I Have Hundreds of Burn-Test Results Across Different Jars and Fragrances — How Should I Organise the Data so Future Wick Selection Becomes Faster Rather than Starting from Zero?

Why hundreds of results still start you from zero

Many candle businesses that have burned hundreds of test candles have the results, and still bracket every new jar from scratch. The results sit in notebooks, phone photos and one-tab spreadsheets with a different layout each season. "C2 fine, bit hot near end" is a true sentence, but it cannot be filtered, counted or compared with the forty other sentences about C2.

Three things usually go missing. The inside diameter: people record the jar's name or listed size, so results from two jars with the same inside cannot be pooled, and two "8 cm" jars that differ inside get pooled by mistake. The direction of a failure: "failed" does not say whether the next test should go up or down a rung. The difference between a candle and a session: when every burn of every candle sits in one free-text cell, you cannot see that the flame grew in session 5, which is the information a late-burn decision needs.

A proper sheet does not make testing unnecessary. CSI publishes no wick-performance data, and no sheet of yours can approve a system that has not been burned to the end. What it does is move the start line: the next test begins one informed step from the answer instead of three blind ones.

One row per session, one row per candle

A candle has facts that never change once poured: its system, lots, measured jar, pour temperature, cure. A session has facts that change every time: weight, pool, flame, carbon. Putting both in one row forces you either to repeat the candle facts on every line (and mistype them) or to cram sessions into columns (S1 flame, S2 flame… S12 flame), which breaks the moment a candle needs a thirteenth session. Three tables avoid both.

In our worked example the business has 312 candle rows and 1,870 session rows, about 6.0 sessions per candle. That ratio is itself useful: candles with only two or three sessions were screens or early fails, and the pivot should not count them as full-life evidence.

IDs that still make sense next year

Use three keys, each built for one job. The system ID is descriptive, because systems are what you search: J150-464-JAS-08-C1 reads as CSI's 150 g jar, CSI 464, Jasmine, 8%, one C1. Two wicks become 2xC1 and a spacing note goes in its own field. The test ID is sequential, because tests happen in order: T26-041 is the 41st test set of 2026. The candle ID is the test ID plus a letter: T26-041-A, -B, -C. A session is a candle ID plus S01, S02 and so on.

Codes for jars, waxes and oils go on a LOOKUP tab with the full product name, supplier and, for jars, the three measured inside diameters of a sample from each lot. When a jar supplier quietly changes the inside, that tab is where you notice it, and a new lot code is how you keep old results honest; the lot measurement routine is the incoming-lot sheet.

The fields, and the words allowed in them

Every observation field gets either a number with a unit or a code from a short list. Free text goes in one notes column only. The codes below are our recommendation; what matters is that they are fixed and that none collides with a product name. In particular, do not use "C1" and "C2" for carbon levels when C1 and C2 are also your wick names.

The VOID rule comes from the masking checklist: a throw score from a session that failed flame, soot, mushrooming, heat or flame-to-glass is not a score. Storing VOID instead of a number keeps a hot candle's flattering scores out of every average you will ever take. The supplier and lot fields matter when you compare wick sources; the full comparison record is the two-supplier post.

Outcome codes that point the next test

Exclusions are the honesty column. A candle that soots because it sat under a fan tells you about the fan, not the wick, and lumping it in with FAIL-OW would push the next test the wrong way. Keeping it as X-DRAUGHT, rather than deleting it, also shows you how often your bench conditions spoil tests, which is worth knowing.

Three pivot views that answer the next question

Build these from the tables with your spreadsheet's pivot tool. Filter to PASS-FL and the FAIL codes; leave X- codes and PASS-3 out of the evidence counts.

View 1 · Diameter band × oil group → build

Rows are diameter bands from measured inside diameters; columns are oil groups; each cell shows the build with the most full-life passes, as passes over full-life tests. This is the diagram's grid, and it answers "where should a new system start?". In the worked example, light-group oils in the 6.0–6.9 cm band passed full life on C1 in 9 of 9 systems (100%). A band is a filter, not proof that two jars behave alike; how close inside diameters need to be before one wick is shared is the one-wick-across-jars post.

View 2 · Build × band → pass rate and failure direction

For each wick build, count PASS-FL, FAIL-UW, FAIL-OW and FAIL-HT per band. This shows where a build runs out: in the worked example, C2 in the 7.0–7.9 cm band failed seven heavy-group systems, five of them as FAIL-OW or FAIL-HT (71%). That says heavy oils in that band need less heat than C2 or a different build, not more.

View 3 · Open questions

List every band × group cell with no PASS-FL, plus every approved system whose evidence rests on a single candle. These are the cells where the sheet cannot help, so they are where testing time should go first; the "open" cells in the diagram are examples. How grouping cuts the number of cells you need to fill is the grouping post, and how an oil earns its group is the behaviour-groups post.

Worked example: slotting a new oil from evidence

You want to add Jasmine at 8% to the line in CSI's 150 g jar (listed 7 cm outside; measured 6.3 cm inside in this example, so band 6.0–6.9). Its group is provisional: PROV, leaning light on its page description, which is not evidence. View 1 shows light-group oils in that band passing full life on C1 in 9 of 9 systems with no FAIL-UW. View 2 shows no C1 failures for light oils in that band. So instead of a blind bracket, the oil goes through the admission routine set out in the behaviour-groups post: a short screen on the group's wick and one full life.

One C1 candle costs ₹221.92 here (₹51.60 of oil, ₹70.02 of wax, ₹94.40 of jar and ₹5.90 of wick); one on C2 costs ₹225.46. If you add twelve oils a year that land in well-evidenced cells, the saving is about ₹8,116.60 in materials, plus the burn hours. If the screen returns FAIL-UW or FAIL-OW, the code tells you which way to bracket, and the oil's group changes from PROV to whatever its candles say.

How these rows become a published internal guide, with confidence levels and retest triggers, is the next step, in the internal wick-guide post.

Buying for a bench that keeps records

Five items, chosen on page statements, price and stock on 29 September 2026. CSI publishes no wick-performance data, so nothing is ranked on burn results. The records themselves cost nothing but discipline; if you already own a scale and a thermometer you trust, buy only what the next test needs.

Frequently asked questions