HMI vs SCADA: when each is the right tool

A small bottling line with one PLC, fifty tags and three operator screens does not need a SCADA system. A wastewater treatment works with eight PLCs across two pump stations, a clarifier loop, a chlorination skid and a remote intake — that one needs SCADA, even if the budget makes the conversation uncomfortable. Most of the confusion about HMI versus SCADA comes from people on the wrong side of that line: a small site sold an oversized SCADA package by an integrator, or a medium site running a single big HMI panel-mount because the original spec never grew up. This page is for the engineer or facility manager who has to decide which tool fits the plant in front of them.

Try the simulator →

TL;DR

• HMI = panel-mount runtime, single-machine scope, fixed-spec hardware. Typical SA HMI panel cost is R30k for a 7-inch through R150k+ for a 22-inch high-end Comfort or PanelView Plus 7.
• SCADA = server-based runtime, plant-wide scope, historian plus alarm management plus reporting plus user-role security. SA SCADA software licence is typically R200k+ per server before tags and engineering.
• The break point in practice: 3 PLCs, 500 tags, batch reports, regulatory operator-handover documentation. Hit any one of those and SCADA starts paying for itself.
• Both can talk to the same controllers — Siemens TIA Portal hosts both WinCC HMI runtime (Comfort panels) and WinCC OA / Unified (SCADA). Rockwell does the same with PanelView Plus and FactoryTalk View SE.
• The architectural mistake that costs the most is choosing SCADA for a site that needs an HMI, then finding out that the OPC layer, the historian licence, the redundant server pair and the SCADA engineer time eat the project margin.

Side-by-side

Criterion	HMI (panel-mount)	SCADA (server-based)
Scope	Single machine or skid	Plant-wide, multiple PLCs
Hardware	Dedicated panel — Siemens Comfort, Rockwell PanelView Plus 7, Pro-face, Beijer	Server — physical or virtualised, often redundant pair
Typical SA hardware cost	R30k (7-inch) to R150k (22-inch high-end)	R0 hardware if virtualised, plus the server farm cost
Typical SA software cost	Bundled with panel hardware in most cases	R200k+ per server licence (WinCC OA, FactoryTalk View SE, Wonderware)
Tag scale	Up to ~1000 tags comfortably	5000 to 250000+ tags
Historian	None or basic ring buffer	Full historian — minutes to years of trended data
Alarm management	Local annunciation, optional SMS	Full alarm prioritisation, shelving, ISA-18.2 compliant workflows
User roles	2-4 levels typical	Active Directory integration, granular permissions
Reporting	None or minimal	Batch reports, shift handover, regulatory submissions
Network footprint	Wired to PLC, sometimes one Ethernet drop	Plant network, OPC UA, possibly DMZ to corporate
Operator training	Hours	Days to weeks
Reference	Wikipedia: HMI	Wikipedia: SCADA

Where each one wins

HMI

A panel-mount HMI wins when the scope is one machine, one operator, one place. The classic case is a packaging line where the line operator stands at the infeed, watches the screen for jam alarms, hits start and stop buttons on the screen, and goes home at the end of the shift. The site does not need plant-wide visibility because there is no plant — there is one line. The budget for trended history is zero because if the line jams the operator cleans it and pushes start. Reporting is whatever the line manager extracts from the PLC tag log once a week.

For that scope, a Siemens TP1500 Comfort or a Rockwell PanelView Plus 7 1500P at R80k-R150k installed delivers everything the site needs: capacitive touch, alarms, a few trends for last-shift visibility, recipe management, and a stable runtime that does not need a Windows update window. The HMI runtime is firmware on the panel — it boots, it runs, it does not get patched on Patch Tuesday. That stability is the reason panel-mount HMIs still dominate machine-level installs even when the software side of automation has gone increasingly server-based.

The argument for HMI gets stronger as the operator-machine distance gets smaller. If the operator is two metres from the panel during normal running, the HMI is the right tool. If the operator is two hundred metres away in a control room watching twenty machines, that is a SCADA conversation.

SCADA

SCADA wins when the operator has to see the whole plant from one place, and when the data needs to live longer than the current shift. A wastewater plant where the operator in the control room is monitoring eight pump stations, a clarifier, a chlorination skid and three bulk tanks needs SCADA. A petrochem unit where the day-shift handover requires a written report of every alarm acknowledged in the last twelve hours needs SCADA. A pharmaceutical batch plant where the regulator asks for batch-traceability records going back five years needs SCADA — specifically a SCADA package with a validated historian and 21 CFR Part 11 audit trails (or its SA-relevant SAHPRA equivalents).

The historian piece is the thing that justifies SCADA most often. A panel HMI can buffer a few hours of trended data; a SCADA historian holds years of it, indexed and queryable, and supports the OEE dashboards, the incident investigations, the regulatory submissions and the engineering trend analyses that make running a complex plant viable. Once a site is generating reports off the historian regularly, replacing it with anything thinner is a step backward.

The other strong argument is alarm management at scale. ISA-18.2 alarm management — alarm rationalisation, prioritisation, shelving, performance metrics on alarm rate per operator per hour — is a SCADA-grade workflow. A panel HMI can annunciate alarms but cannot run the management discipline on top of them. On a high-alarm-density plant (a chemical reactor on startup, a paper machine after a sheet break), the operator is drinking from a firehose and the difference between SCADA-grade and HMI-grade alarm handling is measured in incidents per year.

What this means in SA

The break point in practice is around 3 PLCs, 500 tags or a regulatory reporting requirement. Below that, an HMI does the job. Above that, the SCADA cost starts paying for itself in ways the spreadsheet captures.

SA water utilities run a lot of SCADA. Bulk water suppliers, large municipal water and wastewater operations, and the better-run district municipalities all have SCADA systems controlling pump stations, reservoirs and treatment works across geographic areas larger than any single panel HMI could span. The typical platform is Siemens WinCC OA or WinCC Unified for greenfield, with older sites still on Wonderware (now AVEVA) or Schneider ClearSCADA. The SCADA engineer headcount in SA water is real — every metro has integrators specialising in it.

SA petrochem is similarly SCADA-heavy. The major refineries and synfuels operations run distributed control systems (DCS — a SCADA-adjacent category with tighter integration to the controllers), with WinCC OA or PCS 7 on top in the smaller utility-system corners. Mining beneficiation runs SCADA at the plant level — concentrators, smelters, refineries — usually with a mix of FactoryTalk View SE on the AB side and WinCC on the Siemens side. Cement plants have SCADA for the kiln, mill and packing lines.

SA F&B and packaging tend to live on HMI. A standalone bottling line, a meat-processing line, a small dairy, a juice line — these run on panel-mount HMI, and adding SCADA only makes sense when the site grows to multiple lines on a shared plant network with shared engineering and shared reporting requirements. Many small to mid-sized food processing plants spent good money on SCADA they did not need because an integrator sold them a package; the result is a SCADA server gathering dust because no one trained the operators on the workflows that justify it.

SA mining outside of beneficiation — the mine itself, the underground operations, the conveyor systems — varies. Coal mining has a lot of SCADA at the wash plant. PGM mining has SCADA at the concentrator. Underground signalling and personnel tracking is SCADA-adjacent but usually a separate specialist system. The fragmentation of mining IT means there is rarely a single plant-wide SCADA in the sense petrochem has one.

Common mistakes when picking

• Buying SCADA when HMI fits. The classic case: a single-line bottling plant where an integrator sold a SCADA package on the promise of "future expandability". Two years later the site has the same one line, paying annual maintenance on a SCADA licence that never grew into its scope. If the site is one machine, buy the panel HMI.
• Buying HMI when SCADA fits. The mirror case: a wastewater works with five PLCs running on one panel HMI per station, with no central visibility, no historian, no shift reporting. The integrator saved cost on the original install and the operations team has been working around the gap ever since. If the site is more than three PLCs across more than one location, plan SCADA from day one.
• Skipping the historian sizing. SCADA historian sizing is a real engineering activity — tag count times update rate times retention period equals storage. Sites that skip the sizing end up with disks full inside two years and an emergency archive migration during the next shutdown.
• Putting SCADA on the corporate IT network. SCADA wants its own segment with controlled gateways to corporate IT. Putting the SCADA server on the same VLAN as the office printers is how plants get hit by ransomware that walked in through HR's phishing inbox. The IEC 62443 zones-and-conduits model is the reference; even a small SA plant should know the basic patterns.
• Forgetting the operator training cost. A SCADA system is only as good as the operators using it. Budget the training as part of the project, not as an afterthought. Operators who never learnt the alarm-shelving workflow will silence every nuisance alarm by killing the audio, which defeats the whole alarm management discipline.
• Buying redundancy you do not need. Hot-standby SCADA server pairs are correct for high-availability plants. They are overkill for a small water works where a 30-minute SCADA outage is operationally fine. Match the redundancy to the operational tolerance, not to the integrator's quote sheet.

How to test the trade-off in the simulator

Build the same control problem twice — once HMI-scoped, once SCADA-scoped. Drop an S7-1500 with the rack shown in the image at the top, wire a small process: two pumps, a tank with level transmitter, a flow loop. On the HMI side, mock up a single-screen panel that gives the operator the start, stop, fault reset, level bar graph and last-hour trend. On the SCADA side, mock up the same data but add the historian view, an alarm summary across multiple equipment items, a trend that goes back days, and a shift-handover report.

The exercise teaches the trade-off concretely. The HMI side is faster to build, comes up cleanly, runs in the simulator with no external dependencies. The SCADA side takes longer, needs more configuration, surfaces more edge cases — but produces something an operator could actually use to run a multi-station plant. Doing both in the simulator before specifying for a real site is the cheapest way to internalise where the dividing line sits.

Start the free tier →

Vendor reference

The Siemens reference for HMI is Siemens Comfort Panels documentation — search for "Comfort Panel" in the online support portal for the panel hardware manual and the WinCC TIA HMI configuration guide. The Siemens reference for SCADA is the WinCC Unified or WinCC OA documentation on the same portal. The Rockwell reference for HMI is Rockwell PanelView Plus documentation — search for "PanelView Plus 7" for the panel hardware manuals. The Rockwell SCADA reference is FactoryTalk View SE on the same site. For independent overviews, Wikipedia: SCADA covers the architectural patterns, and the ISA-18.2 alarm management standard is the reference for the alarm-management workflows that distinguish SCADA from HMI in practice.

What we don't claim

This site is not SAQA-registered, not MerSETA-accredited, and not an NQF-registered qualification provider. We do not specify SCADA systems for production sites — that is a job for an integrator with on-site experience and a current vendor partnership. We do not commission HMI panels either. The simulator is a learning tool — practise the panel-screen layout, the alarm hierarchy, the trend configuration, the historian query patterns, in a sandbox where the cost of getting it wrong is zero. Pricing figures above are indicative SA list-price ballparks at the time of writing — get a current quote from the local distributor before you commit to a panel order or a SCADA licence.