[{"data":1,"prerenderedAt":968},["ShallowReactive",2],{"content-query-PSn0OuJoPK":3},{"_path":4,"_dir":5,"_draft":6,"_partial":6,"_locale":7,"title":8,"description":9,"heading":10,"prompt":11,"tags":15,"files":17,"nav":6,"presets":18,"gallery":36,"body":38,"_type":961,"_id":962,"_source":963,"_file":964,"_stem":965,"_extension":966,"sitemap":967},"/tools/lead-scoring-model","tools",false,"","Lead Scoring Model Calculator","Build lead scoring models online from Excel or CSV CRM data. Rank prospects, score conversion probability, and evaluate model quality with AI.","Lead Scoring Model",{"prefix":12,"label":13,"placeholder":14},"Build a lead scoring model","Describe your leads data and conversion outcome","e.g. CRM data with 1,500 leads, binary outcome converted (0/1); features include job title, company size, website visits, email opens, demo requested; logistic regression lead score 0–100; feature importance; threshold for hot leads; AUC and precision",[16],"statistics",true,[19,25,30],{"label":20,"prompt":21,"dataset_url":22,"dataset_title":23,"dataset_citation":24},"B2B SaaS Lead Scoring","Build a lead scoring model: binary outcome = converted (0/1); features = company_size, job_title, page_visits, email_opens, demo_requested, trial_signup; logistic regression; output lead score 0–100 for each lead; feature importance bar chart; ROC curve and AUC; precision-recall at top 20% of leads; threshold recommendation","https://data.cdc.gov/api/views/iuq5-y9ct/rows.csv?accessType=DOWNLOAD","NHANES Mental Health Assessment","CDC",{"label":26,"prompt":27,"dataset_url":28,"dataset_title":29,"dataset_citation":24},"Health Program Enrollment Scoring","Lead scoring for health program enrollment: binary outcome = enrolled (0/1); features = age, income, prior_contact, referral_source, screening_completed; logistic regression; score each person 0–100; feature importance; identify top 25% high-priority candidates; confusion matrix at 0.5 threshold","https://data.cdc.gov/api/views/dppn-5tm3/rows.csv?accessType=DOWNLOAD","NCHS Health and Nutrition Examination Survey",{"label":31,"prompt":32,"dataset_url":33,"dataset_title":34,"dataset_citation":35},"Student Conversion Scoring","Score prospective students by enrollment likelihood: outcome = enrolled (0/1); features = inquiry_source, campus_visit, GPA, financial_aid_inquiry, application_started; logistic regression lead score; rank prospects; feature importance; precision at top decile; score distribution plot by outcome","https://ourworldindata.org/grapher/pisa-test-score-mean-performance-on-the-reading-scale.csv","PISA Mean Reading Performance Scores","Our World in Data",[37],"/img/tools/lead-scoring-model.png",{"type":39,"children":40,"toc":950},"root",[41,50,71,97,102,108,166,172,432,437,443,592,598,756,762,830,836,881,887,909,933],{"type":42,"tag":43,"props":44,"children":46},"element","h2",{"id":45},"what-is-a-lead-scoring-model",[47],{"type":48,"value":49},"text","What Is a Lead Scoring Model?",{"type":42,"tag":51,"props":52,"children":53},"p",{},[54,56,62,64,69],{"type":48,"value":55},"A ",{"type":42,"tag":57,"props":58,"children":59},"strong",{},[60],{"type":48,"value":61},"lead scoring model",{"type":48,"value":63}," assigns each prospect a numeric score — typically 0 to 100 — representing their estimated probability of converting to a customer, patient, student, or program participant. Rather than treating all leads equally or relying on intuition, a data-driven scoring model learns from historical conversions: it identifies which observable characteristics (job title, company size, behavioral signals like demo requests or page visits) are most predictive of conversion and combines them into a single actionable score. Sales and marketing teams use these scores to ",{"type":42,"tag":57,"props":65,"children":66},{},[67],{"type":48,"value":68},"prioritize outreach",{"type":48,"value":70}," — focusing effort on high-score \"hot leads\" where conversion probability is highest, and deprioritizing low-score leads that are unlikely to convert regardless of effort invested.",{"type":42,"tag":51,"props":72,"children":73},{},[74,76,81,83,88,90,95],{"type":48,"value":75},"The statistical foundation is ",{"type":42,"tag":57,"props":77,"children":78},{},[79],{"type":48,"value":80},"logistic regression",{"type":48,"value":82}," (or a gradient-boosted classifier for non-linear patterns), which models the log-odds of conversion as a linear combination of input features. Each feature receives a ",{"type":42,"tag":57,"props":84,"children":85},{},[86],{"type":48,"value":87},"coefficient",{"type":48,"value":89}," — a positive coefficient means more of that feature increases conversion probability, a negative coefficient means it decreases it. The raw predicted probability (0 to 1) is multiplied by 100 to produce the 0–100 lead score. ",{"type":42,"tag":57,"props":91,"children":92},{},[93],{"type":48,"value":94},"Feature importance",{"type":48,"value":96}," (the magnitude of each coefficient) reveals which signals matter most: in a B2B context, a demo request might have a coefficient of +1.8 (strongly predictive), while website visits might add only +0.05 per visit. Understanding feature importance is often as valuable as the scores themselves — it tells the marketing team which engagement signals to invest in creating.",{"type":42,"tag":51,"props":98,"children":99},{},[100],{"type":48,"value":101},"A concrete example: a B2B SaaS company builds a scoring model on 1,600 historical leads (320 converted, 1,280 did not). The model achieves AUC = 0.83, meaning it correctly ranks a randomly chosen converter above a non-converter 83% of the time. Setting a threshold at score ≥ 60, the top 20% of leads capture 62% of all conversions — the sales team can focus on 320 \"hot leads\" and close 3× as many deals per call as they would working the full list. The feature importance chart shows that \"Demo Requested\" (+1.82) and \"Trial Sign-up\" (+1.63) are the strongest predictors, suggesting that the highest-ROI marketing investment is driving prospects to request demos or start trials.",{"type":42,"tag":43,"props":103,"children":105},{"id":104},"how-it-works",[106],{"type":48,"value":107},"How It Works",{"type":42,"tag":109,"props":110,"children":111},"ol",{},[112,123,139],{"type":42,"tag":113,"props":114,"children":115},"li",{},[116,121],{"type":42,"tag":57,"props":117,"children":118},{},[119],{"type":48,"value":120},"Upload your data",{"type":48,"value":122}," — provide a CSV or Excel file with one row per lead, a binary outcome column (1 = converted, 0 = did not convert), and feature columns representing lead characteristics and behaviors.",{"type":42,"tag":113,"props":124,"children":125},{},[126,131,133],{"type":42,"tag":57,"props":127,"children":128},{},[129],{"type":48,"value":130},"Describe the model",{"type":48,"value":132}," — e.g. ",{"type":42,"tag":134,"props":135,"children":136},"em",{},[137],{"type":48,"value":138},"\"outcome = converted, features = company_size, job_title, page_visits, email_opens, demo_requested; logistic regression; score 0–100; feature importance; AUC; precision at top 20%\"",{"type":42,"tag":113,"props":140,"children":141},{},[142,147,149,156,158,164],{"type":42,"tag":57,"props":143,"children":144},{},[145],{"type":48,"value":146},"Get full results",{"type":48,"value":148}," — the AI writes Python code using ",{"type":42,"tag":150,"props":151,"children":153},"a",{"href":152},"https://scikit-learn.org/",[154],{"type":48,"value":155},"scikit-learn",{"type":48,"value":157}," and ",{"type":42,"tag":150,"props":159,"children":161},{"href":160},"https://plotly.com/python/",[162],{"type":48,"value":163},"Plotly",{"type":48,"value":165}," to train the model, compute scores for all leads, plot feature importance, produce the ROC curve, and output a ranked lead list",{"type":42,"tag":43,"props":167,"children":169},{"id":168},"required-data-format",[170],{"type":48,"value":171},"Required Data Format",{"type":42,"tag":173,"props":174,"children":175},"table",{},[176,200],{"type":42,"tag":177,"props":178,"children":179},"thead",{},[180],{"type":42,"tag":181,"props":182,"children":183},"tr",{},[184,190,195],{"type":42,"tag":185,"props":186,"children":187},"th",{},[188],{"type":48,"value":189},"Column",{"type":42,"tag":185,"props":191,"children":192},{},[193],{"type":48,"value":194},"Description",{"type":42,"tag":185,"props":196,"children":197},{},[198],{"type":48,"value":199},"Example",{"type":42,"tag":201,"props":202,"children":203},"tbody",{},[204,242,283,323,362,401],{"type":42,"tag":181,"props":205,"children":206},{},[207,218,223],{"type":42,"tag":208,"props":209,"children":210},"td",{},[211],{"type":42,"tag":212,"props":213,"children":215},"code",{"className":214},[],[216],{"type":48,"value":217},"converted",{"type":42,"tag":208,"props":219,"children":220},{},[221],{"type":48,"value":222},"Binary outcome",{"type":42,"tag":208,"props":224,"children":225},{},[226,232,234,240],{"type":42,"tag":212,"props":227,"children":229},{"className":228},[],[230],{"type":48,"value":231},"1",{"type":48,"value":233}," (converted) or ",{"type":42,"tag":212,"props":235,"children":237},{"className":236},[],[238],{"type":48,"value":239},"0",{"type":48,"value":241}," (did not)",{"type":42,"tag":181,"props":243,"children":244},{},[245,254,259],{"type":42,"tag":208,"props":246,"children":247},{},[248],{"type":42,"tag":212,"props":249,"children":251},{"className":250},[],[252],{"type":48,"value":253},"company_size",{"type":42,"tag":208,"props":255,"children":256},{},[257],{"type":48,"value":258},"Numeric or categorical feature",{"type":42,"tag":208,"props":260,"children":261},{},[262,268,270,276,277],{"type":42,"tag":212,"props":263,"children":265},{"className":264},[],[266],{"type":48,"value":267},"250",{"type":48,"value":269},", ",{"type":42,"tag":212,"props":271,"children":273},{"className":272},[],[274],{"type":48,"value":275},"enterprise",{"type":48,"value":269},{"type":42,"tag":212,"props":278,"children":280},{"className":279},[],[281],{"type":48,"value":282},"SMB",{"type":42,"tag":181,"props":284,"children":285},{},[286,295,300],{"type":42,"tag":208,"props":287,"children":288},{},[289],{"type":42,"tag":212,"props":290,"children":292},{"className":291},[],[293],{"type":48,"value":294},"job_title",{"type":42,"tag":208,"props":296,"children":297},{},[298],{"type":48,"value":299},"Lead characteristic",{"type":42,"tag":208,"props":301,"children":302},{},[303,309,310,316,317],{"type":42,"tag":212,"props":304,"children":306},{"className":305},[],[307],{"type":48,"value":308},"VP Sales",{"type":48,"value":269},{"type":42,"tag":212,"props":311,"children":313},{"className":312},[],[314],{"type":48,"value":315},"Engineer",{"type":48,"value":269},{"type":42,"tag":212,"props":318,"children":320},{"className":319},[],[321],{"type":48,"value":322},"CEO",{"type":42,"tag":181,"props":324,"children":325},{},[326,335,340],{"type":42,"tag":208,"props":327,"children":328},{},[329],{"type":42,"tag":212,"props":330,"children":332},{"className":331},[],[333],{"type":48,"value":334},"page_visits",{"type":42,"tag":208,"props":336,"children":337},{},[338],{"type":48,"value":339},"Behavioral signal",{"type":42,"tag":208,"props":341,"children":342},{},[343,349,350,356,357],{"type":42,"tag":212,"props":344,"children":346},{"className":345},[],[347],{"type":48,"value":348},"12",{"type":48,"value":269},{"type":42,"tag":212,"props":351,"children":353},{"className":352},[],[354],{"type":48,"value":355},"3",{"type":48,"value":269},{"type":42,"tag":212,"props":358,"children":360},{"className":359},[],[361],{"type":48,"value":239},{"type":42,"tag":181,"props":363,"children":364},{},[365,374,379],{"type":42,"tag":208,"props":366,"children":367},{},[368],{"type":42,"tag":212,"props":369,"children":371},{"className":370},[],[372],{"type":48,"value":373},"email_opens",{"type":42,"tag":208,"props":375,"children":376},{},[377],{"type":48,"value":378},"Engagement feature",{"type":42,"tag":208,"props":380,"children":381},{},[382,388,389,394,395],{"type":42,"tag":212,"props":383,"children":385},{"className":384},[],[386],{"type":48,"value":387},"5",{"type":48,"value":269},{"type":42,"tag":212,"props":390,"children":392},{"className":391},[],[393],{"type":48,"value":239},{"type":48,"value":269},{"type":42,"tag":212,"props":396,"children":398},{"className":397},[],[399],{"type":48,"value":400},"18",{"type":42,"tag":181,"props":402,"children":403},{},[404,413,418],{"type":42,"tag":208,"props":405,"children":406},{},[407],{"type":42,"tag":212,"props":408,"children":410},{"className":409},[],[411],{"type":48,"value":412},"demo_requested",{"type":42,"tag":208,"props":414,"children":415},{},[416],{"type":48,"value":417},"Binary flag",{"type":42,"tag":208,"props":419,"children":420},{},[421,426,427],{"type":42,"tag":212,"props":422,"children":424},{"className":423},[],[425],{"type":48,"value":231},{"type":48,"value":269},{"type":42,"tag":212,"props":428,"children":430},{"className":429},[],[431],{"type":48,"value":239},{"type":42,"tag":51,"props":433,"children":434},{},[435],{"type":48,"value":436},"Any column names work — list the features and outcome in your prompt. Categorical features (job title, industry) are automatically one-hot encoded. Missing values are imputed with column medians/modes. More rows (> 500 with both outcomes) produce more reliable models.",{"type":42,"tag":43,"props":438,"children":440},{"id":439},"interpreting-the-results",[441],{"type":48,"value":442},"Interpreting the Results",{"type":42,"tag":173,"props":444,"children":445},{},[446,462],{"type":42,"tag":177,"props":447,"children":448},{},[449],{"type":42,"tag":181,"props":450,"children":451},{},[452,457],{"type":42,"tag":185,"props":453,"children":454},{},[455],{"type":48,"value":456},"Output",{"type":42,"tag":185,"props":458,"children":459},{},[460],{"type":48,"value":461},"What it means",{"type":42,"tag":201,"props":463,"children":464},{},[465,481,496,512,528,544,560,576],{"type":42,"tag":181,"props":466,"children":467},{},[468,476],{"type":42,"tag":208,"props":469,"children":470},{},[471],{"type":42,"tag":57,"props":472,"children":473},{},[474],{"type":48,"value":475},"Lead score (0–100)",{"type":42,"tag":208,"props":477,"children":478},{},[479],{"type":48,"value":480},"Each lead's estimated conversion probability × 100 — higher = more likely to convert",{"type":42,"tag":181,"props":482,"children":483},{},[484,491],{"type":42,"tag":208,"props":485,"children":486},{},[487],{"type":42,"tag":57,"props":488,"children":489},{},[490],{"type":48,"value":94},{"type":42,"tag":208,"props":492,"children":493},{},[494],{"type":48,"value":495},"Log-odds coefficient for each feature — magnitude shows importance, sign shows direction",{"type":42,"tag":181,"props":497,"children":498},{},[499,507],{"type":42,"tag":208,"props":500,"children":501},{},[502],{"type":42,"tag":57,"props":503,"children":504},{},[505],{"type":48,"value":506},"AUC (ROC curve)",{"type":42,"tag":208,"props":508,"children":509},{},[510],{"type":48,"value":511},"Area under the ROC curve — 0.5 = no better than random, 1.0 = perfect ranking; 0.70–0.85 is typical for lead scoring",{"type":42,"tag":181,"props":513,"children":514},{},[515,523],{"type":42,"tag":208,"props":516,"children":517},{},[518],{"type":42,"tag":57,"props":519,"children":520},{},[521],{"type":48,"value":522},"Precision at top N%",{"type":42,"tag":208,"props":524,"children":525},{},[526],{"type":48,"value":527},"What fraction of the top-scoring N% leads actually converted — the most business-relevant metric",{"type":42,"tag":181,"props":529,"children":530},{},[531,539],{"type":42,"tag":208,"props":532,"children":533},{},[534],{"type":42,"tag":57,"props":535,"children":536},{},[537],{"type":48,"value":538},"Threshold recommendation",{"type":42,"tag":208,"props":540,"children":541},{},[542],{"type":48,"value":543},"Score cutoff that optimizes the precision/recall trade-off for a given sales capacity",{"type":42,"tag":181,"props":545,"children":546},{},[547,555],{"type":42,"tag":208,"props":548,"children":549},{},[550],{"type":42,"tag":57,"props":551,"children":552},{},[553],{"type":48,"value":554},"Score distribution",{"type":42,"tag":208,"props":556,"children":557},{},[558],{"type":48,"value":559},"Histogram of scores for converted vs non-converted leads — good separation means the model is predictive",{"type":42,"tag":181,"props":561,"children":562},{},[563,571],{"type":42,"tag":208,"props":564,"children":565},{},[566],{"type":42,"tag":57,"props":567,"children":568},{},[569],{"type":48,"value":570},"Ranked lead list",{"type":42,"tag":208,"props":572,"children":573},{},[574],{"type":48,"value":575},"All leads sorted by score — ready to export to CRM for prioritized outreach",{"type":42,"tag":181,"props":577,"children":578},{},[579,587],{"type":42,"tag":208,"props":580,"children":581},{},[582],{"type":42,"tag":57,"props":583,"children":584},{},[585],{"type":48,"value":586},"Lift curve",{"type":42,"tag":208,"props":588,"children":589},{},[590],{"type":48,"value":591},"Ratio of model precision to random baseline — e.g., \"top 20% of scored leads are 3× more likely to convert\"",{"type":42,"tag":43,"props":593,"children":595},{"id":594},"example-prompts",[596],{"type":48,"value":597},"Example Prompts",{"type":42,"tag":173,"props":599,"children":600},{},[601,617],{"type":42,"tag":177,"props":602,"children":603},{},[604],{"type":42,"tag":181,"props":605,"children":606},{},[607,612],{"type":42,"tag":185,"props":608,"children":609},{},[610],{"type":48,"value":611},"Scenario",{"type":42,"tag":185,"props":613,"children":614},{},[615],{"type":48,"value":616},"What to type",{"type":42,"tag":201,"props":618,"children":619},{},[620,637,654,671,688,705,722,739],{"type":42,"tag":181,"props":621,"children":622},{},[623,628],{"type":42,"tag":208,"props":624,"children":625},{},[626],{"type":48,"value":627},"Basic scoring",{"type":42,"tag":208,"props":629,"children":630},{},[631],{"type":42,"tag":212,"props":632,"children":634},{"className":633},[],[635],{"type":48,"value":636},"outcome = converted; features = company_size, visits, email_opens; logistic regression; score 0–100; feature importance; AUC",{"type":42,"tag":181,"props":638,"children":639},{},[640,645],{"type":42,"tag":208,"props":641,"children":642},{},[643],{"type":48,"value":644},"Threshold setting",{"type":42,"tag":208,"props":646,"children":647},{},[648],{"type":42,"tag":212,"props":649,"children":651},{"className":650},[],[652],{"type":48,"value":653},"at what score threshold does precision reach 50%? compute precision and recall at score = 40, 50, 60, 70, 80",{"type":42,"tag":181,"props":655,"children":656},{},[657,662],{"type":42,"tag":208,"props":658,"children":659},{},[660],{"type":48,"value":661},"Top-N analysis",{"type":42,"tag":208,"props":663,"children":664},{},[665],{"type":42,"tag":212,"props":666,"children":668},{"className":667},[],[669],{"type":48,"value":670},"what % of conversions are captured in the top 10%, 20%, 30% of leads by score? lift chart",{"type":42,"tag":181,"props":672,"children":673},{},[674,679],{"type":42,"tag":208,"props":675,"children":676},{},[677],{"type":48,"value":678},"Segment scores",{"type":42,"tag":208,"props":680,"children":681},{},[682],{"type":42,"tag":212,"props":683,"children":685},{"className":684},[],[686],{"type":48,"value":687},"compute average lead score by industry and job title; which segments score highest?",{"type":42,"tag":181,"props":689,"children":690},{},[691,696],{"type":42,"tag":208,"props":692,"children":693},{},[694],{"type":48,"value":695},"Model comparison",{"type":42,"tag":208,"props":697,"children":698},{},[699],{"type":42,"tag":212,"props":700,"children":702},{"className":701},[],[703],{"type":48,"value":704},"compare logistic regression vs random forest for lead scoring; AUC comparison; feature importance for both",{"type":42,"tag":181,"props":706,"children":707},{},[708,713],{"type":42,"tag":208,"props":709,"children":710},{},[711],{"type":48,"value":712},"Score calibration",{"type":42,"tag":208,"props":714,"children":715},{},[716],{"type":42,"tag":212,"props":717,"children":719},{"className":718},[],[720],{"type":48,"value":721},"plot calibration curve: do predicted probabilities match observed conversion rates? calibrate if necessary",{"type":42,"tag":181,"props":723,"children":724},{},[725,730],{"type":42,"tag":208,"props":726,"children":727},{},[728],{"type":48,"value":729},"Negative signals",{"type":42,"tag":208,"props":731,"children":732},{},[733],{"type":42,"tag":212,"props":734,"children":736},{"className":735},[],[737],{"type":48,"value":738},"which features reduce conversion probability? show negative coefficients; are there any surprising negative predictors?",{"type":42,"tag":181,"props":740,"children":741},{},[742,747],{"type":42,"tag":208,"props":743,"children":744},{},[745],{"type":48,"value":746},"Score over time",{"type":42,"tag":208,"props":748,"children":749},{},[750],{"type":42,"tag":212,"props":751,"children":753},{"className":752},[],[754],{"type":48,"value":755},"plot average lead score by lead acquisition month; is lead quality improving or declining?",{"type":42,"tag":43,"props":757,"children":759},{"id":758},"assumptions-to-check",[760],{"type":48,"value":761},"Assumptions to Check",{"type":42,"tag":763,"props":764,"children":765},"ul",{},[766,783,793,803,820],{"type":42,"tag":113,"props":767,"children":768},{},[769,774,776,781],{"type":42,"tag":57,"props":770,"children":771},{},[772],{"type":48,"value":773},"Representative training data",{"type":48,"value":775}," — the model learns from historical conversions; if past conversions were influenced by sales capacity constraints (sales only called certain leads, so other leads never had a chance to convert), the model learns biased conversion signals; this is called ",{"type":42,"tag":57,"props":777,"children":778},{},[779],{"type":48,"value":780},"survivorship bias",{"type":48,"value":782}," and leads to underscoring leads that were never contacted",{"type":42,"tag":113,"props":784,"children":785},{},[786,791],{"type":42,"tag":57,"props":787,"children":788},{},[789],{"type":48,"value":790},"Sufficient positive examples",{"type":48,"value":792}," — logistic regression requires a reasonable number of conversions in the training data; with fewer than 50 conversions, the model is likely overfit and unstable; aim for at least 100–200 positive cases; with very few conversions, consider regularization (ridge penalty) or simpler models",{"type":42,"tag":113,"props":794,"children":795},{},[796,801],{"type":42,"tag":57,"props":797,"children":798},{},[799],{"type":48,"value":800},"Feature relevance stability",{"type":48,"value":802}," — the model assumes that the features that predicted conversion historically will continue to predict it in the future; if the market or product changes significantly, retrain the model on recent data rather than relying on a model trained on historical behavior",{"type":42,"tag":113,"props":804,"children":805},{},[806,811,813,818],{"type":42,"tag":57,"props":807,"children":808},{},[809],{"type":48,"value":810},"No data leakage",{"type":48,"value":812}," — features must reflect information available ",{"type":42,"tag":134,"props":814,"children":815},{},[816],{"type":48,"value":817},"at the time of scoring",{"type":48,"value":819},", not information that is only known after the conversion decision (e.g., \"meeting held\" or \"proposal sent\" should only be used if they occur before the conversion event being predicted); leakage inflates AUC and produces scores that cannot be used in practice",{"type":42,"tag":113,"props":821,"children":822},{},[823,828],{"type":42,"tag":57,"props":824,"children":825},{},[826],{"type":48,"value":827},"Class imbalance",{"type":48,"value":829}," — if conversions are rare (\u003C 5% of leads), the model may achieve high accuracy by predicting \"no conversion\" for everyone; check precision and recall rather than accuracy; use class-weight balancing or oversampling techniques if imbalance is severe",{"type":42,"tag":43,"props":831,"children":833},{"id":832},"related-tools",[834],{"type":48,"value":835},"Related Tools",{"type":42,"tag":51,"props":837,"children":838},{},[839,841,847,849,855,857,863,865,871,873,879],{"type":48,"value":840},"Use the ",{"type":42,"tag":150,"props":842,"children":844},{"href":843},"/tools/logistic-regression",[845],{"type":48,"value":846},"Logistic Regression",{"type":48,"value":848}," calculator to build the underlying classification model and examine the full regression output with p-values and confidence intervals on each feature coefficient. Use the ",{"type":42,"tag":150,"props":850,"children":852},{"href":851},"/tools/roc-curve",[853],{"type":48,"value":854},"ROC Curve and AUC Calculator",{"type":48,"value":856}," to evaluate model discrimination performance and compare multiple scoring models — AUC is the primary metric for ranking-based lead scoring. Use the ",{"type":42,"tag":150,"props":858,"children":860},{"href":859},"/tools/confusion-matrix",[861],{"type":48,"value":862},"Confusion Matrix & Sensitivity Specificity Calculator",{"type":48,"value":864}," to evaluate precision, recall, and F1-score at the chosen score threshold — different thresholds optimize for different business objectives (high precision vs. high recall). Use the ",{"type":42,"tag":150,"props":866,"children":868},{"href":867},"/tools/ab-test-calculator",[869],{"type":48,"value":870},"A/B Test Calculator",{"type":48,"value":872}," to test whether routing high-scoring leads to a specialized sales team significantly improves conversion rates compared to standard routing. Use the ",{"type":42,"tag":150,"props":874,"children":876},{"href":875},"/tools/conversion-funnel",[877],{"type":48,"value":878},"Conversion Funnel Analysis",{"type":48,"value":880}," to analyze how scored leads flow through the sales pipeline stages — do high-scoring leads advance further in the funnel even when they ultimately don't convert?",{"type":42,"tag":43,"props":882,"children":884},{"id":883},"frequently-asked-questions",[885],{"type":48,"value":886},"Frequently Asked Questions",{"type":42,"tag":51,"props":888,"children":889},{},[890,895,900,902,907],{"type":42,"tag":57,"props":891,"children":892},{},[893],{"type":48,"value":894},"Should I use logistic regression or a machine learning model like random forest?",{"type":42,"tag":57,"props":896,"children":897},{},[898],{"type":48,"value":899},"Logistic regression",{"type":48,"value":901}," is the right choice when interpretability matters — you need to explain to sales why a lead scores high or low, and the feature importance coefficients are directly interpretable. It works well for up to 20–30 features and typically achieves AUC = 0.70–0.80 for standard lead data. ",{"type":42,"tag":57,"props":903,"children":904},{},[905],{"type":48,"value":906},"Random forest or gradient boosting (XGBoost)",{"type":48,"value":908}," can capture non-linear interactions and improve AUC to 0.82–0.88, but the model is harder to explain. The practical recommendation: start with logistic regression for a baseline and stakeholder buy-in; switch to gradient boosting if the AUC improvement is meaningful (> 3–4 points) and you can explain the model using SHAP values. For most CRM-based lead scoring with \u003C 20 features and \u003C 10,000 leads, logistic regression is sufficient.",{"type":42,"tag":51,"props":910,"children":911},{},[912,917,919,924,926,931],{"type":42,"tag":57,"props":913,"children":914},{},[915],{"type":48,"value":916},"What AUC should I aim for, and what does it mean in practice?",{"type":48,"value":918},"\nAUC measures how well the model ",{"type":42,"tag":134,"props":920,"children":921},{},[922],{"type":48,"value":923},"ranks",{"type":48,"value":925}," leads — the probability that a randomly chosen converter scores higher than a randomly chosen non-converter. AUC = 0.50 means the model is no better than random ordering; AUC = 0.70 means 70% of the time a true converter outranks a non-converter. Practical benchmarks: AUC \u003C 0.65 indicates weak signal (consider adding better features); 0.65–0.75 is acceptable for prioritization; 0.75–0.85 is good; > 0.85 is excellent but may indicate data leakage. However, AUC is a ranking metric — what matters most for operational use is ",{"type":42,"tag":57,"props":927,"children":928},{},[929],{"type":48,"value":930},"precision at the score threshold you'll actually use",{"type":48,"value":932},". If your sales team can call 200 leads per week, what fraction of those 200 will convert? That precision figure at the top-200 cutoff is more actionable than AUC.",{"type":42,"tag":51,"props":934,"children":935},{},[936,941,943,948],{"type":42,"tag":57,"props":937,"children":938},{},[939],{"type":48,"value":940},"How often should I retrain the scoring model?",{"type":48,"value":942},"\nRetrain whenever conversion patterns change significantly — new product launches, market shifts, or changes in the lead acquisition mix. A practical trigger: if the model's ",{"type":42,"tag":57,"props":944,"children":945},{},[946],{"type":48,"value":947},"precision at your working threshold",{"type":48,"value":949}," has dropped by 5+ percentage points compared to the initial validation, retrain on recent data. For most B2B SaaS companies with steady state, quarterly retraining is a reasonable cadence. Always hold out the most recent 3–6 months of data as a validation set rather than using random splits — time-based splitting better simulates real deployment conditions where the model is always predicting future leads based on past patterns.",{"title":7,"searchDepth":951,"depth":951,"links":952},2,[953,954,955,956,957,958,959,960],{"id":45,"depth":951,"text":49},{"id":104,"depth":951,"text":107},{"id":168,"depth":951,"text":171},{"id":439,"depth":951,"text":442},{"id":594,"depth":951,"text":597},{"id":758,"depth":951,"text":761},{"id":832,"depth":951,"text":835},{"id":883,"depth":951,"text":886},"markdown","content:tools:085.lead-scoring-model.md","content","tools/085.lead-scoring-model.md","tools/085.lead-scoring-model","md",{"loc":4},1775502475420]