A historian for forty tags

The floor, in numbers

The three smallest packages sold by the tag, read from each vendor's own pricing page on 2026-10-21. Full table — including the products that are not metered by tag — and the method on what a data historian actually costs.

A single packaged machine is usually twenty to eighty signals worth historising: a few temperatures and pressures, motor currents, run states, a counter or two, some setpoints. A small plant with several machines and a boiler house is comfortably inside the low hundreds.

So the entry-level metered historian is somewhere between forty and a hundred times what you need, and it is priced accordingly. You are not being quoted enterprise money because your problem is hard. You are being quoted enterprise money because no historian vendor sells the small size. What does exist below it is either a historian with no tag limit that you host yourself, the next section, or hosted storage that is not sold as a historian, in the one after.

What people actually do instead

1. Nothing — the PLC keeps its own trends

Worth naming first because it is what most small sites do. The HMI has trend screens, the drive keeps a fault buffer, and when something breaks you look at the last few hours.

This is fine until the question is comparative — “is this bearing running hotter than it did in the spring?” — because that needs data from before you started wondering. The cost of doing nothing is invisible right up until the first time you need last year.

2. Free software on a spare PC

Entirely legitimate, and the licence really is zero. Timebase is free with no tag limit, TDengine is free self-hosted up to 5,000 tags, and InfluxDB OSS is free. Add Grafana and you have a historian with dashboards for the price of a machine.

3. Hosted, at the small end

Hosted storage that is not sold as a historian exists at this size too, priced other ways: emoncms.org at £1.00 per feed per year ex VAT (£1.20 inc VAT), and TagoIO's EU Starter at 59€ / month, plus usage beyond what the plan includes, sold by device. Both are in the full comparison, read from their own pages on 2026-10-21. What each keeps, at what detail and for how long, is on those pages to check.

This is what we sell, so weigh it accordingly. Business is 500 tags for €89/month excl. VAT — €111.25 incl. 25% VAT for consumers in Sweden. Pro is 50 tags for €9/month excl. VAT — €11.25 incl. 25% VAT for consumers in Sweden. The free plan is 5 tags and costs nothing, which is enough to prove the plumbing with a real PLC before spending anything.

Full resolution is 1 year on Pro and 2 years on Business, and the hourly minimum, maximum, average and count for every tag is kept 10 years on every plan — so the “is it running hotter than last spring” question stays answerable long after the detail has gone. See how long data is kept.

Ingestion is OPC UA, MQTT, Sparkplug B (beta) or Modbus TCP (beta, for meters and older PLCs) through a collector container you run on the plant network. It connects outward only — no port forwarding, no inbound firewall rule, no PLC exposed to the internet. That constraint is the first thing a controls engineer asks about and the reason the design is shaped this way.

You do not need SCADA to have a historian

A recurring blocker: the historian is sold as a module of a SCADA platform, so it looks like you need the platform. You do not. A historian needs values and timestamps. If your PLC has an OPC UA server — most do now — or anything on the network can publish MQTT, that is sufficient.

Supervisory control is a genuinely separate product and a genuinely large purchase. If what you want is “keep the numbers and tell me when one goes out of range”, buying a SCADA platform to get it is buying a car for the radio.

When we are the wrong answer

• Above 500 tags. That is the ceiling on the largest published plan. If you are at a thousand signals and growing, you have crossed into the range the established vendors are built for, and their smallest package finally fits.
• Data must not leave the site. Some plants have that rule and it is not negotiable. There is no on-premise install here. Timebase is free and runs where you put it.
• You need to pick OPC UA tags from a live tree. Here the collector browses the address space into a spreadsheet file and you delete the rows you do not want. At forty tags that is a few minutes, and nothing is subscribed to that you did not keep. If your team expects to click through a tree with live values, this is not that.
• You need a contractual uptime SLA. The self-service plans have none.
• You need control, alarming on the plant floor, or an HMI. Not what this is.

How to decide, in three questions

1. Count the signals you would genuinely look at a year from now. Not everything the PLC exposes — the ones you would open a trend on during a breakdown. Almost everyone overestimates this by a factor of five.
2. Ask who restores the backup. If there is a name, self-hosted free software is excellent and you should use it. If there is no name, you are choosing between hosted storage and eventually losing the data.
3. Ask whether the data needs to survive the site. A historian in the switchroom shares a fate with the switchroom.

The free plan exists for exactly this decision: 5 tags is enough to point a collector at a real PLC, watch real values arrive, and find out whether any of this is worth paying for — before paying for it.