01
Assessment
Sources, data volumes, freshness needs, users, compliance rules and today’s reporting pain, written down before any platform is picked.
FIG 01 · Data Warehouse Consulting
Data warehouse consulting, from first model to fast dashboards
A data warehouse your team can afford to run. We pick Snowflake, Databricks or BigQuery for your workloads, model it in dbt, and set the guardrails so the bill doesn’t surprise finance.
See how it works
FIG 02 · Quick answer
What is data warehouse consulting?
Data warehouse consulting is planning, building and tuning the central database your reports and AI tools query. Elgrith designs Snowflake, Databricks or BigQuery warehouses, migrates data from SQL Server, Redshift or spreadsheets, models it with dbt, and sets compute limits so costs stay predictable, with every account in your name.
- Snowflake, Databricks and BigQuery compared on your workloads
- 3 platforms
- Projects delivered since 2015
- 1,000+
- Clients across the USA, UK and India
- 250+
- Your warehouse account, dbt project and repository
- 100%
FIG 03 · Why it matters
Warehouse bills follow how you run it, not just what you buy.
Three billing and start-up rules from Snowflake’s and Databricks’ own documentation, checked 24 Oct 2026. We size and schedule around them.
01
60 s
Minimum charge each time a Snowflake warehouse starts
Snowflake bills warehouse credits per second, with a 60-second minimum every time a warehouse starts or resumes. Many short queries on a cold warehouse add up.
Source: Snowflake documentation, Understanding compute cost (checked 24 Oct 2026)(opens in a new tab)02
10%
Cloud services allowance before Snowflake bills it
Cloud services usage is charged only on days when it exceeds 10% of that day’s warehouse usage, so metadata-heavy workloads need watching.
Source: Snowflake documentation, Understanding compute cost (checked 24 Oct 2026)(opens in a new tab)03
2–6 s
Databricks serverless SQL start-up
Databricks says serverless SQL warehouses typically start in 2 to 6 seconds; pro and classic warehouses take about 4 minutes. That changes how aggressively you can auto-stop.
Source: Databricks documentation, SQL warehouse types (updated 11 Sep 2026)(opens in a new tab)
FIG 04 · Definition
Warehouse, lake or lakehouse?
A data warehouse stores cleaned, modeled tables for fast SQL. A data lake stores raw files of any shape, cheaply. A lakehouse puts warehouse-style tables (Delta Lake or Apache Iceberg) on top of lake storage. Most teams end up with a mix; here is when each leads.
| Data warehouse | Data lake | Lakehouse | |
|---|---|---|---|
| Stores | Modeled tables (star schema, marts) | Raw files: logs, JSON, images, exports | Open table formats on lake storage |
| Best for | Finance, BI dashboards and governed reporting | Data science, archives and unknown future uses | Mixed BI, data science and streaming on one copy |
| Typical platforms | Snowflake, BigQuery, Amazon Redshift | Object storage on AWS, Azure or Google Cloud | Databricks, or Snowflake and BigQuery with Iceberg tables |
| Watch out for | Paying to copy data you rarely query | A “data swamp” nobody trusts | More moving parts: formats, catalogs and engines |
The platform matters less than the model and the guardrails. We run your real queries on a shortlist before you commit.
Go deeper: Snowflake vs Databricks in 2026FIG 05 · What you get
What our data warehouse consulting services cover
From the first assessment to the last dashboard hand-off, with cost guardrails and documentation built in rather than bolted on.
02
Platform choice
Snowflake, Databricks or BigQuery scored on your workloads, skills and existing cloud, with a proof-of-concept on your real queries.
03
Architecture
Medallion layers (bronze, silver, gold) or dimensional marts with star schemas and slowly changing dimensions, sized to how you report.
04
Ingestion
Fivetran, Airbyte or custom connectors, with change data capture where freshness matters. Our data engineering team builds the heavier pipelines.
05
dbt modeling
Tested, documented dbt models in your repository, so every metric has one definition and a history of changes.
06
Orchestration
Airflow, Dagster or the platform’s own jobs, with retries, alerts and clear ownership for every scheduled load.
07
Governance and access
Roles, row-level security, masking for personal data and a catalog, set up with least privilege from day one.
08
BI hand-off
Clean marts and a semantic layer ready for Power BI, Tableau or Looker, with our data visualization team if you want the dashboards too.
09
Cost guardrails
Auto-suspend, right-sized warehouses, query tags, budgets and alerts, so spend tracks usage you meant to have.
10
Documentation
Architecture notes, model docs, runbooks and access lists, written for the people who run it after we leave.
Deliverables
11
deliverables, in writing
Included in every build, written into your fixed quote before work starts.
FIG 06 · Migration
Migrating from SQL Server, Redshift, Teradata or spreadsheets
Most warehouses replace something. We move it with change data capture, a parallel run and reconciliation reports, so finance signs off the numbers before the old system is switched off.
- SQL Server
- Amazon Redshift
- Spreadsheets
CDC, parallel run, reconciliation
- Modeled warehouse
- Dashboards
- AI tools
SQL Server and on-prem databases
- From
- SQL Server, Oracle or MySQL reporting databases and SSIS packages
- To
- Snowflake, Databricks or BigQuery, with dbt models replacing stored-procedure logic
Amazon Redshift
- From
- Redshift clusters that are hard to size or share
- To
- Redshift Serverless, or another platform if your workloads fit it better
Teradata and legacy appliances
- From
- Appliance warehouses nearing renewal
- To
- A cloud warehouse, moved subject area by subject area with a parallel run
Spreadsheets and SaaS exports
- From
- Excel workbooks, CSV exports and copy-paste reports
- To
- Automated loads from your CRM, ERP and finance tools into tested models
Pipelines first?
- When
- Your sources need APIs, streaming or heavy transformation before they can land
- Who
- Our data engineering team builds the pipelines; we design the warehouse they feed
Data migration checklist
- Covers
- Inventory, mapping, test loads, reconciliation, cut-over and rollback
- Use it
- Before you book a cut-over window
FIG 07 · Role
What a data warehouse consultant does (and when you need one)
A data warehouse consultant turns scattered data into one trusted model, picks and sets up the platform, and leaves your team able to run it.
04principles
Finds the real requirements
Which questions the business asks, how fresh the answers must be, who may see what, and what today’s reports get wrong.
Chooses the platform with evidence
Runs your heaviest queries on a shortlist instead of picking the platform with the best conference talk.
Designs the model
Facts, dimensions and marts that match how you report, with one definition per metric.
Hands over, then steps back
Code reviews, runbooks and training so your analysts and engineers own it. You need one when reports disagree, loads break or the bill grows faster than usage.
FIG 08 · Snowflake
Snowflake consulting services
Virtual warehouses, credits and catalogs, set up so performance and cost stay under your control. Platform names and defaults checked 24 Oct 2026.
05principles
Virtual warehouses and Gen2
Snowflake now defaults new standard warehouses to Gen2 where the region supports it. We size and test on Gen2 rather than assuming older sizing rules hold.
Per-second billing
Credits bill per second with a 60-second minimum per start, so we tune auto-suspend and group small queries instead of waking warehouses all day.
Resource monitors and budgets
Credit quotas that notify or suspend warehouses at the limits you set, plus query tags so every cost has an owner.
Horizon Catalog and Iceberg
Snowflake Open Catalog is closed to new customers, who are pointed to Horizon Catalog for Iceberg tables. We plan the catalog before the tables.
Snowpark, Cortex and CoWork
Python in Snowpark, Cortex AI functions and Snowflake CoWork (formerly Snowflake Intelligence) used where they help, on governed data.
FIG 09 · Databricks
Databricks consulting
SQL warehouses, Unity Catalog and open table formats on one lakehouse, for teams that mix BI, data science and streaming.
05principles
SQL warehouses: serverless vs pro or classic
Serverless typically starts in 2 to 6 seconds, pro and classic in about 4 minutes (Databricks docs, 11 Sep 2026). We pick per workload and set auto-stop to match.
Unity Catalog
One place for permissions, lineage and audit across tables, files, models and dashboards.
Delta Lake and Apache Iceberg
Delta by default, Iceberg where other engines must read the same tables, with the catalog chosen to limit lock-in.
Lakebase for operational data
Databricks announced Lakebase, its managed Postgres, as generally available on 3 Feb 2026. We use it when apps need low-latency reads next to the lakehouse.
Photon and Lakeflow
Photon for faster SQL and Lakeflow for ingestion and pipelines, where they beat a separate tool.
FIG 10 · Costs
Keeping warehouse costs predictable
We don’t quote platform prices here: list prices change and depend on your contract. We control the things that drive the bill.
05principles
Auto-suspend and auto-stop
Idle warehouses shut down quickly, with timeouts set from your real query patterns.
Right-sized compute
Separate warehouses for loading, BI and ad hoc work, each sized to its job and scaled out only when queues form.
Query tags and owners
Every query tagged by team, tool or job, so you can see who spends what and why.
Budgets and alerts
Limits and alerts in the platform itself, sent to the people who can act on them.
Across your whole cloud
For compute, storage and commitments beyond the warehouse, our cloud cost optimization team runs the full FinOps review.
Cloud cost optimization
FIG 11 · AI
Ready for AI
AI tools are only as good as the data they read. A governed warehouse is the safest source for them.
04principles
Semantic layer
Metric definitions in dbt or the platform’s semantic layer, so people and AI tools get the same answer to the same question.
RAG over warehouse data
Assistants that query governed tables and cite them, built with our LLM integration team.
LLM integrationPermissions that follow the data
Row-level security and masking apply to AI queries too, so an assistant can’t show what a user couldn’t see.
Built by the team behind our flagship
Kenerate AI, our flagship build, designed and engineered by Elgrith. It is an AI image and video studio, not a data warehouse project.
See Kenerate AI
FIG 12 · Measurement
What we measure in a data warehouse project
Trusted numbers, fresh data and a bill that tracks usage. That is what we report, sprint by sprint.
| How we measure it | What we won’t do | |
|---|---|---|
| Accuracy | Reconciliation of warehouse totals against source systems and finance reports | Switch off the old reports before the numbers match |
| Freshness | Load times and data age against the targets in your written scope | Promise real time where a daily load does the job |
| Model quality | dbt tests passing, documented models and one definition per metric | Ship untested models into dashboards |
| Cost | Compute and storage use by warehouse, team and query tag, month on month | Quote a savings figure we can’t trace to your bill |
| Adoption | Dashboards and queries running on the warehouse instead of spreadsheets | Leave a platform only we know how to run |
We agree the measures, targets and reporting rhythm in your written scope.
FIG 13 · Process · 6 steps
Assess first, then build in slices
Free call, assessment, fixed written quote, then subject areas shown working in weekly demos, with reconciliation before every cut-over.
- Step 01
Free call
Tell us what you report on today, where the data lives and what keeps breaking. We ask about volumes, freshness, users and compliance first.
- Step 02
Discovery
A warehouse assessment: sources, data volumes, query patterns, access rules and current costs, plus a platform shortlist tested on your real queries.
- Step 03
Fixed written quote
A written scope and a fixed quote, with the timeline, access and hand-over agreed before any work starts. You own the code and IP, and your warehouse account, dbt project and repository stay in your name.
- Step 04
Sprints with weekly demos
Subject areas built one at a time, ingestion to dbt models to marts, each shown working with real data in a weekly demo.
- Step 05
Launch
Parallel run against the old reports, reconciliation sign-off, then cut-over with a rollback plan.
- Step 06
Support
Tuning, new subject areas or a hand-over to your team, as agreed in your written scope.
FIG 14 · What you get
Everything included
Written into your fixed quote, so you know exactly what you are paying for before work starts.
11items
- Written scope and fixed quote
- Assessment report and platform recommendation
- Proof-of-concept on your real queries
- Architecture and data model diagrams
- Ingestion and change data capture set-up
- Tested, documented dbt project
- Roles, row-level security and masking
- Cost guardrails: auto-suspend, budgets, alerts, query tags
- Reconciliation reports for every cut-over
- Runbooks and hand-over documentation
- Your warehouse account, dbt project and repository stay in your name
FIG 15 · Tech & tools
The tools behind the work
Proven, well-supported tools, chosen for your project and set up in accounts your business owns.
- Snowflake
- Databricks
- BigQuery
- Amazon Redshift
- Microsoft Fabric
- dbt
- Apache Iceberg
- Delta Lake
- Airflow
- Dagster
- Fivetran
- Airbyte
- Python
- SQL
- Terraform
FIG 16 · Who it’s for
Data warehouses for teams that need numbers they can trust.
Same method everywhere: assess first, model in dbt, reconcile before cut-over.
4 audiences
Finance teams
Month-end numbers without the spreadsheet relay.
Revenue, margin and cash reports built on modeled data from your ERP, CRM and billing tools, reconciled to the ledger.
- One definition per metric
- Reconciled to source systems
- Audit trail for every change
- Access by role
Data teams
A platform your engineers and analysts can run.
A clean architecture, tested dbt models and cost guardrails, so your team ships new models instead of firefighting loads.
- dbt in your repository
- Orchestration with alerts
- Cost by team and query
- Code reviews and training
Scale-ups
From product database to first real warehouse.
Move reporting off your production database before it slows the app, on a platform sized for the next few years, not the next decade.
- Right-sized from day one
- Product and revenue metrics
- Ready for AI features
- No lock-in by default
UK teams
Data warehouse consultancy for UK organisations.
UK engagements run through our London client-partnership centre, with delivery from our New Delhi HQ. Models, documentation and reports use UK English, data can stay in UK or EU cloud regions, and working hours are agreed in your written scope.
- London client partner
- UK or EU cloud regions
- Fixed written quote in GBP
- Your data stays yours
FIG 17 · Working with us from the USA and UK
Built in New Delhi. Run on your hours.
Every project gets a point of contact in New York or London and an engineering team in New Delhi. Here is how today’s working days line up, live.
New York
North America
--:--ET
New Delhi team Your working day- Shared working time
- ≈ 3 hours today
- New Delhi is
- 10.5 h ahead
- Live calls, your time
- 9 a.m. to 12 p.m. ET
London
UK & Europe
--:--UK
New Delhi team Your working day- Shared working time
- ≈ 4.5 hours today
- New Delhi is
- 5.5 h ahead
- Live calls, your time
- 9 a.m. to 1:30 p.m. UK
Three ways to work together
- 01
Fixed-scope project
A written scope, one fixed price and a dated plan. Suits websites, apps and MVPs.
- 02
Dedicated team
Designers and engineers on your roadmap every month, billed monthly.
- 03
White label
For US and UK agencies: our work, your brand, unbranded deliverables.
Every engagement
- Fixed quote in USD, GBP or INR after a free 30-minute call
- NDA before discovery. You own all code, designs and content
- A working demo every sprint, not status slides
- Reply within one business day, calls inside your hours
Book a call
Calls on ET or UK hours
FIG 18 · Platforms
BigQuery and other platforms: where each fits
BigQuery and other platforms we build on. No prices here: each vendor’s pricing depends on region, edition and contract, and we model it on your usage in the assessment.
| Platform | How you pay for compute | Strongest fit | Watch out for |
|---|---|---|---|
| Snowflake | Credits for running virtual warehouses, billed per second | SQL-first teams that want separate compute per workload | Warehouses left running and the cloud services allowance |
| Databricks | DBUs for SQL warehouses and clusters | Mixed BI, data science and streaming on a lakehouse | Classic warehouses with slow start-up left idle |
| BigQuery | On-demand per query or capacity editions with slots | Teams already on Google Cloud and Google Analytics exports | Unbounded on-demand queries on large tables |
| Amazon Redshift | Provisioned clusters or Redshift Serverless | AWS-heavy estates that want to stay close to S3 | Cluster sizing and concurrency planning |
| Microsoft Fabric | Capacity shared across Fabric workloads | Microsoft 365 and Power BI-centred organizations | One capacity shared by many workloads |
FIG 19 · Rules and dates
Data warehouse platform changes that affect your design (checked 24 Oct 2026)
Platform defaults and names move fast. This is what changed and when, from each project’s own documentation, checked 24 Oct 2026.
| What changed | When | Source | What it means for you |
|---|---|---|---|
| Snowflake defaults new standard warehouses to Gen2 | In effect (checked 24 Oct 2026) | Snowflake docs, Gen2 standard warehouses | Size and benchmark on Gen2; a few regions still lack it |
| Snowflake bills warehouses per second with a 60-second minimum; cloud services only above 10% of daily warehouse use | In effect (checked 24 Oct 2026) | Snowflake docs, Understanding compute cost | Tune auto-suspend and batch small queries |
| Snowflake Open Catalog closed to new customers; Horizon Catalog recommended for Iceberg | In effect (checked 24 Oct 2026) | Snowflake docs, Open Catalog overview | New Iceberg projects plan around Horizon Catalog |
| Snowflake Intelligence renamed Snowflake CoWork | June 2026 | Snowflake CoWork product page | Older guides use the old name |
| Apache Polaris becomes an Apache top-level project | 19 Feb 2026 | Apache Polaris project blog | An open Iceberg catalog option outside any single vendor |
| Databricks Lakebase generally available (AWS first) | 3 Feb 2026 | Databricks blog | Managed Postgres next to the lakehouse for app data |
| Databricks serverless SQL starts in about 2 to 6 s; pro and classic about 4 min | Docs updated 11 Sep 2026 | Databricks docs, SQL warehouse types | Serverless allows tighter auto-stop |
Sources: Snowflake documentation and product pages, Apache Polaris project blog, Databricks blog and documentation, all checked 24 Oct 2026. Next check: December 2026.
FIG 20 · Compare
Consultant-led build vs in-house only vs lift-and-shift copy
| Consultant-led build with hand-over | In-house only | Lift-and-shift copy | |
|---|---|---|---|
| Time to trusted numbers | Subject area by subject area, reconciled as you go | Depends on spare capacity in your team | Fast to copy, slow to trust |
| Data model | Designed for how you report, tested in dbt | Often grows report by report | Old schema and old problems moved as-is |
| Cost control | Guardrails set from day one | Added after the first surprise bill | Rarely planned |
| Who runs it after | Your team, with runbooks and training | Your team | Whoever copied it |
FAQ
Frequently asked questions
Straight answers about Data Warehouse Consulting with Elgrith.
A data warehouse consultant assesses your data and reporting needs, picks the platform, designs the data model, sets up ingestion, dbt models, access and cost controls, and trains your team to run it. The goal is one trusted source for reports and AI tools, owned by you.
FIG 21 · Related
Keep exploring
Data Engineering
Pipelines, warehouses & AI-ready dataData Visualization
Dashboards, Power BI & TableauLLM Integration
ChatGPT, Claude & RAG in your productCloud Migration
Move to AWS, Azure or Google CloudCloud Cost Optimization
FinOps, rightsizing & commitment planningERP Development
Custom ERP & ERP integrations
Free 30-minute call
Tell us where your numbers live. We’ll design the warehouse.
Share your sources, reports and current platform. After a free call you get a written assessment scope and a fixed quote.
Pick a day for your call
OpenPick a day. You'll choose a time in your own time zone on the next step.