BigQuery consulting · Migration · Cost and slot engineering · 24×7 BigQuery support · Google Cloud data platform
BigQuery Consulting That Starts From Your Job History, Not From a Slide: Cost, Performance and Migration Measured in Slot-Milliseconds and Bytes Billed
MinervaDB BigQuery consulting is delivered by senior engineers with database-internals backgrounds who read INFORMATION_SCHEMA.JOBS before they recommend anything. The practice covers architecture and data modelling, query performance, on-demand versus Editions cost engineering, slot and reservation design, migration from Redshift, Snowflake, Teradata, Oracle, SQL Server and Hadoop, pipelines and lakehouse integration with BigLake and Iceberg, and security and governance. Every recommendation names the view it came from; every change is reversible and dual-run where it is not; and the same team carries 24×7 support afterwards. Vendor-neutral by principle: we resell no Google Cloud, so the honest answer can be "stay where you are" or "this workload belongs in ClickHouse or PostgreSQL".
01 · Why MinervaDB for BigQuery consulting
BigQuery removes the servers and leaves you every other decision
Serverless does not mean designless. Partitioning, clustering, slot capacity, reservation assignments, storage billing mode and query patterns decide whether BigQuery costs a tenth of your old warehouse or three times as much. Those decisions are visible in job history, and that is where BigQuery consulting begins.
Vendor-neutral
No Google Cloud resale, no partner tier to protect. BigQuery consulting recommendations come from your job history and billing export, including when the answer is on-demand with quotas instead of Editions, a smaller reservation, or a workload that belongs on ClickHouse or PostgreSQL.
Database-internals background
The engineers who tune your BigQuery estate have tuned optimizers, storage engines and shuffles on PostgreSQL, MySQL, ClickHouse and Redshift. They read a Dremel execution plan the way they read EXPLAIN ANALYZE: stage by stage, looking for the skew, the spill and the scan that should have been pruned.
Real 24×7 support
Named engineers on watch with S1 acknowledged within 15 minutes, S2 in 12 hours, S3 in 24 hours and S4 in 48 hours, covering jobs, slots, quotas and pipelines, with proactive job-history review for regressions. Google Cloud data platform engineering →
Measured, not promised
Nothing is claimed until the number has moved: slot-milliseconds and bytes billed per query shape, p95 job duration per workload, storage bytes by tier, and the invoice. Savings are estimates until the next billing export confirms them.
02 · BigQuery consulting services
Six disciplines around your BigQuery projects, one accountable team
Most BigQuery consulting engagements start with the health check and cost review and grow into the discipline the job history points at. All six are delivered by the same engineers.
Figure 1. The MinervaDB BigQuery consulting service map: architecture and data modelling, performance engineering, cost optimisation and FinOps, slot and reservation engineering, migration, and pipelines, lakehouse and governance, entered through the health check and sustained by 24×7 support.
Architecture and data modelling
BigQuery consulting starts with the table: partitioning by ingestion time, event date or integer range, clustering by the columns queries actually filter on, nested and repeated fields where they remove joins, dataset and project layout for IAM and billing, materialized views for repeated aggregates, and BI Engine where dashboards justify it.
Performance engineering
Execution plans and stage statistics from INFORMATION_SCHEMA.JOBS, shuffle bytes and spill, skew, slot contention, UDF and JavaScript cost, join strategy, ORDER BY without LIMIT, and the query rewrites and schema changes that move slot-milliseconds and bytes billed together.
Cost optimisation and FinOps
On-demand versus Editions modelled from ninety days of job history, storage billing mode (logical versus physical) chosen from compression ratios, partition expiry and time-travel windows, custom quotas per project and user, and a monthly cost-per-query curve. Cloud database FinOps →
Slot and reservation engineering
Reservations per workload with baseline and autoscale slots sized from JOBS_TIMELINE, assignments by project and job type, idle-slot sharing, query queues and concurrency, and the edition chosen per reservation from the features actually required.
Migration into BigQuery
Redshift, Snowflake, Teradata, Oracle, SQL Server and Hadoop: source inventory, SQL translation with the BigQuery migration service and hand re-engineering, schema redesign for partitioning and clustering, historical load through GCS and the Storage Write API, Datastream CDC to keep in sync, and a dual-run cutover with the source retained. Data modernization →
Pipelines, lakehouse and governance
Pub/Sub, Dataflow and Datastream ingestion, Dataform or dbt transformations, BigLake and Iceberg tables over GCS, Dataplex governance, IAM with column-level and row-level security, policy tags and audit logs, and the evidence pack for GDPR, HIPAA, PCI DSS and SOC 2. Data analytics platform engineering →
03 · BigQuery consulting method
How BigQuery executes a query, and the metric that exposes each stage
Every slow or expensive BigQuery job is slow or expensive at a specific stage: the read that should have pruned partitions, the shuffle that spilled, the sort on one slot. BigQuery consulting names the stage from the job's own statistics before changing anything.
Figure 2. BigQuery consulting view of query execution: capacity, Dremel stages on slots over Colossus columnar storage, partition and cluster pruning, shuffle and join, with the INFORMATION_SCHEMA views and stage statistics that expose each stage.
A BigQuery consulting baseline is ninety days of INFORMATION_SCHEMA.JOBS aggregated by query hash: total slot-milliseconds, bytes processed and billed, duration percentiles, cache hits and referenced tables, joined to the billing export so each query shape carries its cost. Stage statistics add shuffle output and spilled bytes, records per stage and the wait, read, compute and write ratio that separates a slot-starved job from a badly written one.
Changes are made one at a time and are reversible: a clustering change applied to a copy first, a materialized view that can be dropped, a reservation resized in minutes, a quota that can be lifted. Schema rewrites are dual-run with parity checks before a BI connection moves. Blast radius and rollback are written before execution.
-- Heaviest query shapes, last 30 days: slots, bytes, p95
SELECT
query_info.query_hashes.normalized_literals AS query_hash,
COUNT(*) AS runs,
ROUND(SUM(total_slot_ms) / 3600000, 1) AS slot_hours,
ROUND(SUM(total_bytes_billed) / POW(1024, 4), 2)
AS tib_billed,
ROUND(APPROX_QUANTILES(
TIMESTAMP_DIFF(end_time, creation_time, SECOND), 100
)[OFFSET(95)], 1) AS p95_seconds,
COUNTIF(cache_hit) AS cache_hits,
ANY_VALUE(user_email) AS sample_user
FROM `region-us`.INFORMATION_SCHEMA.JOBS
WHERE creation_time >=
TIMESTAMP_SUB(CURRENT_TIMESTAMP(), INTERVAL 30 DAY)
AND job_type = 'QUERY'
AND state = 'DONE'
GROUP BY query_hash
ORDER BY slot_hours DESC
LIMIT 25; 04 · BigQuery consulting for cost
On-demand or Editions: the crossover is measured from your job history, not guessed
BigQuery bills compute two ways and storage two ways. BigQuery consulting models all four against ninety days of your own jobs and storage, then applies the controls that move each line.
Figure 3. The BigQuery consulting cost model: on-demand bytes billed versus Editions slot-hours, logical versus physical storage, how the crossover is measured, and the controls that move each line. Prices for the US multi-region as verified on 28 September 2026.
| Line | Price (US multi-region, 28 Sep 2026) | What moves it | Where it is measured |
|---|---|---|---|
| On-demand compute | $6.25 per TiB scanned; first 1 TiB per month free | Partition filters, clustering, column selection, materialized views, BI Engine, custom quotas | total_bytes_billed per job and query hash in INFORMATION_SCHEMA.JOBS |
| Editions compute | Standard $0.04, Enterprise $0.06, Enterprise Plus $0.10 per slot-hour pay-as-you-go; 1-year and 3-year commitments 10% and 20% lower | Baseline and autoscale sizing, assignments, idle-slot sharing, concurrency and queues | JOBS_TIMELINE and RESERVATIONS_TIMELINE slot usage versus capacity |
| Logical storage | About $23.55 per TiB per month active; about half after 90 days unmodified | Partition expiry, snapshot and time-travel windows, dropping unused tables | TABLE_STORAGE active and long-term logical bytes |
| Physical storage | About double the logical rate, billed on compressed bytes; often cheaper for compressible data | Billing-model switch per dataset after comparing logical and physical bytes | TABLE_STORAGE physical bytes and compression ratio |
Prices from the BigQuery pricing page on 28 September 2026; other regions differ, and commitments change the slot-hour rate. Every saving on this page is an estimate until the next billing export confirms it.
05 · BigQuery consulting for slots and reservations
Reservations sized from the timeline, assigned by workload, tuned monthly
A reservation is a capacity contract you can change in minutes. BigQuery consulting sizes the baseline at the measured median of demand, the autoscale ceiling at the measured p95, and assigns workloads so nothing important falls back to on-demand by accident.
Figure 4. Slot and reservation engineering: an admin project holding reservations per workload with baseline and autoscale slots, assignments by project and job type, idle-slot sharing, and the sizing rule from JOBS_TIMELINE. Names and slot counts are illustrative.
Baseline and autoscale
BigQuery consulting sets the baseline at the p50 of concurrent slot demand over the busiest recurring window and autoscale to the p95; above the ceiling jobs queue and p95 duration rises, below it you pay for slots that never run. Both numbers come from JOBS_TIMELINE per second, and both are re-read monthly.
Assignments and editions
Assignments by project and job type (QUERY, PIPELINE, ML_EXTERNAL); a workload without one falls to on-demand and is billed by bytes. Editions are chosen per reservation: Enterprise Plus only where cross-region disaster recovery or its higher security controls are actually required.
Fairness and queues
Query queues and maximum concurrency per reservation stop one analyst's scan from starving the pipelines; idle baseline slots are shared across reservations unless a workload needs isolation; pending job counts and per-reservation p95 duration are the fairness evidence.
06 · BigQuery consulting for migration
Into BigQuery from Redshift, Snowflake, Teradata, Oracle, SQL Server and Hadoop, dual-run and reversible
A warehouse migration is a data-model redesign with a cutover attached. BigQuery consulting translates the SQL, redesigns the schema for partitioning and clustering, loads and syncs, and proves parity before a single BI connection moves.
Figure 5. BigQuery consulting migration path: assess, redesign, load history, keep in sync with Datastream CDC, verify and cut over with the source retained, and the source-specific traps planned for.
What changes in the model
In BigQuery consulting for migration, distribution and sort keys have no equivalent; partitioning and clustering replace them and are chosen from the query log, not the old DDL. Nested and repeated fields remove joins where the access pattern allows. Stored procedures move to Dataform, dbt or application code, and dialect differences in case sensitivity, NULL ordering and numeric precision are caught by the parity suite, not by users.
How the cutover stays reversible
History is loaded from GCS in Parquet or Avro with row counts and checksums per partition; Datastream or Dataflow keep the target current with lag measured commit to queryable; the reporting suite dual-runs on both warehouses until result parity is signed off; and the source warehouse is retained through the observation window, so the rollback is a connection-string change.
07 · BigQuery health check and cost review
The fixed-scope entry point to BigQuery consulting
Read-only, evidence-based, priced per project, and credited against a support retainer signed within 30 days. Nothing changes in the estate during the review.
What is reviewed
BigQuery consulting reviews ninety days of job history by query shape, user and project; schema design (partitioning, clustering, nested fields, table sprawl); slot utilisation against reservations and autoscale ceilings; storage billing by tier and mode; on-demand exposure and quotas; pipeline freshness; and security posture including IAM, policy tags, row and column security and audit logs.
What you receive
Findings ranked by measured impact with estimated savings, each citing the INFORMATION_SCHEMA view or billing-export field behind it and the reversible change that moves it; a prioritised remediation plan; and a versioned report your team keeps whether or not MinervaDB does the remediation. Typical turnaround is one to two weeks.
Standing caveat: every recommendation on this page is tested on a copy or in a dual-run before it is applied to production datasets, with table snapshots taken first, and storage and reservation changes are made reversible by design.
08 · BigQuery consulting rates
Transparent BigQuery consulting rates
Priced to the market for senior cloud data-warehouse expertise. No minimum block of hours for remote consulting; retainers include named engineers and severity SLAs.
Figure 6. BigQuery consulting engagement types and rates: health check and cost review, migration assessment, remote consulting, 24×7 support retainer and on-site consulting, with how an engagement runs.
Health check and cost review · US $6,500 per project
Ninety days of job history, schema, slot utilisation, storage billing and security reviewed; findings ranked by measured impact with estimated savings; one to two weeks; credited against a retainer signed within 30 days.
Migration assessment · US $9,500
Source inventory, SQL translation scope, schema redesign plan, load and cutover approach, cost model against your current warehouse, and a written recommendation, including "stay where you are" when that is the honest answer.
Remote consulting · US $195 per hour
Hourly BigQuery consulting: architecture, data modelling, performance and cost engineering by senior engineers, billed by the hour with no minimum, available on short notice worldwide.
24×7 support retainer · from US $3,500 per quarter
Named engineers, S1 in 15 minutes, incident response on jobs, slots, quotas and pipelines, proactive job-history review, monthly cost and performance report, onboarding assessment and runbook included. On-site consulting at US $350 per hour plus travel.
09 · FAQ
BigQuery consulting questions we are asked most
Short answers to what data and platform leaders ask before the first call.
What does a BigQuery consulting engagement include?
A discovery call on workloads, projects, reservations and pain points, then a scoped engagement: the health check and cost review, an architecture or data-modelling design, a migration assessment, slot and reservation engineering, or a 24x7 support retainer. Every engagement ends with written deliverables that cite the INFORMATION_SCHEMA view or billing-export field behind each finding, reversible changes with rollback, and an action plan with the metric each item is expected to move.
Should we be on on-demand or Editions?
It depends on your demand curve, and it is measured rather than guessed: ninety days of JOBS give bytes billed per project and user and slot demand per second, and we simulate Editions cost at the measured percentiles against the on-demand bill. Mixed estates are common, batch and ELT on a reservation, ad hoc BI on on-demand with quotas, or the reverse. The crossover is re-checked monthly against the invoice.
How do you cut BigQuery cost without slowing queries?
By moving both numbers together: partition filters and clustering cut bytes scanned and slot-milliseconds at once; materialized views and BI Engine remove repeated scans; reservations sized at the measured median with autoscale to the p95 remove idle capacity; physical storage billing cuts the storage line where data compresses well. Every change is reversible and its effect is read from the next billing export.
Can you migrate us into BigQuery from Redshift, Snowflake or Teradata?
Yes, and from Oracle, SQL Server and Hadoop. The migration assessment inventories the source, scopes SQL translation, redesigns the schema for partitioning and clustering, plans the historical load and CDC sync, and models cost against the current warehouse, including the honest recommendation to stay where you are. Cutover is dual-run with parity checks and the source retained through the observation window.
How do you size slots and reservations?
From INFORMATION_SCHEMA.JOBS_TIMELINE per second: baseline at the p50 of concurrent slot demand over the busiest recurring window, autoscale ceiling at the p95, assignments by project and job type so nothing important falls to on-demand, and the edition chosen per reservation from the features actually required. Reservations are re-read monthly and changed in minutes when the demand curve moves.
Do you provide 24x7 BigQuery support?
Yes. Retainers start at US $3,500 per quarter with named engineers, S1 acknowledged in 15 minutes, S2 in 12 hours, S3 in 24 hours and S4 in 48 hours, incident response on jobs, slots, quotas and pipelines, proactive job-history review for regressions, and a monthly cost and performance report. The onboarding assessment and runbook are included.
Is MinervaDB a Google Cloud partner or reseller?
No. MinervaDB resells no Google Cloud and holds no partner incentives, so BigQuery consulting recommendations carry no commission and can include a smaller reservation, on-demand with quotas, or moving a workload to ClickHouse or PostgreSQL where the numbers say so.
How much does BigQuery consulting cost?
Remote consulting is US $195 per hour with no minimum; on-site consulting is US $350 per hour plus travel; the health check and cost review is US $6,500 per project and credited against a retainer signed within 30 days; the migration assessment is US $9,500; and 24x7 support retainers start at US $3,500 per quarter.
Talk to a senior BigQuery consulting engineer
Bring a billing export, ninety days of INFORMATION_SCHEMA.JOBS and the reservation configuration to the first call. We will tell you where the slot-hours and bytes go, whether on-demand or Editions fits your demand curve, and what we would change first.