Why embedded analytics for connected equipment is different
A machine builder that ships dashboards on sensor data to its customers faces a problem that ordinary embedded BI does not. The answer for most such builders is self-hosted Apache Superset in front of ClickHouse, with per-customer row-level security and SSO embedding, provided the plant data must stay out of shared SaaS. Superset is licensed under Apache-2.0 and has no paid edition, so the software itself does not force a commercial decision, but the operating model still has to be designed: where the metadata database runs, how the caching and worker components are hosted, and who holds the keys. Each customer's plant is a separate tenant, and the row-level rules, not the portal's navigation, are what keep one plant's readings away from another.
Machine telemetry differs from sales or finance data in three ways. Volume is high and continuous, with readings from many sensors arriving every few seconds, so the analytics store must be built for time-series scans and retention, which is why ClickHouse is the usual choice. The data is operational: a spike in spindle temperature or a pressure drop can point to a fault on a production line, so dashboards and the systems behind them sit inside an operational technology context. Standards such as the ISA/IEC 62443 series address securing industrial automation and control systems, and NIST SP 800-82 Rev. 3 covers operational technology security. Ask your security team which zone and conduit model applies before a dashboard reads from any plant network, and confirm with your plant engineers which tags may leave the site at all.
Which regulations and constraints shape the build
This table maps each regulation or standard to what it means for an analytics build that serves customers from machine and sensor data. Read it as a list of scope inputs to confirm with your compliance and OT teams, not as a statement that any of these apply or that a design satisfies them.
| Requirement | What it asks for | What it means for the analytics build |
|---|---|---|
| IEC 62443 series (ISA/IEC 62443) | Security program, risk assessment, system and component requirements for IACS. Asset owners carry program duties (ANSI/ISA-62443-2-1-2024). Suppliers carry lifecycle and component duties (ANSI/ISA-62443-4-1-2018, ANSI/ISA-62443-4-2-2018). | Clarify whether you are the supplier, the asset owner or both. Keep the analytics platform outside control networks, and document zone and conduit boundaries with the OT team before any data flow is designed. |
| ANSI/ISA-62443-3-2-2020 risk assessment | Security risk assessment for system design, used to define zones and conduits. | Include the dashboard tier, the ClickHouse cluster and the portal integration in the risk assessment. Treat the path from plant historian to analytics store as a conduit that needs defined controls. |
| ISA-62443-3-3-2013 system requirements | System security requirements and target security levels for zones. | Confirm the target security level for the data path. Confirm how access to dashboards is authenticated and logged, since customer portal users sit outside the plant network. |
| NIS2 Annex II (manufacturing) | Lists manufacturing under NACE Rev. 2 divisions 26 to 30, including machinery, electrical equipment, and medical and in vitro diagnostic devices. | Confirm whether your company is an in-scope entity. If so, customer dashboards that expose operational data become part of the incident response and supply chain picture. |
| NIS2 incident reporting (Article 23(4)) | Significant incidents: early warning within 24 hours, incident notification within 72 hours, final report within one month. | Log dashboard access, export events and admin changes so that an incident involving the analytics stack can be reconstructed within these windows. Agree who triggers reporting before go-live. |
| NIS2 fines (Article 34(5)) | Important entities face fines of at least EUR 7 million or 1.4% of worldwide turnover. | Use these figures to justify funding for access logging, backups and restore testing, which a regulator will expect to see as evidence of controls. |
| Cyber Resilience Act (Regulation (EU) 2024/2847) | Products with digital elements must meet the regime from 11 December 2027. Article 14 reporting of actively exploited vulnerabilities and severe incidents applies from 11 September 2026 (Article 71(2)). | If the analytics components ship as part of a connected product or its companion portal, track each open-source component version and its vulnerability disclosures. Confirm with counsel whether the portal counts as part of the product. |
| CRA fines (Article 64) | Maximum of EUR 15 million or 2.5% of worldwide annual turnover for breaches of Annex I requirements and Articles 13 and 14, whichever is higher. | Maintain a software bill of materials and a vulnerability handling process for the Superset, ClickHouse and gateway components. Ask your counsel which obligations apply to your product line. |
| Machinery Regulation (EU) 2023/1230 | Applies from 14 January 2027. Annex III, 1.1.9 requires machinery to be protected so that connecting another or remote device does not create a hazardous situation. | Do not let the analytics platform write to, or accept commands from, machine controllers. Keep data flows one-way from machine to analytics where possible, and document this for the machinery technical file. |
| NIST SP 800-82 Rev. 3 (OT security) | Guide to OT security covering SCADA, DCS, PLCs and building systems, with OT performance, reliability and safety requirements. Supersedes Rev. 2 (September 2023). | Use it as the OT reference for how plant networks connect to IT and cloud analytics. Design the analytics tier so that query load and failures cannot affect control or safety functions. |
| Logging and access evidence | Assessors and customers expect records of who viewed or exported what data, and who changed roles or row-level rules. | Plan audit logs at the dashboard, database and identity layers from the start. Confirm retention periods with your compliance lead, since no retention figure applies to this build by default. |
What machine data flows into customer dashboards
The data path for a connected-product dashboard runs from the machine controllers to the customer's screen in distinct tiers, and the design decisions sit at the boundaries between them. PLCs and drives expose process values, SCADA systems poll and display them, and a historian such as the one in the plant stores them over time. A gateway or collector, typically a small agent inside the plant network, reads from a replica or from the historian's own interfaces, tags each sample with the machine, line and customer identifiers, and pushes batches into ClickHouse. Superset then reads ClickHouse through datasets whose SQL is scoped to the tenant. Confirm with your controls team which protocol and historian interfaces the gateway will use, since that choice determines what the collector can read without touching control logic.
Two properties of the data set the design more than any dashboard feature does. The first is volume: sensor streams from many machines at high sampling rates produce time-series tables that grow continuously, so the team must agree on sampling intervals, downsampling and how long raw data is kept against what customers actually query. The second is tagging: every row needs a reliable customer and asset key written at ingestion, because row-level security in Superset depends on those columns being present and correct. A machine that is resold or transferred between customers needs a documented re-keying procedure, or historical data will appear under the wrong account.
The analytics tier should never query the control-system historian directly. Dashboard traffic is bursty and unpredictable, and a long query or a poorly written dataset can compete with the plant's own reads, which is a reliability and safety question as well as a performance one. A separate ClickHouse cluster fed by the collector, or at minimum a dedicated replica, keeps customer load away from the process. Ask the plant engineers to name the systems that must never see analytics traffic, and write that boundary into the network design and the collector's read permissions before the first dashboard is built.
Which open-source platforms fit this job and why
A manufacturer's customer dashboard usually needs three layers: a time-series store that can absorb machine and sensor data at volume, a visualization layer that can be embedded in a service portal with per-customer access, and a collection layer at the edge. The cards below cover the platforms that fit each layer and the license or operational checks that decide whether they fit your product.
Apache Superset for the embedded dashboard layer
Apache Superset is licensed under Apache-2.0 and has no paid edition, so the license question is about how you ship it, not about a per-seat fee. It is built on Flask-AppBuilder and supports OAuth2 providers and LDAP, with AUTH_ROLES_MAPPING translating identity groups into Superset roles. Check how your service portal's identity provider maps customer accounts to those roles, and confirm which embedding and row-level security features your release version provides before you design per-customer views.
ClickHouse for high-volume time series
ClickHouse server is open source under Apache-2.0, and it is built for the write-heavy, range-query pattern of machine telemetry. Replicated tables are coordinated by ClickHouse Keeper, which replaces ZooKeeper, and the replication guide recommends at least three Keeper nodes on dedicated hosts in production. Plan for backups separately from replication, because replication does not protect against human error or corruption, and test restores on a spare cluster before you rely on them.
Metabase as a secondary option
Metabase is AGPL licensed outside its enterprise directory, which uses the Metabase Commercial License. If you modify Metabase and make it reachable by customers over a network, the AGPL obligations on your changes apply, and features in the enterprise directory need a commercial license. Confirm with counsel which embedding features you need and whether they sit in the open-source or commercial part before choosing it over Superset.
Node-RED for edge collection
Node-RED is licensed under Apache-2.0 and suits moving data from PLCs, gateways and historians into ClickHouse at the plant edge. Its role should stay narrow: collection, normalization and buffering, not customer-facing access. Confirm the protocols your equipment speaks and where flows, credentials and buffered data are stored, since those flows often sit close to operational technology.
n8n only for non-core collection tasks
n8n is a fair-code platform under the Sustainable Use License, and files marked .ee require an n8n Enterprise License. Its workflow model suits business-side tasks such as alerting or report delivery more than continuous machine data ingestion. If you use it, keep the enterprise files out of your build unless you hold that license, and review the Sustainable Use License terms before embedding its output in a product you distribute.
Where the analytics stack runs for each customer model
One ClickHouse cluster and one Superset instance run in the manufacturer's cloud account and serve every customer. Tenant separation is handled in the data model and the access layer: each machine or sensor record carries a tenant key, Superset row-level security filters queries by that key, and embedded sessions are issued per customer so one customer's session cannot reach another's rows. This model has the lowest number of systems to patch and back up, and it is the simplest to operate at the start.
- Isolation: Logical only. Confirm how row-level filters are applied to every dataset, including datasets built from custom SQL, and test that a query cannot bypass them.
- Upgrades: One upgrade path for all customers, which means one test cycle, but a bad release affects every tenant at once.
- Support load: Lowest per-customer effort, but noisy queries from one large fleet can slow dashboards for others unless resource limits are set.
- Backups and restores: A single backup target and restore procedure. Restore tests must cover tenant-level recovery, not only full-cluster recovery.
- Plant data: Telemetry from every plant leaves the plant and lands in the manufacturer's cloud, which may conflict with customer data-residency expectations.
Before committing, ask the implementation partner to show the row-level security tests and the tenant-level restore runbook.
Each customer that requires isolation receives its own ClickHouse and Superset deployment, usually in a separate cloud account, subscription or namespace that the manufacturer or the customer controls. Credentials, storage and network paths are separated by design, which simplifies audit answers about who could have accessed a given fleet. The cost is that every deployment is another system to upgrade, monitor, back up and support.
- Isolation: Physical or account-level. Easier to explain to security reviewers than row-level filtering alone.
- Upgrades: Each deployment needs its own change window and rollback plan, so a release calendar must be enforced across all of them.
- Support load: Grows with the number of customers. Standardized images, infrastructure as code and shared runbooks keep this manageable.
- Resource sizing: Each deployment needs its own ClickHouse Keeper nodes for replicated tables, so the minimum footprint per customer is higher.
- Commercial terms: Clarify who owns the hosting cost and who holds the operator role in the customer's cloud account.
Ask how many deployments the partner has operated in parallel and how upgrades are scheduled across them.
For air-gapped or tightly controlled plants, the analytics stack is installed inside the customer's network, and the manufacturer provides software, upgrades and remote support under agreed access rules. Telemetry stays on the plant network, and the manufacturer never holds the raw data unless the customer explicitly exports it. This model fits operational technology environments where network segmentation is governed by the plant's own security program, including the zones and conduits described in ISA/IEC 62443.
- Isolation: Strongest, because data and compute sit inside the customer's boundary.
- Upgrades: Delivered as offline packages with a verification step. Plan for change approvals and maintenance windows set by the plant, not by the vendor.
- Support load: Highest per site. Remote access must be brokered, logged and time-limited, and on-site hardware failures need a local response path.
- Backups: Stored on plant-controlled media or storage. Confirm who owns the backup schedule and restore tests.
- Connectivity: Confirm whether dashboards need external identity providers or license checks, since air-gapped sites cannot reach them.
Ask the partner how offline updates are verified and how remote support sessions are recorded for the plant's audit.
How a rollout moves from pilot to production
The sequence moves a manufacturer from a defined data model and tenant map to a production embedded analytics service, with each phase closing on a named output the next phase depends on. Nothing reaches a customer portal until isolation, restore and upgrade procedures have been tested.
Scope the data model and tenants
The equipment maker's product, data and security owners agree which machines, sensors, sites and customer accounts each dashboard may show, and how each customer maps to a tenant and to its identity provider groups. DeployOpen documents the entities, the tenant key carried on every row and the roles that may see aggregate versus raw data. The step closes with a signed data model and tenant register that every later phase references.
Design the OT boundary and collectors
The engineering and OT teams define how telemetry leaves the plant network, which collectors or gateways read from control systems, and which direction traffic is allowed to flow. The design is checked against the ISA/IEC 62443 zone and conduit model and the asset owner's existing segmentation, so the analytics platform never gains a path back into control networks. The output is a zone and conduit diagram with the collector placement, firewall rules and a list of the data fields that cross the boundary.
Build the ClickHouse schema and retention
The data team creates ReplicatedMergeTree tables for raw sensor readings and rollup tables for the aggregates customers usually query, with tenant identifiers in the sort and partition keys where they support the access pattern. Retention periods are set per table and written down with the reason for each, including any records that must be kept for quality or regulatory review. The output is a tested schema, a retention register and a load test on representative volumes, run in a pre-production environment with realistic data.
Deploy Superset on a production metadata database
The platform team installs Apache Superset with PostgreSQL or MySQL as the metadata database, which the Superset architecture documentation recommends for production, and does not use the SQLite default. A caching layer and Celery workers are added for alerts, async queries and thumbnails, and the SECRET_KEY is set and stored in the organization's secrets manager. The step closes when the stack starts cleanly in staging, the metadata database version is on the supported list and the configuration is under version control.
Test SSO, guest tokens and row-level security
Engineers connect the customer portal's identity provider, map groups to Superset roles, and configure embedding with guest tokens or SSO as the product design requires. They then test row-level security with a pilot customer's real account structure, using scripted queries that try to read another tenant's machines, sites and raw readings. Pilot customers and the security lead sign off on the test results. The output is a row-level security test report with each failing case closed and a record of who approved the pilot access.
Prove backup and restore on a spare cluster
The operations team takes a ClickHouse BACKUP of the telemetry tables and a backup of the Superset metadata database, then restores both to a spare environment. They confirm row counts, tenant isolation and dashboard behavior after restore, and time the restore against the recovery expectations agreed with the business. Replication alone is not treated as backup, since it does not protect against human error or corruption. The output is a dated restore test record with the runbook that was used, and any gaps logged for remediation before go-live.
Go live with a documented upgrade process
The production cutover moves the pilot customers and then the wider customer base into the portal, with monitoring on query load, ingestion lag and failed logins. The team agrees how Superset and ClickHouse upgrades are staged, including a metadata database backup, the Superset upgrade steps, a check of UPDATING.md for breaking changes and a ClickHouse changelog review, with the ClickHouse security policy used to confirm the release line is still supported. The step closes with a go-live approval, an on-call runbook and a scheduled first upgrade test in staging.
What auditors and customers ask to see
Who can see which tenant's data, and how do we prove it?
Access evidence comes from the access layer, not the dashboard tool alone. Superset's authentication setup (Flask-AppBuilder with OAuth2 or LDAP and AUTH_ROLES_MAPPING) determines who logs in and with which role, and the metadata database records dashboard and query activity that you can export. ClickHouse's own logs and its access control definitions (users, roles and row policies) show which account queried which table. Combine these into a log that maps each viewer to a tenant identifier and a timestamp. Confirm with your security team which log retention and export mechanisms apply in your environment, because the exact fields depend on your deployment and configuration.
What row-level security test evidence should we keep?
Keep a repeatable test suite that logs in as a user from each tenant and attempts to read every other tenant's rows through dashboards, charts, SQL Lab and direct API calls. Store the test inputs, expected results, actual results and the commit or configuration version tested. Auditors look for negative tests, not just a demonstration that a tenant sees its own data. Run the suite after every Superset upgrade and every change to role mapping or dataset definitions. DeployOpen can plan and implement security testing, including application and API testing, but the evidence must be retained by your team and reviewed by your assessor.
How do we show ClickHouse backups actually restore?
Keep backup job records and restore test records as separate artifacts. ClickHouse BACKUP and RESTORE can write full and incremental backups of tables, partitions or databases to local disk, S3-compatible storage or Azure Blob Storage. The backup documentation notes that replication does not protect against human error or corruption, so backups are a separate control. The documentation also warns that a backup never test-restored will probably not restore when needed. Record the backup identifier, storage location, date, the restore target (ideally a spare cluster), the row counts or checksums compared, and who signed off. Run restore tests on a schedule and keep the results for assessors.
What change records does a Superset upgrade produce?
Each upgrade should leave a change record built from the upgrade procedure itself. The Superset documentation calls for a metadata database backup, running superset db upgrade and superset init, testing in staging first, and reviewing UPDATING.md for breaking changes between versions. Keep the target version, the reviewed breaking changes, the staging test results, the backup reference, the approver, and the rollback plan. Pair this with ClickHouse change records, since the ClickHouse production guidance recommends a pre-production environment with representative data and reading the changelog before upgrading. Confirm which ClickHouse release line your change board has approved, because supported versions change over time.
How do IEC 62443 and NIS2 map to our security program documents?
Map each document to the control area it supports rather than writing one generic policy. For the IEC 62443 series, ANSI/ISA-62443-2-1-2024 addresses the security program requirements for asset owners, ANSI/ISA-62443-3-2-2020 covers risk assessment for system design, and ISA-62443-3-3-2013 sets system security requirements and security levels. Your analytics platform is typically a system within the wider IACS and IT estate, so the zone and conduit model and security levels need to be stated explicitly. For NIS2, Article 23(4) requires an early warning within 24 hours, an incident notification within 72 hours and a final report within one month for significant incidents, so your runbooks and logs need to support those timelines. Confirm with counsel whether your organization is an in-scope entity, since the manufacturing sector classification covers several NACE divisions.
Does DeployOpen issue audit reports or certifications for our dashboards?
No. DeployOpen plans, implements, hardens and documents the platform and prepares the systems, controls and evidence that an assessment will examine. Audit reports, attestations and certifications are issued only by independent auditors and certification bodies. In practice, DeployOpen can help assemble the access logs, RLS test records, backup and restore records, change records and the mapping to IEC 62443 and NIS2 obligations described above, and can run security testing as a separate engagement. Your assessor will decide what evidence is sufficient. Ask any partner to state this boundary in the written scope.
Will customers' auditors need the Cyber Resilience Act evidence too?
Possibly, and the timing matters. Regulation (EU) 2024/2847 applies to products with digital elements from 11 December 2027, but manufacturers' Article 14 obligations to report actively exploited vulnerabilities and severe incidents have applied since 11 September 2026 (Article 71(2)). If your connected product or its portal is in scope, you need a documented process for vulnerability handling and incident reporting that covers the analytics components, including Superset and ClickHouse versions in use. Fines under Article 64 reach EUR 15 million or 2.5% of worldwide annual turnover, whichever is higher, for breaches of certain obligations. Confirm applicability with your legal and product compliance team, since scope depends on the product and its classification.
Can we produce these records ourselves, and what does a partner add?
Yes, many teams can produce these records with internal staff if they assign owners and schedule the work. The recurring items are access log exports, monthly or per-release RLS test runs, quarterly restore tests, and change records for every upgrade. The one-time items are the initial test suite, the document mapping and the runbooks. A partner typically adds value in building the automation, running independent security testing, and writing the mapping against the standards. Most DeployOpen engagements are fixed-scope, with deliverables and timelines agreed in writing before work starts, and ongoing support can be provided through a retainer. Ask any provider to itemize which evidence they will build and which your team must operate.
What to ask an implementation partner before hiring
Start with tenant isolation, because a manufacturer's customers will see each other's machine data if it is wrong. Ask the partner to show, in a running environment, how row-level security is enforced per customer in Superset and in ClickHouse. The useful answer names the mechanism on both sides: which Superset row-level security rules are attached to which roles, how those roles are derived from the identity provider, and whether ClickHouse enforces the same restriction through its own users, roles and row policies so that a direct query cannot bypass the dashboard layer. Ask for a test plan that logs in as customer A and attempts to read customer B's machines, sensors and sites, and ask who keeps that test in the regression suite after go-live.
Next, ask how guest tokens or SSO embedding are implemented, and how the embedding service proves which customer is asking. Ask which embedding method is planned, who signs the tokens or assertions, how long they live, and what happens to a session when a customer's contract or equipment fleet changes. Ask how SECRET_KEY is generated, stored and rotated using PREVIOUS_SECRET_KEY, and how group-to-role mappings through AUTH_ROLES_MAPPING are reviewed. Ask how OT data crosses the boundary: which historian or gateway exports data, whether plant networks reach the analytics tier only through a one-way or brokered path, and who approves that path against the zones and conduits defined under ISA/IEC 62443. Ask for the written data flow diagram before any build starts.
Then test the operational claims. Ask how a ClickHouse restore is tested, not just how backups are scheduled. The backup documentation warns that a backup never test-restored will probably not restore properly, so ask for the restore schedule, the spare cluster used, and the measured restore time the partner has achieved on representative data. Ask which ClickHouse support line the partner will run and confirm it against the security policy, which lists versions 26.9, 26.8, 26.7 and 26.3 as supported with security updates. Ask how the partner will move between stable and LTS lines, and who owns the changelog review before each upgrade. For license obligations, ask in writing how AGPL terms apply if Metabase is modified and distributed inside the customer portal, and whether any feature you need sits in the commercial directory. Superset is Apache-2.0 with no paid edition, which makes it the simpler licensing answer, and the partner should say which path it recommends and why.
Questions buyers ask before hiring help with embedded analytics
What drives the cost of embedded analytics for connected equipment?
Cost follows four drivers: the number of customer tenants, the volume and retention of machine and sensor data, the isolation model you choose, and how much OT integration the work requires. Each tenant adds dashboards, row-level security rules and access testing. Time-series volume sets the size of the ClickHouse cluster, its backup storage and the replication design. A shared-table model with row-level security costs less to build than separate databases per customer, but it needs stronger testing to prove that one customer never sees another's data. OT integration adds connectors, change control and plant-side review. Ask any implementation partner for a quote that itemizes these four drivers and states which parts are one-time and which recur, such as hosting, backups, patching and retainer support.
How long does it take to get a first embedded dashboard live?
The first dashboard depends on how quickly the data owners confirm the data model, how many machine signals feed it, and whether the customer portal's identity system is ready for single sign-on or guest tokens. A narrow first release, such as one equipment family and one set of sensor metrics for a pilot group of customers, moves faster than a full fleet rollout. Work runs under a written, fixed-scope agreement, so the timeline is set in that document rather than estimated here. Ask the partner which milestones the schedule depends on, which of them sit with your team, and what happens if plant access or data-owner approvals arrive late.
What does the manufacturer need to supply before implementation starts?
The manufacturer must supply four things: named data owners who can approve the metrics and retention rules, a documented path to the machine and sensor data, a decision on how plant networks connect to the analytics platform, and the identity provider that customers will use. Data owners settle which signals are customer-visible and which are internal. OT access needs an agreed method, such as a read-only export or a one-way data path, and the security team must approve it. The identity provider must support OAuth2 or LDAP so that Superset or Metabase can map groups to roles. Confirm these inputs with your plant engineering, OT security and customer identity teams before signing a scope.
Can our internal team run embedded analytics without outside help?
A team with Linux, Kubernetes or VM operations, PostgreSQL administration, ClickHouse experience and identity integration skills can run the platform, but the workload is specialized. The Superset architecture includes a metadata database, an optional caching layer and optional Celery workers, and alerts, async queries and thumbnails depend on the caching and worker components. ClickHouse needs at least three ClickHouse Keeper nodes in production, and backups must be restored and tested on spare clusters. Before deciding, list the people who would own upgrades, backups, row-level security changes and on-call response, and check whether they can cover those tasks during a plant incident.
How are upgrades handled for Superset and ClickHouse?
Upgrades follow a documented sequence that someone must own. For Superset, the upgrade documentation says to back up the metadata database, run superset db upgrade and superset init, test in staging first, and read UPDATING.md for breaking changes. ClickHouse stable releases ship roughly monthly, LTS releases ship twice a year and are supported for a year, and its security policy lists supported versions, so choosing a release line is a governance decision. Ask the partner how staging mirrors production, how rollback works, and whether upgrades, patching and testing are included in a retainer. Retainers cover ongoing support, upgrades and continuous testing, and their terms belong in the written scope.
How do retainers differ from the initial implementation fee?
The initial implementation is usually a fixed-scope project with agreed deliverables. Ongoing work, such as platform upgrades, monitoring, backup verification, security testing and changes to dashboards or tenants, is often handled through a retainer. Ask what the retainer covers, how change requests are priced, how response expectations are written, and what happens when a customer tenant is added. Confirm whether the retainer includes restore tests and staging upgrades. DeployOpen does not publish a price list, so request a written quote that separates one-time work from recurring support.
Does a plant-facing analytics platform need OT security review?
Yes. Any path between plant systems and an analytics platform should be reviewed against the ISA/IEC 62443 series, which organizes requirements into general, policies and procedures, system and component groups. ANSI/ISA-62443-3-2-2020 covers security risk assessment for system design, which fits the connection design, and ANSI/ISA-62443-2-1-2024 covers program requirements for asset owners. NIST SP 800-82 Rev. 3 covers OT security and states that performance, reliability and safety requirements matter. Keep customer dashboards on a separate tier from operational control systems, and have OT engineers approve any read path. Ask your partner which of these standards their assessment covers and how findings are documented.
Which regulations affect customer-facing machine data dashboards?
Several regulations may apply, depending on where your products are sold and which sectors you serve. NIS2 Annex II lists manufacturing as a critical sector, including electronic, electrical, machinery and transport equipment, and in-scope entities face incident reporting deadlines in Article 23(4). The Cyber Resilience Act, Regulation (EU) 2024/2847, applies to products with digital elements from 11 December 2027, while manufacturers' Article 14 reporting obligations have applied since 11 September 2026. The Machinery Regulation applies from 14 January 2027. DeployOpen prepares systems, controls and evidence for compliance assessments but does not issue audit reports or certifications. Ask your legal and compliance team to confirm which obligations apply to your products.
How DeployOpen works on this
DeployOpen plans, implements, customizes, integrates, hardens, documents and supports self-hosted open-source platforms inside the client's own infrastructure. For manufacturers embedding machine and sensor dashboards, that means the ClickHouse, Superset and portal integration runs in your cloud accounts or data centers, so plant data stays out of shared SaaS.
How we scope
- Written scope lists the customer tenants, machine data sources, dashboards and portal pages to be embedded.
- Scope sets the row-level security model that keeps each customer's equipment and sensor data separate.
- Engagements are fixed-scope, with deliverables and timelines agreed in writing before work starts.
How we build
- Deploy ClickHouse with replicated tables and ClickHouse Keeper nodes in your infrastructure for time-series storage.
- Configure Superset with its metadata database, caching layer and workers, and connect SSO for portal users.
- Test per-customer isolation with cross-tenant access attempts before any customer sees a dashboard.
How we hand over
- Documentation covers architecture, tenant mapping, role configuration and the upgrade and rollback procedure.
- Backup and restore runbooks include scheduled restore tests on a spare cluster, not only backup jobs.
- Handover includes evidence records for access control, logging and configuration that your auditors can review.
How we support
- Ongoing support, upgrades and continuous security testing are available as retainers after handover.
- Upgrades follow a staging-first process, with changelog review for ClickHouse and UPDATING.md checks for Superset.
- Confirm with us which support hours, response commitments and patch windows your written scope will include.

