- Market
- Saudi Arabia
- Written for
- CTO, CIO
- Decision
- Adopt Snowflake as the analytical warehouse with its economics understood
Information security for software delivery

ISO/IEC 27001
ISO/IEC 27001 certified
The economics
A warehouse billed by the second changes how you should run it
We do not have an office in Saudi Arabia
We serve organisations there from our Riyadh office and our engineering hubs in Lahore and Karachi, travelling for the parts of an engagement that need to happen in person.
On a traditional warehouse you bought capacity and your habits did not matter much; the box cost the same busy or idle. Snowflake inverts that. Storage is billed like cloud storage — cheap, almost never the issue — while compute is billed in credits per second that a virtual warehouse runs. That inversion is the platform’s appeal: analytics teams get elastic compute without capacity planning, and a heavy month costs more than a quiet one instead of forcing a hardware purchase. It is also the trap, because every default — warehouse size, auto-suspend timing, how many warehouses exist — is a spending decision someone made once, usually early, and rarely revisits.
Run well, the model is genuinely elegant: a warehouse per workload class — loading, transformation, BI, ad hoc — each sized for its work, suspending within a minute of going idle, resuming in seconds when needed. Concurrency is solved by scaling out briefly rather than sizing up permanently. Run by default, the same model produces the familiar pathology: one oversized warehouse serving everything, kept warm all day by a trickle of queries, its bill explained by nobody. The platform does not impose the discipline; the implementation has to.
The capability most buyers underestimate is sharing. A Snowflake account can grant another party governed, read-only access to live data — no files exported, no copies drifting out of date, access revocable in one statement. For a group with subsidiaries, a company reporting to a parent, or a business monetising data to partners, that alone can justify the platform: the nightly CSV hand-off, with its failures and its staleness, simply stops existing.
The choice
Snowflake, a lakehouse, or the warehouse you already run — judged by workload
We build on Snowflake and on Databricks, which is exactly why we have no platform to defend. The honest comparison is by workload mix, and our data warehouse page covers the modelling discipline all three options share.
| Option | Primary workload | Team it assumes | Cost behaviour | Where it wins outright | Best for |
|---|---|---|---|---|---|
| Snowflake | SQL analytics and ELT on structured and semi-structured data | SQL-first analysts and analytics engineers | Per-second credits — excellent under discipline, drifts without it | Governed live sharing across organisations, near-zero administration | The organisation whose data work is analytics, and whose sharing needs are real. |
| A Databricks lakehouse | Mixed engineering, ML and streaming alongside SQL analytics | Data engineers comfortable with code and notebooks | Compute-time billing — controllable with policies, unforgiving without | One platform where ML trains on the same governed data analysts query | The organisation whose workload mix genuinely spans engineering and ML — see our Databricks page. |
| The warehouse you already run | Whatever it serves today, reliably | The team that already operates it | Known, budgeted and already paid for in skills | No migration risk, no re-training, no new bill to govern | The organisation whose current platform meets its needs — moving has to be earned by a concrete limit, not by fashion. |
Credit discipline
The five habits that decide what a month of Snowflake credits costs
Snowflake spend is behavioural. Each of these is set once by an engineer and paid for monthly by finance, which is why we put numbers and owners on all of them.
| Variable | Effect on price | Why |
|---|---|---|
| Warehouse size versus query need | Each size step doubles the burn rate | A size chosen for the heaviest monthly job burns double credits on every routine query all month. Splitting workload classes onto right-sized warehouses is the single highest-value adjustment. |
| Suspension timing | Idle minutes are billed minutes | A warehouse that lingers ten minutes after each query bills those minutes forever. Suspension tuned per workload — aggressive for ad hoc, looser where cache warmth pays — recovers the difference. |
| Queries that scan too much | The same answer at several times the cost | Poor clustering, missing filters and select-star habits make Snowflake read data it immediately discards. Query-profile review is a cost tool on this platform, not only a performance one. |
| Serverless features left on | Background credits without a face | Search optimisation, materialised views and automatic clustering each earn their keep on the right table and quietly bill on the wrong one. Each is enabled per table, with the benefit measured. |
| Copies where shares belong | Storage, pipelines and drift, all avoidable | Duplicating datasets between accounts and teams pays three times — storing the copy, running the pipeline, reconciling the drift. Snowflake’s native sharing exists to make all three unnecessary. |
Ask any supplier quoting this
- How many warehouses will exist, sized how, and what workload class does each serve?
- What will the credit consumption report show per team and per pipeline, and who reviews it?
Our role
What we engineer on Snowflake and what your analytics team runs from there
- 01
The fit, judged before the build
We assess your workload mix against Snowflake, a lakehouse and your current platform, and give you the recommendation with its reasons — including when it is not Snowflake.
- 02
The estate, built with its meters on
Warehouse topology, resource monitors, per-team tagging and the ELT pipelines — engineered so consumption is attributable from the first month, not reconstructed after a surprise.
- 03
The sharing, replacing the exports
Where data moves between your entities or to partners, we replace file hand-offs with governed shares — designed with the access boundaries your agreements actually require.
Your analytics team runs the estate day to day; we leave the warehouse topology documented, the monitors alerting, and a quarterly review habit that keeps the credit line explained.
Common questions
What buyers ask
Why did our Snowflake bill jump when our data barely grew?+
Because Snowflake bills compute, not data — the bill tracks how warehouses ran, and data volume is usually a minor term. The usual culprits: a warehouse resized upward for one heavy job and never resized back, a new dashboard whose auto-refresh keeps a warehouse from ever suspending, a serverless feature enabled on a large table, or an ELT schedule that started running hourly. The account usage views hold the answer — credit consumption per warehouse per day, laid against the change dates — and part of a well-built estate is a report that makes this a five-minute question rather than a forensic project.
Can we give a partner access to our data without sending them files?+
Yes — this is one of the strongest reasons to choose Snowflake. A secure share grants another account read access to specific tables or views: the partner queries live data in place, nothing is copied, and revoking access is one statement. For a partner without their own Snowflake account, a reader account gives them a governed window you control. Compared with the nightly file export, everything improves at once — no pipeline to fail, no staleness, no copies of your data sitting in someone else’s storage, and a clean audit line on what was shared with whom. What remains yours to govern is the contractual side: which fields, which partners, and under what agreement.
Should we choose Snowflake or Databricks if we want both analytics and machine learning?+
Weigh where the centre of gravity sits. If the organisation’s daily bread is SQL analytics and the ML ambition is early or modest, Snowflake keeps the analytics excellent and can host lighter ML workloads, with the option of adding a dedicated ML platform later. If ML and heavy data engineering are core to the operating model — models in production, feature pipelines, streaming — the lakehouse’s case strengthens, because training on the same governed data the analysts use removes a whole class of duplication. Both vendors are expanding into each other’s territory, so the durable question is not which platform can do everything, but which one your team will run well for the work you actually have. We implement both, and our recommendation comes with the workload analysis that produced it.
In this cluster
Size your Snowflake workloads with us
A workload assessment that maps your queries to right-sized warehouses and states, with reasons, whether Snowflake or a lakehouse fits your mix.
Start the conversation