A PI historian that ran fine at 5,000 tags starts choking at 50,000. Trend queries that used to return in seconds now spin for a minute. Archive files balloon well past what anyone budgeted for storage. None of this happens because PI itself is a bad product — it happens because tag configuration, compression settings, and archive management rarely get revisited as a historian grows, until performance quietly degrades to the point where engineers stop trusting the trends they're looking at. See how iFactory restores PI historian performance with a Book a Demo.
Your PI Historian Isn't Broken. It's Just Never Been Tuned.
iFactory rationalizes tag configuration, tunes compression settings, and restructures archive management so your PI historian returns fast, trustworthy trend data again — without a forklift migration to a different platform.
The Warning Signs Of A PI Historian That Needs Attention
Historian degradation is gradual, which is exactly why it's easy to miss until it becomes disruptive. Most plants notice one or two of the symptoms below long before they connect them to an underlying tag management or archive configuration problem.
| Symptom | Usual Root Cause |
|---|---|
| Trend queries take noticeably longer than they used to | Archive fragmentation or an oversized query spanning too many uncompressed tags |
| Archive storage growing faster than tag count would explain | Compression settings left at default rather than tuned per tag type |
| Duplicate or near-identical tags for the same physical point | No tag naming standard enforced as new equipment and projects were added over time |
| Engineers no longer trust a given trend without double-checking it | Undetected compression artifacts or stale tags no longer receiving live data |
| New tag requests take days to configure correctly | No documented tag creation standard, so every request is handled ad hoc |
Every One Of These Symptoms Has A Fixable Root Cause.
iFactory audits your PI historian's tag configuration, compression settings, and archive structure, then fixes what's actually causing the slowdown.
Every Tag Should Justify Its Own Existence
Tag counts in a mature PI historian tend to grow indefinitely, because adding a new tag is easy and nobody is ever assigned to remove tags that stopped being useful years ago. Tag rationalization is the process of reviewing every tag against a simple set of questions to decide whether it should be kept, merged, or retired.
Is This Tag Still Receiving Live Data?
Tags connected to decommissioned equipment or retired instruments often keep collecting stale or flat-lined values long after anyone stopped relying on them for anything.
Does It Duplicate An Existing Tag?
The same physical measurement point sometimes gets configured twice under different names by different engineers over the years, doubling storage and query load for no analytical benefit.
Does It Follow The Naming Standard?
A tag that doesn't follow the plant's naming convention is effectively invisible to anyone searching the historian by asset or process area, even if the underlying data is perfectly valid.
Is The Scan Rate Appropriate?
A slow-changing tank level tag scanning every second wastes storage just as surely as a fast transient pressure tag scanning too slowly loses the detail an investigation would need.
Exception and Compression Deviation — The Two Settings Most Historians Get Wrong
PI's exception and compression algorithms decide how much of the raw incoming data actually gets written to the archive, and both are controlled by deviation settings specified per tag. Left at default values across every tag regardless of its behavior, these settings routinely either store far more data than necessary or, worse, throw away detail that later turns out to matter.
Exception Deviation
Controls how much a value must change at the source before it's even sent from the interface to PI. Set too tight, it floods the system with noise; set too loose, it can miss meaningful process changes entirely.
Compression Deviation
Controls how much a value must change before it's actually written to the archive, using a swinging door algorithm that keeps points forming a straight-line trend from being stored redundantly.
Per-Tag-Type Tuning
A vibration tag, a temperature tag, and a discrete on/off status tag each need fundamentally different deviation settings; applying one blanket default across all of them is the single most common cause of both storage bloat and lost data fidelity.
Sizing, Splitting, And Maintaining Archive Files So Queries Stay Fast
PI stores data in archive files that are created and closed on a schedule, and how those files are sized and maintained has a direct impact on query performance, particularly for trends that span long historical ranges. The comparison below covers the archive strategies most commonly used and when each one applies.
| Strategy | Typical Archive Duration | Best Fit |
|---|---|---|
| Fixed Small Archives | Daily or weekly | High tag-count systems where query speed on recent data matters most |
| Fixed Large Archives | Monthly or quarterly | Lower tag-count systems where minimizing archive file count is the priority |
| Auto-Sized Archives | Variable based on data volume | Systems with uneven tag activity across process areas or seasonal load variation |
Regardless of the sizing strategy chosen, archives need periodic defragmentation and shift scheduling to keep the most actively queried data in the fastest-accessible files, which is one of the maintenance tasks most likely to be skipped once a historian has been running smoothly for a while.
Our PI system had grown to over 40,000 tags over twelve years with no tag review ever conducted, and trend queries that used to take two seconds were taking almost a minute. After iFactory ran a tag rationalization pass, we retired nearly 9,000 stale or duplicate tags and retuned compression deviation settings across the rest. Query response times came back down to what they were years ago, and our archive growth rate dropped by more than half.
Frequently Asked Questions
Q: Will optimizing our PI historian risk losing any of our existing historical data?
No. Tag rationalization and compression tuning are applied going forward and do not delete or alter data already written to existing archives. When a tag is identified as a true duplicate or genuinely obsolete, it is typically marked inactive rather than deleted outright, preserving the historical record while stopping any further unnecessary data collection. The archive restructuring work is similarly non-destructive, focused on defragmentation and reorganization rather than removal of existing records. Reach out through Support Contact to discuss the specific safeguards used during an optimization project.
Q: How do you decide which compression deviation setting is right for each tag without breaking existing trends?
The process starts by analyzing each tag's actual historical behavior — how much it naturally varies, how fast it changes, and what precision downstream reports and alarms actually require — rather than applying a single formula to every tag uniformly. Tags feeding safety-critical alarms or precise regulatory reporting are tuned conservatively to preserve full fidelity, while slow-changing or low-criticality tags can tolerate more aggressive compression without any meaningful loss of analytical value. Every proposed setting change is validated against historical trends before being applied, so no tag's usable trend shape changes unexpectedly.
Q: Our historian has grown organically for over a decade with no documentation. Can it still be rationalized?
Yes, and this is actually the most common starting point for a rationalization project rather than an unusual edge case. The process begins with an automated inventory of every existing tag, cross-referenced against actual data collection activity, naming patterns, and where available, the CMMS asset register, to reconstruct which tags are genuinely active, which are duplicates, and which correspond to decommissioned equipment. A Book a Demo can walk through what this inventory process looks like for a historian with no existing documentation.
Q: Does tag rationalization require taking the historian offline or interrupting live data collection?
No, tag rationalization and compression retuning are applied while the historian continues collecting live data, since the changes affect configuration settings rather than the underlying PI server availability. Archive restructuring and defragmentation work is scheduled during lower-activity periods to minimize any performance impact on live queries, but does not require a full historian outage. Plants typically see this as a rolling optimization process applied in phases across tag groups rather than a single disruptive cutover event.
Q: How often should tag rationalization and archive optimization be repeated once it's done the first time?
Historian health is not a one-time fix; new tags get added continuously as equipment changes and projects are commissioned, and without a maintained naming standard and periodic review, the same drift that caused the original problem will gradually reappear. Most plants that have gone through an initial rationalization project schedule a lighter review annually, alongside enforcing the naming and configuration standards established during the initial project for every new tag added going forward. Discuss an ongoing historian health program with a Book a Demo.
Get Your PI Historian Back To Fast, Trustworthy Trend Data.
iFactory rationalizes tags, tunes compression, and restructures archives so your PI historian performs the way it did when it was new.
