Data Engine

Why your dashboards are true, and never stale.

Anything can draw a chart. The question is whether you'd forward it to your CFO. This is the engine underneath every Mammoth dashboard. Teams have run production pipelines on it since 2017.

Try free Talk to an engineer

1B+
rows on a single canvas: 100+
tested source connectors: Hourly
to event-driven refresh: 1–7 yrs
activity-log retention

The premise

Dashboards don't fail because the chart was wrong.

They fail in three specific ways, and none of them is a visualisation problem. Each one is a pipeline problem.

It went stale

Right the day it was produced. Nothing was watching the source, so nobody noticed when it stopped being right.

Fixed by scheduled refresh

Nobody can reproduce it

The cleaning happened by hand in someone's spreadsheet. The figure can't be defended because it can't be re-derived.

Fixed by recorded pipelines

Three versions exist

Three people built the same metric three ways. All three numbers are circulating in different decks.

Fixed by shared pipelines

How it works

Connect, prepare, automate, govern.

Four steps, one canvas. The handoffs between separate tools were where the time and the errors used to live.

Connect

Read the source. Don't export it.

Point Mammoth at the warehouse, the database, the SaaS app or the folder. No stale copy in a downloads folder becomes the version of truth.

100+ tested connectors, from CSV to Snowflake
SFTP drops and webhooks for supplier and partner feeds
Custom connectors built by AI and validated against live data
Each source counts as one connection, however many tables

PipelineRe-runs daily

1orders.csv4.2M rows
2Drop duplicates61 removed
3Fill blanks: region2,140 filled
4Standardise dates3 formats → ISO
5Join store_masteron store_id
6Weekly Salespublished

Prepare

The messy part, recorded once

Blanks, duplicates, five spellings of one region, three date formats, supplier codes that don't match. Resolved in a pipeline you can read, re-run on every new batch.

No-code steps anyone on the team can follow
Raw SQL for anything easier to express that way
AI profiles the data and proposes the obvious fixes
Row counts shown at every step, so nothing vanishes quietly

Raw SQL stepOptional

select
  region,
  date_trunc('week', order_date) as wk,
  sum(net_amount)              as revenue,
  count(distinct order_id)     as orders
from orders
where order_date >= current_date - 84
group by 1, 2
order by 2, 1

Drops into the same pipeline as the no-code steps, and re-runs on the same schedule.

Automate

It keeps itself current

Every dashboard sits on a schedule or an event trigger. The source changes, the pipeline re-runs, the numbers move. Nobody rebuilds anything on Monday morning.

Daily scheduling on Starter, hourly on Pro
Event triggers on Pro and Enterprise, so a run starts when a file lands
Export destinations to SFTP or a database
Failures surface as alerts, not as silently wrong numbers

RefreshAutomated

Weekly SalesDaily · 06:00Ran 4m ago
Store MarginHourlyRan 12m ago
Close PackOn file arrivalWaiting
Cash ForecastDaily · 07:00Ran 1h ago

Govern

Trust every number

Role-based access, row-level security, draft mode before anything goes out, and restorable checkpoints. Every change lands in an activity log with retention you configure.

Custom roles and connector-level governance on Pro
Row-Level Security, so one dashboard is correctly scoped per viewer
Draft mode so viewers never see work in progress
Checkpoints you can roll back to, with full version history

Version historyRestorable

v14 · livePublished 2 days ago
v13Margin calc corrected
v12Added ANZ region
v11Initial publish

Capabilities

What's in the engine

Every capability is on every plan, including Free. Plans differ on scale: storage, refresh frequency and how many people build.

No-code transforms

Filter, join, pivot, derive, deduplicate and standardise, as steps anyone on the team can read and re-run.

Raw SQL

Write SQL where SQL is easier. Same pipeline, same schedule.

AI suggestions

Mammoth profiles data on arrival and proposes the fixes it can see: mixed date formats, inconsistent categories, duplicates.

Auto-sync

Daily down to hourly, or triggered the moment a file lands.

Draft mode

Change a calculation and see it before anyone else does. Viewers never see a half-finished state.

Checkpoints

Every published state is versioned and restorable, so you can show what a figure was on the day it was reported.

Questions engineers ask first

Do I need a data warehouse?

No. Mammoth reads files, databases, SaaS apps and warehouses directly. If you have a warehouse, it connects to it; if you don't, CSVs and a production database are a perfectly normal starting point.

How much data can it handle?

Teams run 1B+ rows on a single canvas. Storage scales by plan: 1 GB on Free, 10 GB on Starter, 50 GB on Pro, and custom on Enterprise, with 10 GB expansion blocks on paid plans.

What happens when the source schema changes?

The pipeline flags the change rather than silently producing wrong numbers. You'll see which step is affected, and the previous checkpoint stays live until you resolve it.

Can I see exactly what was done to the data?

Yes. That's the design. Every transform is a listed step with the row counts it affected, and the activity log records who changed what and when, with retention from 7 days on Free up to 7 years on Enterprise.

Is this a separate product from the dashboards?

No. It's the same canvas. The dashboard is what the pipeline publishes, which is why the numbers on it can be trusted and why they update on their own.

Keep reading

ConnectorsCheck your sources are supported.SecurityAccess control, compliance and data handling.vs AlteryxThe closest comparison on preparation depth.

Point it at your messiest source.

The engine is the part that's hard to believe from a description. Connect something real and see what it does with it.

Try free Book a demo

21-day Pro trial
No credit card
Viewers always free