What Happens to OEE Data When the Network Drops?

Key takeaways

• An outage loses records, not minutes. The line still stopped; the question is whether anyone knows.
• Store and forward means buffer at the edge, then send later, in order.
• Stamp time at the source, or replayed events land at the wrong time.
• MQTT with Sparkplug tells the server a gateway died, even if someone cut its cable.
• Test it in the pilot: pull the cable for an hour and check the report.

When the link between the line and the OEE system drops, the line keeps running and stopping. The question is whether those stops are still recorded. The answer depends on one gateway feature, store and forward. Without it, the shift report gets a hole or a fake stop.

Fabrico, OEE and light MES software with AI cameras, logs downtime automatically from the machine signal. How the gateway on your network handles an outage is a question for Fabrico's technical team. This guide covers what to ask any vendor, and how to test the answer.

What goes wrong with OEE data when the network drops?

Without a buffer, an outage can bend the numbers in two opposite ways.

• The gap reads as running. No stop events arrive, so the shift looks better than it was.
• The gap reads as one long stop. The whole outage lands as downtime, so the shift looks worse.
• Late events get the wrong time. If the server stamps events when they arrive, a 10:05 stop shows up at 12:01.
• Resent events double up. The same stop is counted twice.
• Operators cannot confirm reasons if the tablet at the line needs the server to work.

Each of these breaks trust. After one bad shift report, people stop believing the screen.

What is store and forward?

Store and forward means the gateway keeps each event in local storage while the link is down. Once the link is back, it sends the backlog in its original order, with its original timestamps.

Many industrial edge gateways offer offline buffering, sometimes as a paid extra. The feature only helps if the timestamps are set at the source. An event stamped with the gateway's clock at 10:05 stays a 10:05 event, whenever it arrives.

How does MQTT tell you that a gateway went offline?

MQTT is a light publish and subscribe protocol that many gateways use to send plant data. Sparkplug B adds fixed topic names, a fixed message format and session state on top. Sparkplug 3.0 became the standard ISO/IEC 20237 in November 2023.

Two Sparkplug ideas matter for outages:

• Birth and death messages. A gateway announces itself when it connects. Its death message is registered with the broker in advance, as the MQTT "Will" message. If the gateway drops without saying goodbye, the broker sends that message for it, even if the cable was cut.
• Report by exception. The gateway sends a value only when it changes, not on a fixed timer. That keeps the backlog small.

So the server always knows which gateway is silent, and since when. Our guides to MQTT in manufacturing and OPC UA vs MQTT cover the protocols in more depth.

Worked example: a two hour outage on one shift

Sample line (illustrative numbers, not a customer). A filler runs a 480 minute shift. The link to the OEE server is down from 10:00 to 12:00. During those 120 minutes, the filler had 9 stops worth 26 minutes. Outside the outage it had another 30 minutes of stops, so 56 stop minutes in total.

The true Availability is (480 minus 56) / 480 = 424 / 480 = 88.3%. Here is what the report shows under each setup:

One outage moves the number by more than 5 points either way. Only the buffered setup gives the shift its real score, and it keeps all 9 stops for the reason step.

The buffer itself is small. With report by exception, two hours of state changes and counts from one line is a short list of events, not a video file.

What should you ask an OEE vendor about outages?

1. How many hours does the gateway buffer? It should cover your longest likely outage, such as a weekend of IT work.
2. Does every event keep its source timestamp?
3. Is the backlog sent in order, and are duplicates removed?
4. Can operators still confirm reasons at the line during the outage?
5. How does the system show that a gateway is offline, and who gets told?
6. What happens after a power cut, not only a network cut?
7. How do we test it? Agree a cable pull during the pilot.

What can the plant do on its own side?

• Keep clocks in sync. Point PLCs and gateways at the plant's time server.
• Power the gateway and its switch from a UPS, so a short power dip is not a data gap.
• Wire the gateway where you can. Plan the Wi-Fi for tablets and cameras on its own.
• Alarm on silence. Tell someone when a gateway has been offline for longer than an agreed time.

Where the data is processed also matters. Our guide to edge vs cloud for shop floor data explains the trade-offs, and IoT gateway vs edge device explains the hardware.

Where does Fabrico fit?

Fabrico is a computer vision enabled Manufacturing Performance Platform. It is the OEE and downtime layer of a light MES, with AI cameras on stops.

• Automatic downtime logging from the machine signal.
• The operator confirms the reason in one tap, so each stop gets its cause.
• OEE live by line and shift, with every loss priced in money.
• A pilot on one line, the right place to test an outage before you scale.

Ask Fabrico's technical team how the gateway on your network handles an outage, then test it on the pilot line. There is no free plan and there are no tiers: each plant gets a quote.

Frequently asked questions

What is store and forward in IIoT?

It is a gateway feature that keeps data in local storage while the network is down. When the link comes back, the gateway sends the stored data in its original order and with its original timestamps. For OEE, it means an outage does not leave a hole in the stop log.

Does OEE data get lost when the Wi-Fi drops?

It depends on the gateway. With a buffer and timestamps at the source, nothing is lost; the events arrive late but keep their real time. Without a buffer, stops during the outage vanish or the whole gap becomes one long stop. Ask your vendor which one applies.

How long should an OEE gateway buffer data?

Long enough to cover your longest likely outage. For many plants that is a full weekend, when IT often works on the network. Ask the vendor for the figure in hours for your line, and check it during the pilot with a planned cable pull.

What is an MQTT death certificate?

It is the message a Sparkplug gateway registers with the MQTT broker when it connects, as its "Will". If the gateway drops off without a clean goodbye, the broker publishes that message for it. The server then knows the gateway is offline, even if its cable was cut.

Should OEE be computed at the edge or in the cloud?

Many systems split it. The edge captures states and counts with their timestamps and keeps them during outages. The server or cloud computes OEE across shifts, lines and plants. What matters most is that the raw events are stamped and kept at the source.

Test an outage on your own line

Bring one line and its last shift report. We will show what the timeline looks like, stop by stop.

Request a demo · Contact us