Marketing ops · Production

DoorDash builds ML-powered Marketing Automation platform to optimize spend across tens of thousands of campaigns

The problem

DoorDash was manually managing tens of thousands of marketing campaigns across multiple channels, a process both time-consuming and sub-optimal at scale, compounded by sparse or clustered campaign spend data that made reliable cost curve construction impossible without ML augmentation.

Workflow diagram · grounded in source
1
Attribution data ingestion
integration
“we use our internal attribution data to assign credit for every conversion to a specific channel based on modified last-touch attribution”
2
ML synthetic data generation
ai_action
“train an ML model to predict the expected number of conversions for any campaign, at any spend level. The ML model generates synthetic data, augmenting our real data. Then we fit cost curves to the combined synthetic and real data.”
3
Cost curve fitting
ai_action
“Then we fit a concave curve (to guarantee diminishing returns) to these data points”
4
Marginal-value budget allocation
ai_action
“Starting with zero spend for all channels, repeatedly allocate a dollar to the channel that has the highest slope at its current spend. Do this until the budget is reached, and we have our optimal spend targets.”
5
Automated bid publication
output
“it optimally allocates budget to each campaign and publishes bids to our channel partners”
Reported outcome

The Marketing Automation platform has begun managing some of the marketing team's weekly budget and is expected to lower marketing costs by 10 to 30 percent while reaching the same number of customers, freeing marketers to focus on strategy rather than bid management.

Reported metrics
Marketing cost reduction (projected)10 to 30 percent
Grid search parallelization speedupgreater than 100x speedup over no parallelism
Reported stack
Databricksscikit-optimizejoblib.ParallelGoogleFacebook
Source
https://careersatdoordash.com/blog/optimizing-marketing-spend-with-ml/
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The Marketing Automation platform has begun managing some of the marketing team's weekly budget and is expected to lower marketing costs by 10 to 30 percent while reaching the same number of customers, freeing markete…

What tools did this team use?

Databricks, scikit-optimize, joblib.Parallel, Google, Facebook.

What results were reported?

Marketing cost reduction (projected): 10 to 30 percent; Grid search parallelization speedup: greater than 100x speedup over no parallelism (source-reported, not independently verified).

How is this marketing ops AI workflow structured?

Attribution data ingestion → ML synthetic data generation → Cost curve fitting → Marginal-value budget allocation → Automated bid publication.

Related marketing ops cases