Aditum Data: connect your industrial data in minutes

Q: Does it replace our historian or data platform?

No. Aditum reads your historians and other systems read-only and delivers prepared data into the platforms you already run, such as Snowflake, Databricks, Azure Data Lake or PostgreSQL. It can also do the whole job itself, from source to live dashboards, P&IDs and 3D models. What is different is how it handles many sites: each KPI is defined once and every site is mapped to it, so a new site is a mapping, not a new project, and every site's figure is calculated the same way. That gives you speed (a new source in minutes), ease (set up without code), traceability (each number can be traced to its source) and reach (results land in your own tools or on dashboards in Aditum).

Q: We already do this in Python and Databricks. Why change?

Keep what works. The time goes into the repetition: per source, per site, and again every time a tag is renamed. Aditum takes over that repeated work without code and delivers into Databricks or Snowflake, and bespoke steps can still run as sandboxed Python.

Q: Why do pilots stall at one site?

Because the logic is tied to one site's tag names, units and meters, so it gets rebuilt for the next. In Aditum, the KPI is defined once, with an owner and a version, and the next site maps its own tags to it, with any exceptions approved and recorded.

Q: What do we install, and what does IT have to open?

One small read-only Windows or Linux service per source, under its own low-privilege account, opening outbound HTTPS on port 443 only. There is no inbound port, and nothing is written back to the plant. It is designed along the zones-and-conduits principles of ISA/IEC 62443.

Q: What does the AI see, and does it calculate anything?

By default it sees only metadata, such as table structure, column statistics, tag names and units. It sees data rows only if you opt in, and never document text. AI drafts pipelines and dashboards for your approval, the platform calculates every number, and your admin can switch AI off or keep all requests on your own Bedrock or Azure contract.

Q: Where is our data hosted?

In the EU, with the application in Frankfurt and the database in the Netherlands. The platform can also run in your own Azure as a project; AWS and Google Cloud are on request. Edge data can stay on site in local mode, but then it doesn't reach Aditum's dashboards.

Q: How do we know a number is right?

Saved pipelines run without AI, so the same input always gives the same output. You can check any dashboard figure against your own SQL reference, and the result is recorded. Each KPI result also states which exclusion rules it applied.

Q: What does it take to get started?

A 45-minute working session to choose one KPI and three sites. Share a few anonymised tags, and we build the demonstration on your own data. Contact sales@aditumdata.com.

Core capabilities

Everything between the plant and a comparable KPI.

It is not another database. Historians keep the raw time series, a lakehouse, where you have one, keeps storage and compute, and analytics tools keep the analysis. Aditum Data is the governed path between them.

01ConnectFrom the plant to a clean pipeline

01

Industrial-native connectors

Connectors for the protocols and systems already in your plants.

OPC UAModbus TCPMQTTHistoriansMES · LIMS · ERP

5 to 7.5 min per new source

02

Edge extractors

A small Windows or Linux service at each site, owned by your team.

Outbound onlyNo inbound portsBuffers & replays

Typically installed in a day

03

No-code pipelines

Merge, filter, transform and map with explicit rules, previewed on sample data before they run.

No-codeSQLPython, if you want it

Versioned · one-step rollback

02GovernOne basis, rules you can read

04

One KPI definition

Define a KPI once. Each site maps its own tags and records any approved exception.

Units & basisOperating statesSite exceptions

Same number at every site

05

Quality travels with the data

Quality flags stay with every value. Trips, start-ups and frozen sensors are excluded by explicit rules.

Late valuesFrozen sensorsTrips & start-ups

Every exclusion explained

06

AI-assisted, never AI-calculated

AI proposes mappings and dashboards. You preview and approve. It never calculates a published number.

Real tags onlyYou approveCan be switched off

No AI in published numbers

03DeliverTo your people and your platform

07

Live dashboards

Anyone authorised asks in plain language and gets a live dashboard.

A KPIA comparisonA monitor

Under a minute · secure link

08

Your platform, your format

Prepared data lands where you choose, in open formats you own.

DatabricksSnowflakeBigQueryPostgreSQLInfluxDBAVEVA PI

No proprietary format

09

Secure by design

Designed to the zones-and-conduits principles of ISA/IEC 62443. Managed in Frankfurt (EU).

Company SSOEncrypted credentialsJSON export

No inbound port into the plant

From the OT team

Onboarding an asset, then owning the pipelines.

Jasem Elayeb is OT Architect and Product Owner for Cognite Data Fusion at Wintershall Dea. He describes how the team shortened new-asset onboarding and kept pipeline ownership in house.

“We have reduced all this time from one and a half months to five minutes. You tell them, right after the meeting: your data is there.”Jasem Elayeb, OT Architect and Product Owner, Cognite Data Fusion, Wintershall Dea

Request the white paper Book a technical demo

Jasem ElayebOT Architect and Product Owner, Cognite Data Fusion · Wintershall Dea

How it works

From historian to dashboard in under ten minutes.

“Which of our units uses the least steam per tonne of product, measured on the same basis?”This is the path that answers it, once the extractor is installed and access is approved.

Example for

01At each siteYour OT network

PIAspenTechHistorian

800xASiemensPCS 7 · WinCC

ProficyKepwareOPC UAOPC UA Modbus

MQTT

Joined withMESLIMSERPFiles

Batch, recipe, grade and lab results enrich the OT time series

Edge extractorWindows or Linux · buffers and replays

Outbound

TLS 443
outbound only

02Aditum Data5 to 7.5 min to configure

One engineering basisManaged in Frankfurt (EU)

Extracteach source
Mergebatch · lab
Filtertrips · frozen
Transformunits · basis
Maptags to KPI
Enrichasset · state

Versioned · audit log · one-step rollback

routed

03Your people and platformBehind company sign-in

Live dashboards

Site A

Site B

Site C

A KPI, a comparison or a monitor, in under a minute

Your platform

Databricks

Snowflake

InfluxDB

PostgreSQL

Open, native formats you own

5 to 7.5 min to configure a sourceonce to define a KPI< 1 min to a live dashboard

Jan KaluzaOT Architect, Data Hub and Governance · Wintershall Dea

Customer use case

Wintershall Dea: six weeks to five minutes, per asset.

“We created working use cases in five minutes. It really was five minutes.”Jan Kaluza, OT Architect, Data Hub and Governance, Wintershall Dea

Book a technical demo Talk to an industrial data architect

Proven at Wintershall Dea

6 weeks to 5 minto a working pipeline, per asset

2 peopleran every pipeline, across every asset

Tens of millionsof data points per day

Kept in placethe existing industrial data platform

Define once, use at every site

One KPI definition. Every site mapped to it.

One shared definition · version 1.3 · owner: group energy leadExample

Specific steam consumption, t steam per t product

Boundary: steam delivered across the unit boundary
Basis: saleable product, dry basis
States: running ≥ 30 min; start-up, shutdown, cleaning excluded
Quality: bad quality, frozen values and gaps excluded

Site AFI-2104.PV · t/h

Mass flow; product from the batch system, dry basis

No exception

Site BU30_STM_HDR_FLOW · m³/h

Volumetric meter; product from orders, wet basis

Exception: converted to mass; wet to dry via lab moisture

Site CFT-7710 · kg/s

Header shared with a second unit

Exception: steam allocated by metered fraction

One comparable figure per site

Tags and exceptions are illustrative.

Running in production

Six failure modes, handled explicitly.

These occur in every plant network. Left unnoticed, each one quietly changes a number that someone will use to make a decision. This is how Aditum Data handles each of them today.

Site network interruption

The edge connector buffers data in memory and on disk and replays it in order on reconnect. A very long outage can exceed the disk buffer, and a stale source opens an incident.

Source tag renamed

In scheduled pipelines, the run reports the missing tag with its likely new name, and the other sites keep writing. A tag that stops reporting opens an incident. The mapping is re-pointed once, and the change is versioned and recorded in the audit log.

Late-arriving or corrected value

Any stored pipeline version can be rerun for a past period, so the period is recalculated with the corrected value.

Frozen sensor or bad-quality reading

Named rules exclude frozen values, bad-quality readings, shutdowns and gaps from KPI and dashboard calculations, and each result states which rules it applied.

Edge connector stops sending

It is shown offline three minutes after its last heartbeat, and an alert opens within ten minutes, by email or in Teams, Slack or a webhook. Owners get one email per root cause, not one per symptom. A connector that sends heartbeats but no data shows as “Heartbeats only”, never as healthy.

Mapping change with an error

Every pipeline change is saved as a version, with one-step rollback, and recorded in the audit log. KPI definitions are versioned too.

How to compare

Build it, buy a platform, or add a layer?

Whichever approach you choose, the same questions show whether it will hold across a network. Ask them of any vendor, including us.

Question to ask any vendor

Aditum Data's answer

01 Speed

How long does adding one source, or one KPI at one more site, take, and was that measured or estimated?

5 to 7.5 minutes per source once access is in place. At Wintershall Dea, a working pipeline per asset went from six weeks to five minutes. In a fixed-scope test, configuration time per source and per site is measured on your own sites.

02 Calculations

Where are calculations run? Is any published number produced by a language model?

In standard numerical libraries, never by a language model. Every result references its source tags and definition version.

03 Exceptions

Show us a site exception, such as a wet basis or a shared header, defined, approved and traceable.

Each site maps its own tags to one shared definition and records any approved exception: wet to dry via lab moisture, or steam allocated by metered fraction.

04 Corrections

What happens to a late, corrected or frozen value?

Any stored pipeline version can be rerun for a past period, so the period is recalculated with the corrected value. Named rules exclude frozen values, bad-quality readings, shutdowns and gaps, and each result states which rules it applied.

05 Access

Can people outside the data team get an answer themselves, and how long does it take?

Yes. An authorised user asks in plain language and has a live dashboard in under a minute, at a secure link behind company sign-in.

06 Exit

On exit, what format are our data and logic in, and where do they already live?

Your data already lives in your own destination, in open, native formats. Definitions and mappings are stored as JSON and can be retrieved through the API.

07 Fit

Does this replace what we already run, or feed it?

It feeds it. Aditum Data does not replace your historian, lakehouse or analytics tools. Engineers keep their analysis tools, now fed with comparable data from every site.

Aditum Data removes the plumbing, not the people. Engineers keep their analysis tools and data teams keep their platform. Both receive prepared, comparable data in minutes rather than weeks.

Technical evaluation

See one of your KPIs prepared the way your sites actually work.

In 45 minutes with your OT and analytics teams we take one KPI across three sites, show the mappings and transformation steps on an anonymised sample, and agree the acceptance criteria before any plant access.

Inspect the pipeline, not a slide deck
Compare how the same definition maps at each site
Decide with criteria your engineers wrote

Book a technical demo

FAQ

Questions from OT and data teams

Does it replace our historian or data platform?

No. Aditum reads your historians and other systems read-only and delivers prepared data into the platforms you already run, such as Snowflake, Databricks, Azure Data Lake or PostgreSQL. It can also do the whole job itself, from source to live dashboards, P&IDs and 3D models. What is different is how it handles many sites: each KPI is defined once and every site is mapped to it, so a new site is a mapping, not a new project, and every site's figure is calculated the same way. That gives you speed (a new source in minutes), ease (set up without code), traceability (each number can be traced to its source) and reach (results land in your own tools or on dashboards in Aditum).

We already do this in Python and Databricks. Why change?

Keep what works. The time goes into the repetition: per source, per site, and again every time a tag is renamed. Aditum takes over that repeated work without code and delivers into Databricks or Snowflake, and bespoke steps can still run as sandboxed Python.

Why do pilots stall at one site?

Because the logic is tied to one site's tag names, units and meters, so it gets rebuilt for the next. In Aditum, the KPI is defined once, with an owner and a version, and the next site maps its own tags to it, with any exceptions approved and recorded.

What do we install, and what does IT have to open?

One small read-only Windows or Linux service per source, under its own low-privilege account, opening outbound HTTPS on port 443 only. There is no inbound port, and nothing is written back to the plant. It is designed along the zones-and-conduits principles of ISA/IEC 62443.

What does the AI see, and does it calculate anything?

By default it sees only metadata, such as table structure, column statistics, tag names and units. It sees data rows only if you opt in, and never document text. AI drafts pipelines and dashboards for your approval, the platform calculates every number, and your admin can switch AI off or keep all requests on your own Bedrock or Azure contract.

Where is our data hosted?

In the EU, with the application in Frankfurt and the database in the Netherlands. The platform can also run in your own Azure as a project; AWS and Google Cloud are on request. Edge data can stay on site in local mode, but then it doesn't reach Aditum's dashboards.

How do we know a number is right?

Saved pipelines run without AI, so the same input always gives the same output. You can check any dashboard figure against your own SQL reference, and the result is recorded. Each KPI result also states which exclusion rules it applied.

What does it take to get started?

A 45-minute working session to choose one KPI and three sites. Share a few anonymised tags, and we build the demonstration on your own data. Contact sales@aditumdata.com.

Get in touch

Discuss architecture with an industrial data specialist.

Source compatibility, security boundaries, deployment and multi-site operation. Or request a 45-minute technical session to select one KPI and three sites and draft the acceptance criteria.

Salessales@aditumdata.com