Mixpanel launches Predict, which applies machine learning to predict user engagement in as little as 30 minutes, working to speed up diagnostics to 30 seconds
David Lumb / Fast Company :
Context & Ripple Effects
By late 2015, predictive analytics was becoming table stakes for SaaS: a month earlier Zendesk had launched its own machine-learning tool to flag customer-service problems before they occur (predictive support analytics), while Amazon opened a general-purpose ML platform for developers (Amazon's managed machine-learning platform) and LinkedIn bought Refresh.io to fold predictive insights into its products. Mixpanel's Predict is the product-analytics answer to that wave — instead of exporting event data elsewhere, users run engagement forecasts inside the tool, with results in as little as 30 minutes and diagnostics targeted down to 30 seconds.
First-order effects
- Product teams using Mixpanel get engagement predictions without building or operating their own models, shortening the loop from anomaly to diagnosis from days to minutes.
Second-order effects
- Other analytics and workflow vendors are pushed to match: Zendesk has already embedded prediction in support, People.ai raised a Series B led by Andreessen Horowitz to predict sales outcomes from communication touchpoints, so raw dashboards alone start losing pricing power against bundled intelligence.
Third-order effects
- If the pattern holds — through Crunchbase predicting startup funding trajectories from seventeen years of data and Simile emerging from stealth with $100M to predict human behavior — the durable asset is each company's proprietary behavioral dataset, with machine-learning prediction layered on top as the default monetizable feature.
The trend: Application software companies are converting the behavioral data they already collect into built-in machine-learning predictions, making prediction a standard feature layer rather than a separate product category.