
| Prompt Success Rate | Prompts achieving intended outcome | Successful Prompts ÷ Total | 86% |
| Recall | Correct detection of positives | TP ÷ (TP+FN) | 89% |
| RMSE | Prediction error for regression | √(Σ(Error²)/n) | 2.3 |
| ROC-AUC | Model discrimination ability | Area under ROC curve | 0.95 |
| ROUGE Score (new) | Measures overlap between generated and reference summaries | Overlap-based recall score | 0.51 |
| Throughput | Predictions processed per second | Requests ÷ Seconds | 500/sec |
| Time to First Token (TTFT) (new) | Latency before the first output token appears — key UX metric for LLMs | Time from request to first token | 300 ms |
| Token Cost | Cost per LLM request | Tokens × Cost/Token | ₹0.08/query |
| Toxicity Rate (new) | Share of harmful/unsafe AI outputs | Toxic Responses ÷ Total Responses × 100 | 0.5% |
| Training Cost (new) | Total cost to train or fine-tune a model | Compute + Data + Labor Cost | ₹40 Lakhs |
Core 30: Metrics Every AI Executive Should Know by Heart
(A–Z; unchanged — this is the original curated priority list. Each term links to its full definition above.)
Activation Rate · AI Adoption Rate · AI ROI · ARR · Automation Rate · CAC · Conversion Rate · CSAT · Customer Churn · Customer Retention · DAU · Deployment Frequency · Drift Rate · F1 Score · Gross Margin · Groundedness Score · Hallucination Rate · Latency · LTV · LTV:CAC Ratio · MAU · Model Accuracy · MRR · MTTR · NPS · Precision · Recall · ROI · SLA Compliance · Throughput
Worth considering for future revisions: North Star Metric, Rule of 40, TTFT, and WAU.
Great leaders don't memorize every metric, they know which metric to look at, when to look at it, and what action to take next.
Keep this guide bookmarked, revisit it often, and let it grow with your career. Because in the world of AI and business, you can't improve what you don't measure... and you definitely can't explain it in a meeting if you don't know what it means. 😉
Top comments (1)
Great reference! One thing I’d add is that metrics are most valuable when viewed as a system rather than in isolation. For example, improving AI accuracy while latency, token cost, or hallucination rate worsen may actually reduce overall product value. Likewise, a high AI ROI means little if user adoption stays low. The real challenge isn’t tracking more metrics—it’s understanding the trade-offs between them and knowing which ones align with the product’s North Star