The four words inline, online, atline, and offline are the most commonly confused vocabulary in process analytical technology. They are not synonyms for fast, slow, automated, or manual. They describe where the sample is relative to the process stream at the moment of measurement. Getting the distinction wrong in a user requirements specification, a validation protocol, or a vendor request for quotation can change which instrument is appropriate, which qualification protocol applies, and how regulators read the submission.

This piece is a short reference for engineers and quality assurance staff who write or review documents that hinge on these terms. The four definitions come from ASTM E2363, the standard terminology document referenced by the U.S. Food and Drug Administration’s 2004 PAT guidance and by ICH Q8 and Q14. The wording below paraphrases ASTM E2363; consult the standard itself for the exact text used in regulatory correspondence.

Where the definitions come from

ASTM E2363, first published in 2006 and reaffirmed most recently in 2021, fixes the vocabulary used by pharmaceutical regulators and pharmacopeias for sample placement. The FDA’s 2004 PAT guidance uses the four terms in its discussion of measurement modes but does not formally define them; ASTM E2363 closed that gap. ICH Q8(R2), Q13, and Q14 reuse the same four terms when they refer to process measurements.

The standard also defines adjacent vocabulary that is often discussed alongside placement - invasive versus non-invasive, destructive versus non-destructive, univariate versus multivariate - but those describe other axes of a measurement and should not be conflated with placement. A measurement can be inline and non-invasive (a transmission probe through a sight glass), or inline and invasive (an immersion probe in a reactor), or offline and non-destructive (a benchtop near-infrared spectrum of a withdrawn tablet).

The four placements

Inline

Inline means the sample interface sits in the process stream. The analyzer measures the process directly; no portion of the material is removed. An immersion Raman probe in a crystallizer is inline. A flow-cell near-infrared probe inserted into a tablet press feed frame is inline. A pH sensor in a bioreactor is inline.

Inline measurements give the shortest possible loop between the process and the analyzer, which is why they are the placement of choice when sub-second to few-second cadence is required for feedback control. They also impose the strongest constraints on the probe: it must tolerate the process environment - pressure, temperature, cleaning regimes, sterilization cycles - for the duration of a campaign.

Online

Online means the sample is diverted from the process stream into a measurement loop and may be returned to the process. The measurement is on a representative withdrawn portion, not on the bulk process material. A slipstream that pulls a few millilitres per minute through a flow cell and recycles back to the reactor is online. An automated dilution-and-measure cycle that sends a sample to an analyzer and discards the diluted aliquot is also online if the sampling cadence is automated and continuous.

The boundary between inline and online is where most misuse happens. A probe physically inserted into the bulk process is inline; a flow cell fed by a sampling pump is online, even if the cadence is one measurement per second. The placement reflects where the measurement happens, not how fast it happens.

Atline

Atline means the sample is removed from the process and analyzed in close proximity to the process stream, but not in the stream itself. A benchtop Raman next to a tablet press, fed by an operator pulling a tablet every five minutes, is atline. A handheld near-infrared analyzer used by a shift operator on a sample drawn from a reactor sample port is atline. The defining feature is physical proximity to the process combined with a discrete sample handling step.

Atline measurements typically have longer cadence than online ones - minutes rather than seconds - because the handling step is manual or semi-automated. They are common where the process tolerates the cadence but the analytics do not justify a permanent probe installation: small campaigns, multi-purpose plants, and feasibility-study deployments.

Offline

Offline means the sample is removed from the process and analyzed in a separate location, typically a quality-control laboratory. The cadence is whatever the lab can support, the analytical methods are whatever the lab has qualified, and the measurement may take hours or days to feed back to the process. A high-performance liquid chromatography assay run in the QC lab on a retained sample is offline. A reference near-infrared measurement run on a calibration sample in a laboratory hutch is offline.

Offline is the default for release testing and for reference methods used to calibrate inline or online analyzers. It is not a failure mode of process analytics; it is a deliberate choice when the analytics required cannot be reproduced near the process.

Why the boundary matters

The four placements differ in three ways that matter for project scoping:

  • Cadence and loop time. Inline can support sub-second feedback; online supports seconds to minutes; atline supports minutes; offline supports hours or longer. A control strategy that requires per-second adjustment cannot be served by an atline analyzer even if the analytics are identical.
  • Sample representativeness. Inline measures the bulk; online measures a diverted portion that may not represent the bulk if the diversion is non-isokinetic or if the loop dwell time alters the sample; atline and offline depend on the discipline of the sampling procedure. ICH Q13 spends considerable text on sample representativeness for continuous processes.
  • Qualification burden. An inline probe in a pharmaceutical reactor falls under the equipment qualification regime; an atline benchtop falls under a different one; an offline lab method falls under analytical procedure qualification per ICH Q14. The placement decision drives which validation protocol the project needs.

These three axes are why the placement choice belongs in the user requirements specification, not in the implementation phase. The URS template we publish has placement as one of the first questions because every downstream design decision depends on it.

How placement maps to instrument families

Most analytical techniques can be deployed across all four placements, but with different ease.

  • Raman and near-infrared spectroscopy map cleanly to all four. Probes are commercially available for inline immersion; flow cells exist for online; benchtops support atline; laboratory units support offline. The Raman versus NIR decision framework discusses the trade-offs.
  • Mid-infrared (FTIR) is more constrained. Attenuated total reflectance probes support inline and online; benchtop FTIR is common offline; atline FTIR is feasible but less common.
  • Mass spectrometry is rarely inline. Online membrane-inlet and slipstream configurations exist; atline and offline are typical.
  • Chromatography is almost always offline or atline. Online HPLC exists but imposes substantial engineering for sample handling.
  • Particle sizing (focused-beam reflectance, laser diffraction) is commonly inline or online; less commonly atline.

When a vendor describes an instrument as “process” or “process-ready,” they usually mean it has a probe or flow cell that supports inline or online placement. That is a marketing claim; the underlying qualification work is the user’s.

Adjacent terms worth keeping straight

A few neighbouring terms get pulled into placement discussions and should not be:

  • Real-time describes the time relationship between measurement and decision, not placement. An offline laboratory method can produce a result fast enough to be considered timely for a slow process; an inline probe with a slow chemometric model can be slower than its placement implies. Spectrane style prefers explicit cadence figures (per-second, per-minute, sub-second) over the word real-time.
  • Non-invasive describes whether the probe contacts the sample, not whether the probe is in the process stream. A sight-glass transmission measurement is inline and non-invasive. A wetted immersion probe is inline and invasive.
  • In-process is a regulatory phrase used by FDA and EMA for measurements taken during manufacture; it is broader than ASTM’s inline. An atline measurement of a tablet during a press run is in-process but not inline.

Practical guidance for writing about placement

For specifications, validation documents, and journal articles, three habits prevent most of the confusion:

  1. State the placement explicitly using the ASTM term. “An inline Raman probe in the crystallizer” is unambiguous; “real-time Raman monitoring” is not.
  2. Specify the cadence numerically. “One measurement per 2 seconds” tells the reader more than “real-time,” and the cadence does not change with placement.
  3. Cite ASTM E2363 once when the placement vocabulary is introduced, so the reader knows which definitions are operative. The standard is short and the citation costs nothing.

A working vocabulary is not a stylistic preference. It is the shared reference frame that lets engineers, quality assurance staff, regulators, and vendors discuss the same project without translating each other’s words. For PAT specifically, that frame is ASTM E2363 plus the four placements above. For more on how the surrounding regulatory terminology evolved, see our reading guide on PAT vocabulary and the primer on PAT itself.