I Run a Candle Business with 20 SKUs — How Do I Decide Which Raw-Material Changes Require a Quick Confirmation Test Versus Complete Revalidation?

The four tiers, and what each one tests

A small candle business that runs 20 SKUs changes something almost every month. A wax bag arrives in a different size. A fragrance is out of stock. The jar supplier sends a slightly different mould. Testing every SKU in full every time is impossible, and testing nothing is how a range quietly drifts until a customer complains. The way out is to sort changes into tiers before anyone pours a test candle, and to let the tier decide the depth of the test, the number of candles and who signs the release.

The tier system below uses the vocabulary every page in this series shares. A test result you already hold is KEEP when the change cannot plausibly touch it, CONFIRM when a short side-by-side check against a control candle is enough, REPEAT when the full test must be run again, and NEW when the change creates a question your old file never needed to answer. A tier is simply a pre-agreed mix of those four words.

Two design choices sit under this table. First, the number of candles per variant — three new and three controls in the deeper tiers — is a working practice, not a standard. Three lets you see whether one odd candle is a fluke; one lets you notice a gross problem and nothing more. Second, a Tier 3 change does not mean every SKU gets a full test. It means the bracket SKUs — the ones most likely to fail — get the full test, and every other SKU gets a confirmation. How to choose brackets is further down, and it is where most of the saving comes from.

Triage: scoring a change in two minutes

A tier table only works if everyone puts the same change in the same tier. The scorecard below makes that repeatable. Answer five questions, score each 0, 1 or 2, add them up, and read the tier. It is a working scheme we built for this page, not an industry standard — adjust the thresholds once you have a few changes on record.

Run a few real changes through it and the logic becomes obvious. A new order of the same wax, in the same bag size, from the same listing scores 0 + 0 + 2 + 0 + 0 = 2 on the example range (it touches all 20 SKUs) — Tier 1 on the strict reading. Many businesses decide that a same-listing, same-pack reorder should sit at Tier 0 with a retained sample, because it is the most frequent event they have; that is a reasonable local rule as long as it is written down. What is not reasonable is deciding it case by case while the pouring jug is warm.

Why score reversibility? Because the same change is a different risk depending on your shelf. A wick swap with 400 old wicks still in the drawer can be undone the moment a test candle misbehaves. The same swap after the last old wick has gone into a candle cannot. The first is a test; the second is a commitment, and commitments need deeper evidence before they are made. Pre-qualifying backups is how you keep that fifth question at zero.

Applying it to a 20-SKU range

The title gives 20 SKUs and nothing else, so this page uses a labelled example range: ten fragrances, each poured in two formats. One format is a 200 g candle in CSI's White Matte Glass Jar (its fill weight is not published — confirm 200 g at your fill line with a weighed pour). The other is CSI's Golden Aluminium Tin, which has a published 85 g fill. One wax, one wick type, 8% fragrance load throughout. Your range will differ; what matters is that you can describe it the same way — which SKUs share which materials.

This map is the whole trick. Before you score question 3 you look up the material here, and the SKU count falls out. It also shows why wax and wick changes are the expensive ones: they sit under every candle you sell. A fragrance substitution, by contrast, touches two SKUs and a jar change touches ten. If your range has three wax types or two wick series, draw the map with more rows; the logic does not change. The approved-specification post turns this map into a formal specification per SKU.

Read the last two rows carefully. A wax or wick change touches all 20 SKUs, so it lands in Tier 3 — but the test plan is 56 candles, not 120. Full testing on every SKU would be 20 × (3 new + 3 control) = 120 candles. The bracketed plan is 4 SKUs in full (4 × 6 = 24) plus the other 16 confirmed one-to-one (16 × 2 = 32). That is 64 fewer candles to pour, burn, weigh and record. The next section shows how to choose the four so that the saving does not cost you the evidence.

Bracketing: which SKUs carry the full test

Bracketing means testing the extremes and inferring the middle. If the widest container and the smallest both pass with the new wax, and the fragrance that was hardest to get right passes, the SKUs between them are unlikely to fail — but "unlikely" is why they still get a confirmation candle. The brackets are chosen from your own test records, not from a list of oils someone thinks are risky. CSI publishes no discolouration, throw or wick-behaviour data for any oil, and nor should you assume any.

The material-value line prices every test candle as if it were a production candle: ₹262.80 for a 200 g jar (184 g of wax at ₹0.60/g, 16 g of oil at the example average, wick and jar) and ₹114.99 for an 85 g tin. It excludes labour, electricity and the burn-station time, which on a 120-candle plan is the larger cost — replace it with your own figure. The oil figure uses your existing fragrance stock, which is why the priced kit below buys only the new wax, containers and wicks. The ten example oils run from ₹3.00/g (Citrus Lemon Fragrance Oil) to ₹7.90/g (Dark Vanilla Fragrance Oil) at the 100 g size on the 24 September 2026 pull.

The retest matrix behind each tier

A tier is a shorthand. Underneath it, every result line in your test file gets one of the four words. The matrix below shows the default word for each line at each tier, for the example range. The individual component posts refine it — wax supplier, fragrance supplier, wick and jar — this is the range-level summary a production lead can pin to the wall.

Notice that Tier 0 keeps almost everything and still confirms appearance on the first pour. That is deliberate: CSI does not publish lot numbers, lot-to-lot consistency figures or change notices for any material, and nor do most suppliers to small businesses. The first pour of every new order is your cheapest early warning. Keep a retained sample of each order — a sealed bag of wax, a small bottle of oil, three wicks — labelled with the order date, so a later complaint can be traced back to a delivery.

One protocol for every tier

Whatever the tier, a test candle only means something if it was made and burned the same way as its control. The protocol below is a working practice that you may change, not a published standard: pour the control and the new-material candle on the same day from the same batch of everything else, cure them side by side, and burn them side by side in the same room, away from draughts. The 3–4 hour burn cycle with a full cool-down between cycles is a commonly used cycle; the 48-hour, 7-day and 14-day checkpoints are labelled examples.

The CONFIRM rows only ask whether the pair looks the same. That is the right question for a one-to-one check: with a single candle on each side you cannot separate a small real difference from candle-to-candle variation, but you can spot a clear one. When you spot it, the SKU is promoted to a full 3 + 3 test and becomes a bracket in its own right.

Pass, fail or escalate — the decision rule

A tier system collapses if the decision at the end is a matter of taste. Write the rule down once and apply it to every change. The rule below is the one this series uses. Its first line is not negotiable: no worse than the approved control on every safety line — flame, soot, jar heat, end-of-life behaviour — and within your jar supplier's guidance and the recognised candle fire-safety standard you work to (ASTM F2417 is a commonly cited US standard; this page quotes no numbers from it, and CSI publishes no numeric flame-height or jar-temperature limit).

Priced kit for a Tier 3 wax change

The heaviest event on the example range is a wax change, so here is its kit. The candidate is Luxury Soy Wax CSI 464, rated by its wax maker up to 10% fragrance load — the example range runs at 8%, so no formula has to move to fit it. If any of your SKUs run above 10%, that SKU cannot simply be moved to CSI 464: it becomes a formulation change and a NEW line. The kit buys only what the test needs new: the new wax for 28 new-material candles, and jars, tins and wicks for all 56 candles (the controls use your existing wax and all candles use your existing oils). A 5% pour-loss allowance on the wax is a labelled working assumption.

That is ₹5,587.60 for the bracketed plan. The same change tested in full on all 20 SKUs would need 8,259.3 g of new wax (1 × 5 kg + 3 × 1 kg + 1 × 500 g, ₹4,319.20), 60 jars, 60 tins and 120 wicks — about ₹11,075.20 in purchases, and more than twice the burn-station time. The wax bags are priced at the listed rate: CSI 464 is ₹507.40 a kilo at 1, 5 and 10 kg, so there is no bulk break to chase, and four 1 kg bags cost less than one 5 kg bag here because 4 kg covers the need. Controls are poured from your current approved wax; if that wax is already gone, you have no true control and every SKU must be treated as NEW — a retained finished candle from an earlier batch is the fallback reference.

Release, record and route

The tier also sets the paperwork. A Tier 0 event needs a line in the order log and a retained sample. Tier 1 adds the scoring sheet. Tiers 2 and 3 add a written decision, an updated approved specification for each affected SKU and an updated approved-supplier list. The point of recording is not bureaucracy: in six months, when a customer complains about the lavender tin, you need to know which wax delivery and which wick pack it came from, and whether that combination was ever tested.

Everything else in the series plugs into this hub. The table below routes each kind of change to the page that owns it, including the four earlier cluster hubs on wax, fragrance, wick and jar changes.

One last habit makes the whole system cheaper over time: after each change, write down whether the tier was right. If a Tier 1 confirmation kept turning up differences, that change type belongs in Tier 2 for your range. If Tier 3 wax changes never moved a CONFIRM pair, you may be able to bracket with three SKUs instead of four. The scheme on this page is a starting point; your own records are what make it accurate.

Frequently asked questions