DataRev
Project cost calculator
Not just the cloud bill. A data and AI project costs technology (the platform, month to month), people (who builds it and who runs it), process (governance, training, coordination) and AI/agents (design, development and deployment). Adjust the assumptions and all four recalculate instantly.
Pick which items are in scope. The summary below adds up only the active ones.
Project summary
What the selected items add up to, on whichever platform you pick below.
Monthly (recurring)
$687
One-time
$100,365
First year
$108,609
One-time plus twelve months of run rate
- Technology$318/mes
- People$369/mes · $28,480 one-time
- Process$15,185 one-time
- AI / Agents$56,700 one-time
Who will build it
The team this quote demands, turned into people. Derived from the same lines that produce the cost, so the two cannot contradict each other.
| Role | People | Person-days | Cost |
|---|---|---|---|
| Data product managerHolds scope and sequencing. Without it the backlog is set by whoever shouts loudest. | 1 | 6.3 | $6,320 |
| Architect / tech leadDesigns storage, access and system boundaries. Triggered when where the data lives is still an open question. | 1 | 10 | $14,000 |
| Data engineerBuilds the pipes: ingestion, integration and the loads that run unattended. | 1 | 36.2 | $32,580 |
| BI analystTurns the model into an answer the business uses on Monday. | 1 | 34 | $22,100 |
| ML engineerTakes the model into a production system with committed latency and availability. | 1 | 15 | $16,500 |
| Governance and quality leadCatalog, lineage, quality and access policy. Cisco 2025: only 51% have their data centralized. | 1 | 4.5 | $4,275 |
| Change management and adoptionMaking people actually use it. Cisco 2025: only a third have a formal change-management plan. | 1 | 5.4 | $4,590 |
| Total | 7 | 111.4 | $100,365 |
Rounded up: half an engineer does not show up on Monday. Shortening the window does not reduce person-days, it raises headcount — and with it the cost of coordination. Roles activate on their own according to what the quote includes: a project with no AI carries neither an ML engineer nor a Forward Deployed one.
What you have today
Your current estate. Drives what the migration costs.
ERP, CRM, operational databases, APIs
Everything living in databases and files today
What you want
The service level you want to reach. Drives the monthly cost of running it.
What dashboards and analysts read. Usually several times the stored volume.
Who will use it
Drives licensing.
Editing-licence authors/developers (QuickSight Author, Power BI Pro creator, Looker Studio editor). Not the delivery team doing the migration — that is below, under Migration.
Viewer count is the single biggest lever in this comparison. That is where the platforms actually diverge.
What you need
This is not entered here: it comes from the use cases you picked in the planner.
3 use cases selected in the planner · Change the selection
Keeping a platform running costs people's time. Switching this off makes the open option look cheaper than it is.
Estimated monthly cost500 GB
Open source
Supabase + Airbyte + dbt + Metabase
$1,292
USD / month · $15,508 annual
- Platform$392
- Licences$0
- Operations$900
Worth pointing out
- Zero per-user licences: the structural advantage of this column, and it grows with headcount.
- The real cost sits in the ops line, not the platform line. Ignore it and this column looks free, which it is not.
Google Cloud
CheapestBigQuery + Cloud Storage + Looker Studio
$687
USD / month · $8,250 annual
- Platform$318
- Licences$99
- Operations$270
Worth pointing out
- On-demand still wins. You pay per byte scanned, so partitioning and clustering cut the bill directly.
- Looker Studio is free to read. That is the main reason the licence column is low here.
AWS
S3 + Redshift Serverless + Glue + QuickSight
$1,307
USD / month · $15,678 annual
- Platform$563
- Licences$384
- Operations$360
Worth pointing out
- Redshift Serverless bills per second while working. Concentrated workloads make it very efficient; a warehouse left awake all day does not.
- QuickSight charges readers, unlike Looker Studio. With many readers that line grows fast.
Microsoft Azure
Microsoft Fabric + OneLake + Power BI
$2,047
USD / month · $24,560 annual
- Platform$1,063
- Licences$714
- Operations$270
Worth pointing out
- At this headcount the small capacity wins. The crossover sits around 526 users: past that, jumping to F64 and dropping viewer licences is cheaper.
- Fabric bills one compute pool for ingest, transform, warehouse and Power BI refresh. It simplifies the invoice but lets one bad job starve the dashboards.
Comparison by layer
Each bar sums platform, licences and operations. The proportion between layers matters more than the total.
- Platform
- Licences
- Operations
Portable engines: Snowflake and Databricks
Neither is a cloud. They are engines that run on top of AWS, GCP or Azure, so they compare as a swap for the native engine — not as a fifth column. Object storage, orchestration and BI licences still land on the host cloud's bill.
All three clouds, together
The same nine scenarios (three engines × three hosts) in one table, so you can compare without switching tabs.
| Engine | GCP | AWS | Azure |
|---|---|---|---|
| Nativo | $318 | $563 | $1,063 |
| Snowflake | $1,044 | $1,095 | $1,007 |
| Databricks | $819 | $871 | $784 |
Detail by cloud
Redshift
Native$563
USD / month
- S3 (data lake)$12
- Redshift Managed Storage$7
60% of volume — the modelled subset only
- Redshift Serverless$183
488 RPU-hours · 8 RPU base
- Glue (ETL)$11
25 DPU-hours
- Orchestration (MWAA)$350
The host cloud's own engine. Simpler integration and a single invoice, at the price of being tied to that vendor.
Snowflake
$1,095
USD / month
Engine $740 · Host cloud $355
- Snowflake compute$729
M warehouse · 243 credits × $3 (Enterprise)
- Snowflake storage$11
0.5 TB × $23 (capacity rate)
- Object storage on AWS$5
- Orchestration on the host cloud$350
Bills per second while the warehouse is awake, not per byte read. Auto-suspend is the lever: dropping it from 10 minutes to 1 usually cuts the bill without anyone noticing.
Databricks
$871
USD / month
Engine $509 · Host cloud $362
- Databricks SQL Serverless$507
Small warehouse · 724 DBU × $0.7 · underlying compute included
- Jobs Compute (ETL)$2
15 DBU × $0.15 · 3-4× cheaper than an interactive cluster
- Delta Lake on AWS$12
No storage markup: the data lives in your own object storage
- Orchestration on the host cloud$350
The expensive mistake here is running production pipelines on interactive clusters: 3-4× the cost for the same work. On AWS and GCP you also get two invoices, one from Databricks and one from the cloud.
Only the engine and platform layers are compared. BI licences and operations do not change with the engine, so including them would only blur the comparison.
One-time migration cost
Everything above is the monthly run rate. Getting there costs extra, and it is usually the number that decides the deal. Each stage is done by a different role, at a different rate — not one generic consultant.
Technical delivery
The people who build the technology
- Discovery and architecture· Architect / tech lead5 d · $7,000
- Integrate 5 sources· Data engineer15 d · $13,500
- History backfill and reconciliation· Data engineer0.2 d · $180
- Rebuild reports and semantic layer· BI analyst12 d · $7,800
Adoption process
What almost always gets left out of an estimate — and why estimates come in low
- Governance, quality and access· Governance and quality lead4.5 d · $4,275
- Change management and training· Change management and adoption5.4 d · $4,590
- Project management and coordination· Data product manager6.3 d · $6,320
48 consulting days
$34,932 – $61,131
Deliberately a range: a point estimate for migration work is always wrong.
AI / agents implementation
A fourth cost, separate from the data migration: designing, building and deploying agents with a client. A project may need one, the other, or both.
- Build — BI analyst· BI analyst22 d · $14,300
- Build — Data engineer· Data engineer21 d · $18,900
- Build — ML engineer· ML engineer15 d · $16,500
- Build — Architect / tech lead· Architect / tech lead5 d · $7,000
63 consulting days
$45,360 – $79,380
Deliberately a range: a point estimate for migration work is always wrong.
How to read these numbers
- This is an order-of-magnitude estimate to frame a conversation, not a quote. A real project is sized by measuring the workload, not estimating it.
- Not included: cross-cloud egress, private networking, support plans, negotiated commitments or the initial migration. All move the number materially.
- Rates were verified on 2026-08-09 against each vendor's official pricing page. They change often: re-verify before committing a number to a client.
- Reference rates are US East. Mexico Central and other LATAM regions typically run 5-20% higher.
Where each rate comes from
Not every figure carries the same weight, and it is worth knowing which is which before defending them to a client.
- OfficialFabric capacity and the F64 threshold — F2, F4, F64 and F2048 come from the Azure page. The intermediate SKUs (F8 to F256) are derived: the ladder is exactly linear at $131.40 per CU-month, checked against all four published points. The F64 threshold is confirmed by footnote 3 on the Power BI pricing page.
- OfficialPer-user licences — Power BI Pro $14 and PPU $24 from microsoft.com. QuickSight Author $24 and Reader $3 from aws.amazon.com.
- OfficialAWS storage and ETL — S3 Standard $0.023/GB and Glue $0.44/DPU-hour, both from the AWS pages. Redshift Managed Storage $0.024/GB from the same source; the $0.375/RPU-hour is derived from the published $1.50/hour minimum over 4 RPUs.
- OfficialSnowflake credits and Databricks DBUs — Snowflake credits ($2/$3/$4) and $23/TB storage from snowflake.com. Base DBU rates from databricks.com. The fine-grained Databricks SQL SKU breakdown ($0.70 serverless) comes from a third party.
- SecondaryBigQuery query and storage — The $0.06 slot-hour is on Google's page. The $6.25 per TiB scanned and the $0.02/$0.01 per GB storage are not: that table renders in JavaScript. Three independent sources agree, but confirm before quoting.
- AssumptionOrchestration, VMs and ops hours — Composer, MWAA, the open-stack VMs and the monthly ops hours are DataRev assumptions, not published rates. They are also the lines most worth replacing with the client's real figures.
- AssumptionWorkload conversion factors — Credits per TiB scanned, DBUs per TB processed, hours a warehouse stays awake, and migration days per source. Reasoned orders of magnitude, not measurements. This is where the result moves most.
- AssumptionTeam day rates — Architect/lead $1,400, data engineer $900, BI analyst $650, ML Engineer $1,100, Forward Deployed AI Engineer $1,500 — DataRev's own rates, editable on screen. Not a quote — a starting point for the scoping conversation.
Rate sources
What now?
Let's review your result together and decide which use case to tackle first. The session is free and commits you to nothing.