A 2026 SCADA platform shortlist is won or lost on six measurable criteria — driver openness, tag-count licensing, historian retention, redundancy topology, HMI authoring tooling, and IEC 62443 zone segmentation — before a single demo is booked [S4][S1].
Supervisory Control and Data Acquisition sits one layer above the PLC and RTU layer, polling field devices and serving trends, alarms and control screens to operators, so its selection criteria are fundamentally about I/O breadth, lifecycle cost, and cyber-segregation fitness for the connected Industrial Control System (ICS) [S4].
Driver Coverage and Open-Protocol Architecture
Native driver count and openness to OPC UA over TCP are the single most cited selection filters in 2026 RFPs, because every additional proprietary gateway adds a failure node and a per-tag licence fee [S4].
Open-architecture telemetry stacks such as those supplied by Sage Designs target exactly this criterion, framing SCADA as a hardware-plus-software bundle that avoids the closed-stack penalty on municipal and industrial retrofits [S1]. Free or low-cost stacks — WinTr, LAquis, and the open-source Go-language projects catalogued on SourceForge — now ship native Modbus TCP, Modbus RTU, S7 MPI/PP, Profinet, and Hostlink drivers out of the box, so any shortlisted platform missing two of these five is a hard reject on a brownfield PLC fleet [S3][S4][S6].
Engineers running modern IEC 61131-3 PLC fleets should specifically verify OPC UA Pub/Sub over TLS 1.3 in writing; running only legacy OPC DA leaves a single firewall-piercing DCOM surface that IEC 62443-3-3 zone modelling will flag as non-compliant on any 2026 plant audit.
Tag Licensing Model and Real Cost per I/O Point
Per-tag perpetual licences on legacy SCADA stacks typically run in the USD 5–15 range per I/O point, while modern platform subscriptions price per "concurrent tag" or per "named device", and the difference can swing a 5,000-point project by an order of magnitude [S3][S4].
Where vendor pricing is opaque, ask for the published price list line for an additional 500 tags after year one; this single number exposes the licensing model more honestly than any data sheet. A second filter is software-channel availability: platforms that sell only through regional system integrators carry a 20–35% channel margin that pure-direct OEM sales do not, and that margin is rarely visible in the headline licence quote.
Engineers should also pin down the upgrade model. Open-source Go-language SCADA projects and freeware such as LAquis charge USD 0 for the runtime but shift cost onto integrator hours, while commercial stacks charge annual support at roughly 18–22% of licence list — a meaningful line item for any plant owning the software longer than five years [S4][S6].
Historian Retention, Time-Series Database and Reporting
Process historians embedded in 2026 SCADA platforms typically support 90 days of raw 1-second retention before compression, with the major commercial stacks offering 1–5 years of on-disk archive at configurable compression ratios (commonly 10:1 lossless to 100:1 lossy) [S4].
The selection test is concrete: ask the vendor to demonstrate a 10,000-tag historian running at 1-second scan, then measure (a) disk consumption per day, (b) retrieval latency for a 30-day trend of 200 tags, and (c) export format — CSV, Parquet, or OPC UA HD raw. Any retrieval latency above 4 seconds for that query is a hard shortlist reject, because operator HMI screens depend on sub-second trend rebuild during trip diagnostics.
Reporting integration is the second axis: a stack with native WebHMI or REST/JSON export to a corporate BI layer (Power BI, Grafana, InfluxDB) collapses a parallel data-collection project, while a stack limited to vendor-proprietary reports typically doubles the historian-to-board cost over a 10-year horizon.
Redundancy Topology and Failover Time
Redundant SCADA server pairs, a primary-plus-hot-standby pair, remain the default for any process line whose unplanned downtime exceeds USD 10,000 per hour, and a 2026 shortlist must specify failover time in writing rather than in marketing copy [S3].
Acceptable failover targets: warm-standby server pairs under 30 seconds state-sync, hot-standby pairs under 5 seconds, and triple-redundant (2-out-of-3 voting) server clusters under 2 seconds for safety instrumented functions. Anything slower than 30 seconds is acceptable only for non-critical utilities telemetry [S3].
Network-level redundancy — dual NICs on independent managed switches, PRP or HSR ring topology at the PLC layer — must be priced and demonstrated separately, because the SCADA software's redundant-server feature does not by itself cover Layer-2 switch failure. A common RFP trap is a "redundant SCADA" quote that fails to include the redundant licence cost; vendor contracts should make redundant-server pricing an explicit line.
Cyber-Security Posture: IEC 62443 Zone Segmentation
2026 shortlisting without an explicit IEC 62443-3-3 SL-2 (or SL-3 for safety-critical plants) target is malpractice, because the standard's zone-and-conduit model now drives both insurance underwriting and regulatory audit outcomes [S1][S4].
Verify in writing: role-based access control (RBAC) at the user-group level, Active Directory / LDAP integration, signed firmware updates, audit-log retention of at least 90 days, and TLS 1.2 or higher on every external protocol including OPC UA. Sage Designs explicitly markets a 2004-pivot to security products for utility infrastructure, framing SCADA selection as inseparable from perimeter defence [S1].
Engineers should also confirm that the platform can host an industrial DMZ without external connectors; platforms that require a vendor cloud service for "advanced features" create an outbound tunnel that is increasingly rejected by IT/OT converged network policies.
HMI Authoring, Screen Reuse and Operator Workflow
HMI authoring tool speed — measured in screens authored per engineer-day — is the second-largest hidden cost after licensing, and the 2026 stacks differentiate sharply here: vector-based web HMI authors typically run 3–5× faster than legacy Win32 tag-database editors [S3][S7].
Concrete selection test: build a 20-tag, 4-screen mimic (one motor, one PID loop, one analog trend, one alarm summary) in a vendor-supplied trial, then rebuild it on the second vendor. The time delta is the truest productivity metric the data sheet will not give you. The WinTr 5.x release explicitly added multi-screen designer and small-screen target rendering for this reason, recognising that operator-station hardware is no longer one-size-fits-all [S3].
For deeper control-room context, the HMI vs SCADA decision map covers where an HMI panel is the right tool and where the supervisory layer is mandatory, which keeps a small retrofit from being over-specified.
Who Should NOT Spec the Mainstream Closed-Stack SCADA
Similarly, plants that need 5-second-or-better failover for safety instrumented functions should bypass server-pair redundancy entirely and specify a SIL-2/3-certified safety PLC plus a non-redundant SCADA; redundant SCADA software does not raise the SIL of the underlying control loop, and paying for it on a SIL-1 process is wasted budget.
Conversely, multi-site municipal utilities with thousands of remote RTUs and a regulatory mandate to report to a state PUC should not specify open-source stacks: the total cost of ownership over a 10-year horizon, including integrator on-call and patching, generally exceeds the closed-stack licence-plus-support quote once remote-site count crosses roughly 50 RTUs [S1].
2026 Shortlist Logic and Trackable Signals
Build the shortlist in three passes: (1) driver-coverage cut to 2–3 platforms, (2) tag-cost and historian-retention cut to 2, (3) redundancy and IEC 62443 fit cut to 1 finalist. Anything that requires a custom driver is a reject on pass 1; anything that cannot quote a per-tag price in writing is a reject on pass 2; anything that fails to document failover time and SL target is a reject on pass 3 [S1][S3][S4].
Two trackable signals to watch into late 2026: (a) IEC 62443-3-3 SL-3 certification announcements from major SCADA vendors, which will reset the cyber-security shortlist floor; and (b) the release of OPC UA FX (Field eXchange) compliant platforms, which finally collapse the controller-to-SCADA network layer into a single converged bus and will re-rank the redundancy and driver-coverage criteria for any 2027 retrofit.
Spec-level background on the components involved: scada software, and pressure transmitter.