Before swinging door became the standard, historians reduced data with a simpler pair of tests known as boxcar and backslope. Boxcar checks whether a new reading has moved far enough from the last stored value; backslope checks whether it has strayed from the slope the signal was already following. Understanding this older method explains both how early data reduction worked and why the trends it produced can look stepped and jagged compared to modern compression.
Boxcar & Backslope Compression in one line: Boxcar and backslope compression is a legacy two-test data reduction method: the boxcar test discards a new point if it stays within a tolerance of the last archived value, and the backslope test discards it if it stays within tolerance of a line projected forward from the previous slope. A point is stored only when both tests fail. Because boxcar effectively holds the last value flat, the method tends to reconstruct signals as steps, which is why it was largely superseded by swinging door.
The boxcar test is the simpler of the pair and works exactly like a last-value deadband. It compares each incoming reading to the value that was last written to the archive, and if the new reading is within a set tolerance of that stored value, it is discarded. In effect boxcar assumes the signal is holding flat at the last stored level, and it only reacts when the reading has moved far enough away to break that assumption. The name comes from the boxy, flat-topped shape this produces when the held value is drawn as a horizontal line.
The backslope test adds an awareness of trend that pure boxcar lacks. Instead of comparing to the flat held value, it projects the slope the signal was already following, the line through the previous points, forward to the current time, and checks whether the new reading falls within tolerance of that projection. If the signal keeps moving along the slope it was on, backslope predicts it correctly and the point can be dropped. Backslope catches the ramps that boxcar alone would either mangle or store excessively.
In the combined method the two tests run together and a point is stored only when both fail, meaning the reading is neither close to the last stored value nor close to the projected slope. As long as either test succeeds, the point is discarded because at least one simple model still reconstructs it within tolerance. Storing only when both models break down is what gives the method its data reduction while keeping some ability to follow trends.
The characteristic weakness of boxcar and backslope is the stepped, staircase look of the reconstructed trend. Because the boxcar test measures everything against a single held value, the method has a strong bias toward representing the signal as flat segments that jump to a new level when the tolerance is finally broken. On a slowly drifting signal this can turn a smooth curve into a series of plateaus and abrupt steps that do not match how the process actually behaved.
The backslope test softens this by tracking slope, but the pairing is still cruder than swinging door because neither test maintains a two-sided tolerance corridor around a fitted line. Swinging door proves that a single straight line stays within tolerance of every intervening point before it discards them, which yields smooth, faithful ramps. Boxcar and backslope make two separate one-sided comparisons and can commit to a flat or projected segment that later turns out to have hidden a real excursion within the tolerance.
There is also an error-bound difference that matters for trustworthy data. Swinging door guarantees the reconstructed value is within the deviation of the true signal at every point in between. The boxcar and backslope pair does not offer as clean a guarantee, so the reconstructed trace can accumulate distortion in ways that are harder to reason about. That combination of visible stepping and weaker error bounds is why the industry moved on.
This method is worth knowing because it lives on in older historian archives and in some field devices and RTUs that still reduce data before sending it upstream. When a team migrates an old plant to modern monitoring, the historical trends may carry the stepped signature of boxcar and backslope, and engineers who recognise it will not mistake those steps for real process behaviour. It is a fingerprint of the compression, not of the signal.
For oil and gas operators consolidating decades of history, the practical lesson is that not all archived data is equally faithful. Data that passed through boxcar and backslope was reduced with a coarser model than a modern swinging door historian would use, so fine excursions in that old data may simply not be there to recover. Knowing the compression method behind an archive helps set realistic expectations for what analysis it can support.
Merobix applies modern slope-aware compression when it ingests field data into the cloud, so new history is stored with a clean, bounded error rather than the stepped artifacts of the older two-test approach. When operators bring legacy trends alongside live data on the same dashboard, they can see the difference in fidelity directly, which makes it easier to trust the new archive while treating the stepped old segments for what they are.
Boxcar compares each new reading to the last stored value and discards it if it is close, effectively assuming the signal holds flat. Backslope instead projects the previous slope forward and discards the reading if it stays near that projected line, so it follows ramps. In the combined method a point is only stored when both tests fail.
The boxcar test measures against a single held value, which biases the reconstruction toward flat segments that jump to a new level only when the tolerance is finally exceeded. That creates the staircase look. It also lacks the two-sided tolerance corridor of swinging door, so it cannot guarantee a smooth, bounded fit through the discarded points.
It has largely been replaced by swinging door in modern historians because swinging door gives smoother reconstruction and a cleaner error bound. However, older archives and some legacy field devices and RTUs still contain data reduced this way. Recognising its stepped signature helps engineers avoid mistaking a compression artifact for real process behaviour.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.