$30
Monthly Base Cost
Workers Paid ($5) + Workers for Platforms ($25)
848K
Total Worker Requests
Across 22 days of development + scale testing
15,625
Survey Sessions
16 studies, including 15K+ scale test sessions across 2 studies
$0.00
Overage Charges
Zero. Every service within included tier.
Usage Breakdown by Service
| Service | Metric | Usage | Included in Plan | Overage Rate | Cost |
|---|---|---|---|---|---|
| Workers for Platforms | Requests | 848,313 | 20,000,000 /mo | $0.30 per million | $0.00 |
| CPU time | ~4.2% of included | 60M CPU-ms /mo | $0.02 per million ms | $0.00 | |
| Durable Objects | Requests | ~410,000 | 1,000,000 /mo | $0.15 per million | $0.00 |
| SQLite storage | ~50 MB | 5 GB | $0.75 per GB-mo | $0.00 | |
| Queues | Messages | ~610,000 | 1,000,000 /mo | $0.40 per million | $0.00 |
| R2 | Storage | 224 KB | 10 GB /mo | $0.015 per GB-mo | $0.00 |
| Operations | ~15,000 | 10M Class A /mo | $0.36 per million | $0.00 | |
| D1 | Rows read | ~30,000 | 25B rows /mo | $0.001 per million | $0.00 |
| Storage | 3.1 MB | 5 GB | $0.75 per GB-mo | $0.00 | |
| ClickHouse Cloud | Storage | 19.3 MB | Serverless tier (~$50/mo base) | ~$50.00 | |
| Total Monthly Cost | ~$80/mo | ||||
What This Paid For
| Metric | Count | Notes |
|---|---|---|
| Studies published | 16 | Each deployed as isolated UserWorker + DO class |
| Survey sessions | 15,625 | Each with its own Durable Object + SQLite |
| Answers collected | 584,422 | Stored in DO SQLite + ClickHouse |
| Events tracked | 606,376 | card_answered, session_started, session_complete |
| Scale test: 100 concurrent | 100% pass | 56ms p50, 0 SSE drops |
| Scale test: 500 concurrent | 100% pass | 56ms p50, 0 SSE drops, burst mode |
| Scale test: 1000 concurrent | 100% pass | 56ms p50, 0 SSE drops, 300/wave |
| Scale test: 500 realistic (5 min) | 100% pass | 492 concurrent SSE held for 5+ min |
| Scale test: 500 realistic (coffee_habits) | 100% pass | 55ms p50, 500 concurrent SSE, schema-driven |
| Worker invocations (all time) | 848,313 | Across 22 days, 2 studies scale tested |
Cost Per Survey Engagement
Based on actual usage: 848K requests for 15,625 sessions = ~54 requests per session average (varies by study size)
$0.00
Cost Per Response
All 15,625 sessions within included tier
$0.017
At Overage Rate
If every request were billed: 56 reqs × $0.30/M
$0.00
Per Response: Idle Cost
No traffic = no charge. $0 idle. Always.
Monthly Cost Breakdown
Rival AI Responding - Infrastructure
Workers Paid (base)$5/mo
Workers for Platforms (base)$25/mo
Durable Objects$0 (included)
Queues$0 (included)
R2 storage$0 (included)
D1 database$0 (included)
ClickHouse Cloud (serverless)~$50/mo
ETL pipeline$0 (not needed)
Load balancer / proxy$0 (CF edge handles it)
DevOps / monitoring$0 (serverless, no ops)
Total monthly~$80
$0 idle. Scales with usage. No ops.
Note: This cost does not include any AI/LLM costs (authoring, AI probes, voice interviews). Those are billed separately by the model provider.
Cost Projection at Scale
Based on ~56 worker requests per survey completion at $0.30 per million requests after 20M included
10 studies × 1,000 respondents
~$80
560K requests
Well within included tier
Zero overage
Well within included tier
Zero overage
50 studies × 2,000 respondents
~$82
5.6M requests
Still within included tier
Zero overage
Still within included tier
Zero overage
200 studies × 5,000 respondents
~$94
56M requests
36M over included
$10.80 overage
36M over included
$10.80 overage
500 studies × 10,000 respondents
~$164
280M requests
260M over included
$78 overage
260M over included
$78 overage
The Economics
15,625 survey sessions. 584,422 answers. 16 studies. Scale tested to 1,000 concurrent (burst) and 500 concurrent (realistic, 5+ min hold). Two different studies tested with a schema-driven framework - zero study-specific code. Total infrastructure cost: $80/month.
No servers to manage. No databases to scale. No ETL pipelines to maintain. No load balancers to configure. No idle costs. CPU-time billing means you pay for value delivered, not infrastructure reserved.
At 500 studies with 10,000 respondents each (5 million total completions per month), the total responding infrastructure cost would be approximately $164/month.
Note: This report covers responding infrastructure only. AI/LLM costs (for authoring, AI probes, voice interviews) are billed separately by the model provider and are not included in these figures.
No servers to manage. No databases to scale. No ETL pipelines to maintain. No load balancers to configure. No idle costs. CPU-time billing means you pay for value delivered, not infrastructure reserved.
At 500 studies with 10,000 respondents each (5 million total completions per month), the total responding infrastructure cost would be approximately $164/month.
Note: This report covers responding infrastructure only. AI/LLM costs (for authoring, AI probes, voice interviews) are billed separately by the model provider and are not included in these figures.