---
title: "The Evolution of Kit: Automating Marketing Using Machine Learning"
description: "Kit began as a heuristic, rules-based virtual employee that simplified Facebook advertising for Shopify merchants by selecting products and budgets while defaulting other settings to best practices. The post describes its evolution into an artificially intelligent assistant that uses merchant spending behavior and store signals such as traffic and sales to recommend marketing actions. Kit's machine-learning design combines regression to predict likely spending with classification to estimate whether a budget and store state are likely to generate sales; training uses feature engineering, Apache Spark, Google Cloud ML Engine, and monitored model metrics. For real-time predictions, Bigtable stores engineered features, while a Tornado web container delegates model computation to TensorFlow Serving. The resulting system powered one third of Kit's marketing campaigns."
---

# The Evolution of Kit: Automating Marketing Using Machine Learning

[Shopify](https://yomu.fyi/company/shopify) · 2023-10-18 · Mar 4, 2020

**Type:** Explainer

## Summary

Kit began as a heuristic, rules-based virtual employee that simplified Facebook advertising for Shopify merchants by selecting products and budgets while defaulting other settings to best practices. The post describes its evolution into an artificially intelligent assistant that uses merchant spending behavior and store signals such as traffic and sales to recommend marketing actions. Kit's machine-learning design combines regression to predict likely spending with classification to estimate whether a budget and store state are likely to generate sales; training uses feature engineering, Apache Spark, Google Cloud ML Engine, and monitored model metrics. For real-time predictions, Bigtable stores engineered features, while a Tornado web container delegates model computation to TensorFlow Serving. The resulting system powered one third of Kit's marketing campaigns.

## Context

The heuristic implementation gave every business owner the same hardcoded budget options and required decisions about products and campaign setup. The post states that these choices could create friction, particularly for inexperienced marketers, and that budgeting was an impactful factor in campaign success.

## Approach / What changed

Kit replaced static recommendations with machine-learning-driven recommendations. Regression predicts a business owner's likely budget from historic spending, while classification estimates the likelihood of sales using budget and store attributes. The architecture includes Spark-based feature engineering, Google Cloud ML Engine training, metric monitoring, Bigtable feature storage, and real-time prediction through a Tornado web container and TensorFlow Serving.

## Takeaways

- The recommendation models address two separate problems: predicting the budget a business owner is likely to spend and predicting whether that budget is sufficient to generate sales.
- Model monitoring uses threshold alerts and z-score-based outlier detection to identify abnormal metrics, feature problems, or changing data distributions before models are used.
- Machine-learning-driven recommendations powered one third of Kit's marketing campaigns, according to the post.

**Tags:** [Apache Spark](https://yomu.fyi/topic/apache-spark), [Google Cloud](https://yomu.fyi/topic/gcp), [Machine Learning](https://yomu.fyi/topic/machine-learning), [Model Serving](https://yomu.fyi/topic/model-serving), [Monitoring](https://yomu.fyi/topic/monitoring)

- Source: [Shopify](https://shopify.engineering/evolution-kit-automating-marketing-machine-learning)
- Source URL: https://shopify.engineering/evolution-kit-automating-marketing-machine-learning
- Ingested by Yomu: 2026-08-31T01:13:07.555Z

[Read original post](https://shopify.engineering/evolution-kit-automating-marketing-machine-learning)