How to Design a Meaningful Saggar Qualification Trial

2026-10-15

A trial is supposed to answer one question: can this saggar work in my process?

Too often, it answers none.

Samples arrive, a few cycles run, nothing visibly breaks, and the trial closes with a vague conclusion — "no problems so far." Three months later the same issue appears in production, and nobody can say whether the saggar was ever actually qualified or simply tested.

The gap is rarely the test itself. It is the design around it. A qualification trial becomes usable evidence only when six elements are deliberately set:

  1. The question — what decision the trial result will support
  2. Variable control — what is held fixed, what is varied
  3. Controls — what the result is compared against
  4. Sample size — how many pieces the question requires
  5. Duration — how long the trial runs
  6. Acceptance criteria and records — what counts as a pass, and what is written down

This guide walks through those six elements. It does not tell you what result to expect — that depends on your material, your furnace, and your parts. It tells you how to design a trial whose result you would be willing to defend.

Why Matching Production Conditions Is Necessary but Not Sufficient

Matching your production conditions is the obvious requirement, and it is not negotiable. A trial run at conditions your process never sees tells you nothing about your process.

But it is frequently treated as if it were the whole requirement. Trials are then run under production conditions with no comparison group, on a single piece, for two cycles, with no written acceptance criteria — and the result is still recorded as "passed."

Condition matching establishes relevance. It does not establish evidence. Six elements together do that, and each one answers a different gap:

  • The question — a trial that cannot be answered yes or no cannot conclude anything
  • Variable control — if several things changed at once, no result can be attributed
  • Controls — a number has no meaning without something to compare it to
  • Sample size — one piece cannot show consistency
  • Duration — time-dependent problems are invisible in short trials
  • Acceptance criteria — a judgement made after seeing the data is not a criterion

Remove any one of them and the remaining five do not compensate. That is why condition matching, while mandatory, is only one requirement among six.

Element 1 — The Question the Trial Must Answer

Start with the decision, not the test.

Write down what you will do differently depending on the outcome. "Test sample A for several cycles and observe wear" is an activity. "Adopt the tested material for this line instead of maintaining the current incumbent tooling" is a decision.

The second statement can be answered. The first cannot.

A workable question has three properties:

  1. It names the decision. The trial result feeds a specific choice.
  2. It is answerable yes or no. Not "how did it perform" but "does it meet the stated criterion."
  3. It is scoped to your conditions. Your firing cycle, atmosphere, part geometry and throughput — not a general claim about the product.

If you cannot write the question in one sentence with a yes-or-no answer, the trial is not ready to be designed.

What the question determines

The question sets everything downstream. What you must hold fixed, what the comparison group is, how many pieces you need, how long it runs, and what you must measure — all of it follows from what you are trying to decide.

A trial without a fixed question tends to accumulate measurements instead of producing an answer. More data does not compensate for an undefined question.

Element 2 — Variable Control

Once the question is fixed, list every condition that could affect the result. Then divide them into two groups: those held constant across all test pieces, and those deliberately varied.

Conditions that must be held fixed typically include:

  • Furnace and position within the furnace
  • Atmosphere composition and flow
  • Ramp profile, hold temperature and dwell time
  • Loading pattern and part placement
  • Number of cycles accumulated before the trial

Conditions you may deliberately vary are usually limited to one or two — most often the saggar type itself.

The single-variable rule

Change one group at a time. If the trial simultaneously changes saggar type, cycle count and loading pattern, then a result — good or bad — cannot be attributed to any of the three. The trial will need to be repeated to resolve each, and usually it will not be.

Documenting the fixed set

Write down the fixed conditions and keep them in the trial record. This serves two purposes. It lets someone else repeat the trial, and it lets you detect drift when a later trial is compared against this one.

"Fired as usual" is not a record. "Held at the production ramp profile, dwell as per the standard cycle, loaded in the standard pattern" is.

Element 3 — Controls

A measurement without a reference point is just a number. The control group is what gives your test result meaning.

Why a comparison needs a reference

If you measure a new saggar after several cycles and it looks acceptable, acceptable compared to what? The control answers that. It converts "seemed fine" into "measured against the incumbent, the difference was within the stated range."

Using the current production setter as control

The most useful control in production is usually the material you already run. It carries the same furnace, the same atmosphere, the same parts and the same operators — so a difference between control and test is attributable to the variable you changed.

Two practical points:

  • Control and test pieces should be handled and fired together. Separating them reintroduces the variables you worked to remove.
  • The control must be a comparable geometry. A control piece with different dimensions or capacity is not a control.

When a control is not available

Sometimes no incumbent material exists — a first qualification, or a line change. In that case, establish the reference differently:

  • Test two candidate materials against each other, so at least one comparison exists
  • Run the same new material in two positions or two furnaces to expose position and furnace effects
  • Keep a set of unused pieces from the same lot as a baseline for before-and-after inspection

Without any reference, the trial can still produce observations. It cannot produce a qualification conclusion. Record that limit explicitly rather than letting the result read as a pass.

Element 4 — Sample Size

A single piece proves that one piece performed as observed. It cannot show whether the next piece will behave the same way — and variation between pieces is exactly what qualification is supposed to detect.

Choosing a count that matches the question

There is no universal sample count for saggar trials, and any supplier or article offering one is offering a convention, not a rule. The appropriate count depends on what your question is:

What you are trying to establish What the sample count must cover
Whether a failure mode occurs at all Enough pieces that the mode would plausibly appear if it is present
Whether performance is consistent Enough pieces to observe variation, not just a single outcome
Whether a specific dimension is stable Enough pieces to see the spread of that dimension
Whether a change improves on the incumbent Enough pieces on both sides for the comparison to mean anything

Each row points to a different requirement, and none of them can be read off a general rule.

What a small sample can and cannot establish

A small sample is legitimate. What it cannot do is support a claim about consistency or reliability. If you test three pieces and all three look good, the honest statement is "three pieces showed no visible defect under the stated conditions" — not "the saggar is reliable."

Reporting a small-sample result honestly

State the count in the record, every time. Write what was observed, at which cycle, by what method. State what the result does not establish. A record that reads "piece count as stated, cycles as stated, no visible cracking observed at the stated inspection points, consistency not established" is a usable trial. A record that says "passed" is not.

Element 5 — Duration

Operating strictly under production cycle profiles does not guarantee detection, because many critical failure modes develop cumulatively over time rather than during initial cycles.

Matching duration to the failure mode

The useful question is: how long does the failure mode take to appear?

If the problem you are investigating appeared after extended use in production, a trial of a few cycles cannot detect it — you would be running the test for a shorter time than the problem takes to develop. Short trials systematically understate time-dependent problems, and they do so in a predictable direction: they make questionable material look acceptable.

This is why duration should be justified by the failure mode under investigation rather than by convenience. If the mode is time-dependent, the trial has to run long enough to give it a chance to appear. If the mode is immediate — cracking on first loading, reaction at the setting surface — a short trial is sufficient and a long one adds cost without adding information.

Interim inspection points

Planned intermediate inspections serve a specific purpose: they let you observe a trend rather than only an end state. Define what is measured, at which cycles, and how it is recorded, before the trial starts.

Two cautions. First, an inspection that involves handling the pieces can itself introduce effects — so keep handling consistent across the group. Second, do not extend a trial because the current result looks good. The duration should be fixed in advance, for the reason stated above, or the trial becomes open-ended.

Element 6 — Acceptance Criteria and Records

The last element is the one most often skipped, and it is the one that determines whether the trial result can be used later.

Agreeing criteria before the trial starts

Criteria set after seeing the data are not criteria — they are rationalisations. Write them down first, and write them so that a given observation either meets them or does not.

Each criterion needs three parts:

  1. What is measured — a specific characteristic, in stated units
  2. How it is measured — instrument, method and measurement points
  3. What result counts as acceptable — a stated boundary, agreed before the trial starts

For the boundary itself, two rules matter. First, it must come from your own process requirements — the tolerances your downstream steps or part specifications impose — not from what a sample happens to show. Second, where your internal data does not define that boundary, state the boundary as unresolved rather than adopting an outside value. A figure that looks authoritative but has no established basis behind it will be quoted back as if it were an industry requirement, and it will eventually be used to settle an argument it was never fitted to settle.

What to measure

Characteristics commonly recorded in saggar trials include:

  • Dimensional change at agreed measurement points
  • Mass change
  • Surface condition and any deposit on fired parts
  • Warping or contact-face flatness
  • Crack initiation and progression

Define the measurement points and the instrument before starting, and use them unchanged for every piece in the trial.

A record format that survives beyond the trial

The record should be readable by someone who was not present, months later, with no memory of the trial. Alongside the fixed conditions and the varied conditions already listed under Element 2, it needs:

  • The trial question and the decision it supports
  • Piece count, piece identification and where each piece was placed
  • Results in stated units, with the measurement method noted
  • The acceptance criteria as agreed in advance
  • What the result does not establish

If the trial generates a problem, this record is what allows the cause to be identified rather than guessed at.

What a Trial Cannot Tell You

A well-designed trial still has limits, and stating them is part of the result.

Outside the trial envelope, nothing is established. The trial covers the conditions it ran under. Behaviour at higher temperatures, different atmospheres, heavier loads, longer cycles or different part geometries is not covered by that result.

Time-dependent behaviour remains open if the trial was short. A trial that ran fewer cycles than the failure mode requires has not tested that mode.

A result on one material does not transfer to another without separate verification. Results obtained on a different supplier, a different grade or a different geometry would need their own trial before they can be relied on, even within the same product family.

A trial cannot validate future production. Tooling wears, recipes drift, operators change. A qualification result describes the conditions at the time it was run.

A trial does not replace process control. If your own furnace conditions vary widely, a saggar trial will absorb that variation and produce an inconclusive result regardless of its design.

Summary

A qualification trial becomes evidence through six elements, not through condition matching alone:

  1. The question — written so it can be answered yes or no
  2. Variable control — what is held fixed, what is varied, documented
  3. Controls — a reference to compare against, or an explicit note that none was available
  4. Sample size — chosen to match the question, reported with the count every time
  5. Duration — justified by the failure mode, not by convenience
  6. Acceptance criteria and records — agreed in advance, traceable afterwards

Set these before the trial starts, keep the record to a standard that survives the trial, and state plainly what the result does not establish. The value of a trial is not that it passed. It is that the result can be defended later.

Explore Saggar Options for Your Process

A meaningful qualification trial starts with a clear understanding of your firing conditions, material requirements, and acceptance criteria. Explore our mullite saggar and molded graphite saggar product pages, or contact us to share your application requirements and discuss suitable options.