A trend is a chart that plots tag values against time, and it is the single most useful tool an operator has for understanding what a process is actually doing rather than what it is doing right now. A numeric readout tells you a pressure is 82 psi; a trend tells you it has been climbing steadily for an hour, which is a completely different piece of information. Building a good trend is not hard, but there are a handful of choices that decide whether the finished chart is readable or a tangle of overlapping lines, and whether it answers a question in one glance or hides the answer. This guide walks through building a trend display from scratch: adding the pens, giving each one a color and an axis, setting the time window and how often it updates, deciding between watching live data and reviewing history, and handling long time ranges without drowning the chart in points.
How to build a trend in SCADA in one line: To build a trend in SCADA, create a new trend display and add a pen for each tag you want to plot, give each pen a distinct color and assign it to an axis whose range suits that value, then set the time window and refresh rate. Choose real-time mode to watch data arrive live or historical mode to review a past window, and for long ranges enable min/max decimation so the chart stays fast and still shows the peaks and dips. Start with two or three related pens rather than crowding one chart, and save the layout as a template if you will reuse it.
A pen is a single line on the trend, and each pen is bound to one tag. Building a trend starts with adding a pen and pointing it at the tag you want to see, then repeating for every value you want on the same chart. The first decision that matters is restraint: it is tempting to throw a dozen pens onto one trend, but a chart with too many lines becomes an unreadable knot where no single value can be followed. The most useful trends group a small number of related pens, a pump's suction pressure, discharge pressure, and flow together, so the relationships between them are the point of the chart rather than an accident of crowding.
Each pen needs a color, and the colors should be chosen so that adjacent lines are easy to tell apart at a glance, avoiding two similar shades that blur together when the lines cross. On a display that operators will read for hours, high-contrast colors against the chart background matter more than looking pretty. Many trend tools let you also set the line thickness or style, which is a second way to distinguish pens, a solid line for the measured value and a dashed line for its setpoint, for instance, so the two read as clearly related but different.
The axis is the choice that most often separates a good trend from a confusing one. If two pens have very different ranges, a pressure that swings between 0 and 100 and a flow that swings between 0 and 5000, plotting them on a single shared axis flattens the small-range pen into a nearly straight line hugging the bottom. The fix is to give each pen, or each group of similar pens, its own axis with a range that suits it, so a 0 to 100 axis on the left for pressure and a 0 to 5000 axis on the right for flow. Setting each axis range deliberately, rather than letting it autoscale wildly, also keeps the chart stable so the eye can compare today's shape to yesterday's without the scale jumping around underneath it.
The time window is how much history the chart shows at once, the span from the left edge to the right edge. A one-hour window is right for watching a process respond to an adjustment in real time, while an eight-hour or twenty-four-hour window is better for spotting a slow drift or a daily cycle. The window you pick should match the question you are asking, because too short a window hides slow trends and too long a window compresses fast events into invisibility. Most trend tools let the operator change the window on the fly, so the sensible default is a span typical for that screen, with the freedom to zoom out when something looks off.
Refresh rate is how often a live trend redraws with new data, and it interacts with the scan rate of the underlying tags. There is no benefit to refreshing a chart faster than the data actually arrives from the field; a tag scanned once every ten seconds gains nothing from a trend that redraws every second, it just repaints the same value. A sensible refresh keeps the live edge of the chart moving smoothly without asking the system to do pointless work, which matters especially for remote sites where data arrives less frequently than on a wired plant floor.
The deeper choice is between real-time and historical mode. A real-time trend shows data as it arrives, with the newest value at the right edge and the whole chart scrolling left as time passes, which is what an operator wants while actively running the process. A historical trend instead shows a fixed window in the past, drawn from stored data, which is what you want when investigating an event that already happened, a trip overnight, a pressure excursion last Tuesday. Historical mode depends on the values having been logged to a historian in the first place, so a value that was never configured for logging can be watched live but cannot be reviewed after the fact. Deciding early which values you might need to investigate later is really a decision about what to log, and building a trend is often where that gap becomes obvious.
When you stretch a historical trend across a long time range, a week or a month, the number of stored points can become enormous, far more than a chart has pixels to draw or a slow connection has bandwidth to send. Plotting every raw point is both slow and pointless, because the screen cannot show more detail than it has pixels. The answer is decimation, which reduces the point count for display. Simple decimation just drops points, but the better approach for process data is min/max decimation, which for each pixel column keeps both the highest and lowest value in that slice of time. This preserves the spikes and dips that a naive thinning would erase, so a brief pressure surge that lasted two minutes still appears as a spike even on a month-long chart.
Min/max decimation matters most exactly where bandwidth is scarce, which is why it is a natural fit for cloud monitoring of remote sites. In a cloud SCADA platform such as Merobix, an operator can open a month of history on a tag from a distant, unmanned site in a browser, and the platform sends a decimated series that renders instantly instead of shipping every raw sample over a cellular link. Because min/max decimation keeps the extremes, the operator still sees the overnight low and the afternoon peak that matter for spotting a problem, without waiting on a huge download. This is the mechanism that makes long-range historical review practical for sites that report over metered or intermittent connections.
Once you have a trend arranged the way you like it, with the right pens, colors, axes, and window, it is worth saving it as a template rather than rebuilding the same layout every time. A template lets you reproduce a standard trend for the next pump, the next well, or the next site by just repointing the pens at the new tags, which keeps trends consistent across a facility and saves an enormous amount of repeated setup. For an operator building their first trend, the workflow is therefore: get one trend right by hand, understand why each choice was made, and then capture that work as a template so the second and hundredth trend take seconds instead of minutes.
This almost always means both pens are sharing one axis whose range is dominated by the larger value. A pressure that ranges 0 to 100 plotted on the same axis as a flow that ranges 0 to 5000 gets squashed into a nearly flat line at the bottom. The fix is to give each pen its own axis with a range suited to that value, so a left axis for pressure and a right axis for flow, so both curves use the full height of the chart.
A real-time trend shows data as it arrives, scrolling so the newest value stays at the right edge, which is what you want while actively running a process. A historical trend shows a fixed window in the past, drawn from logged data, which is what you want to investigate an event that already happened. Historical review only works for values that were configured to be logged to a historian, so a value watched only live cannot be reviewed after the fact.
A long time range holds far more stored points than a chart has pixels to draw or a slow link has bandwidth to send, so plotting every raw sample is slow and pointless. Decimation reduces the point count for display, and min/max decimation is preferred for process data because it keeps both the highest and lowest value in each slice of time. That preserves brief spikes and dips that simple point-dropping would erase, which is exactly what makes long-range review over cellular links practical.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.