Upload a CSV, pick your count column and drivers, and get a real count regression — automatic overdispersion testing, the right model chosen for you, and every effect as a plain-language rate ratio. Free.
Free analyses run on up to 10,000 rows. Larger files are randomly sampled to that size — sign up to analyze your full dataset.
Fitting count models and testing overdispersion...
Sent to — rate ratios with confidence intervals, driver ranking, overdispersion verdict, count distribution, R code, and AI insights.
Analyze another fileThe analysis fits a Poisson regression of your count on the mapped drivers, then computes the Pearson dispersion statistic (chi-squared divided by residual degrees of freedom). If dispersion exceeds 1.5 — the signature of overdispersed real-world counts — it refits a negative binomial model (falling back to quasi-Poisson if needed) and makes that the headline. Every coefficient is exponentiated into a rate ratio with a 95% confidence interval: the multiplicative change in the expected count per unit of each driver.
Use it whenever the thing you want to explain is a count of events — orders, tickets, defects, visits, clicks per row — and you want driver effects that respect the count's non-negative, integer, skewed nature.
Not for continuous outcomes like revenue or duration (use linear regression), not for binary outcomes (use classification drivers), and not for counts with wildly different exposure windows per row.
Built for: Analysts and operators explaining event volumes — orders, tickets, defects, visits
Typical data source: A spreadsheet or CSV with one events-per-row count column and several candidate driver columns
One count column plus candidate drivers. For example, weekly orders per store:
Minimum 30 rows · Best with 200-20,000 rows with 2-6 drivers
Standard-library analysis: model event counts (orders, tickets, defects, visits) on the drivers you choose — the statistically correct way. Fits a Poisson regression, tests for overdispersion, and automatically upgrades to a negative binomial (or quasi-Poisson) model when the data demands it. Every effect is reported as a rate ratio — 'each unit of X multiplies the expected count by Y' — with confidence intervals and significance, plus a driver ranking and a model-quality readout. Works on any dataset: map a count column and 1-8 drivers.
Each driver's multiplicative effect on the expected count, with confidence intervals and plain-language significance.
The drivers ranked by statistical influence on the count, scaled 0-100.
The observed counts' shape — skew, zeros, range — the evidence that a count model was the right call.
The overdispersion verdict, which model was used and why, and how well it explains the counts.
Plain-English interpretation — what the numbers mean, what's significant, and what to do next.
What drives my order counts?
Map your per-row order count and the columns that might explain it. You get each driver's effect as a rate ratio — 'each extra promo email multiplies expected orders by 1.2' — from a model built for counts, not a linear regression that pretends counts are continuous.
See our FAQ for details on pricing, data privacy, and how the analysis works. Every report includes a Methodology section showing the statistical test, assumptions checked, and diagnostics run.
Run any analysis on your own data — validated R analyses, interactive reports, AI insights, and PDF export.
Try Free — No Credit CardTell us what went wrong, in your own words. We capture the page you're on automatically, so no need to describe where you are.