Comparing cost: Building a custom data pipeline vs. buying a CDP

Most teams drastically underestimate the ongoing cloud bill for a custom pipeline. You're not just paying for compute. You're funding a dedicated team to build, monitor, and scale it.

For a typical SaaS with 1M MAU, here's the baseline infrastructure cost for a simple collector + warehouse setup on AWS:

```yaml
# Monthly AWS Estimate (~$8k-$12k)
- Kinesis Data Streams: $1,200 (2 shards, 1M events/day)
- ECS Fargate (Transform): $900 (4 vCPU, 8GB RAM)
- Lambda (Enrichment): $450 (5M invocations)
- S3 Storage: $300 (10TB processed data)
- Redshift Serverless RPUs: $5,000+ (variable)
- Data Transfer & Misc: $1,000
```

Now add:
* 0.5 FTE DevOps engineer ($7k/month)
* Unplanned cost spikes from traffic surges
* Reserved Instance management overhead
* Cost of failed events and reprocessing

A commercial CDP at this scale runs $15k-$25k/month. The build-vs-buy math only works if your data volume is massive and your engineering time is free. For most, the CDP is cheaper once you factor in labor.