Every rule, written down.

The one claim being made

At its strongest — a Tier 1 result — BecSpec is saying exactly this:

That is a statement about the register. It is not a statement that the products will behave the same way in the paddock. Rates, water rates, adjuvant requirements, withholding and re-entry periods, plant-back intervals, crop safety and resistance group codes are on the approved label, which is not part of this dataset. Two products can be identical on every field BecSpec can see and still carry different directions for use.

The three tiers

Every product registered for the host and pest you selected falls into exactly one tier. Each tier is a weaker claim than the one before it, and they are never merged into a single ranked list — the difference between them is the most important thing on the page.

Tier 3 is an answer to a different question. Tiers 1 and 2 answer “what else is this same product?”. Tier 3 answers “what else is approved for this job?” — those are alternatives for the task, not substitutes for the product, and the interface says so above every result in it.

Making active constituents comparable

Matching on the raw PubCRIS constituent code is not safe enough, and it fails in the direction that costs the user: it declares obvious generics to be different chemistry and hides them. Three faults in the published export, each verified against the committed file:

• One chemical, two codes. Fifteen active names in scope map to more than one constituent code. S-metolachlor is both ALMEC1 and ALMEC3; cyproconazole is both TZCP and TZCP1.
• One chemical, two spellings. “TRICLOPYR PRESENT AS THE BUTOXYETHYL ESTER” and “TRICLOPYR BUTOXYETHYL ESTER” are separate codes for one substance. Eighteen such pairs exist in scope, including glyphosate, 2,4-D, MCPA and dicamba salts.
• Arithmetic noise. Strengths are exported as floating point, so one product records tebuconazole at 49.799999 g/L where its twin records 49.800000. Another records its unit as g /kg where the rest of the file uses g/kg. Left alone, these split identical products into different groups.

So constituent names are uppercased, whitespace collapsed, and the “present as” phrasing folded away, so that one substance recorded two ways becomes one active. Amounts are rounded to six significant figures — far finer than any registered strength is specified to, and coarse enough to absorb the representation error. Exactly duplicated constituent rows are dropped; the export contains 39 of them in scope, including one product carrying the same clethodim row twice.

What is deliberately not merged

This is where a matching engine becomes dangerous, so the line is drawn early and stated plainly.

• Salts and esters are never collapsed onto the parent acid. Glyphosate isopropylamine salt, potassium salt and mono-ammonium salt stay three distinct actives. So do the 2,4-D and MCPA esters. They are registered separately, formulated differently and applied at different rates.
• Units are never converted. A product at 500 g/L is not compared to one at 500 g/kg. The conversion needs a density this dataset does not carry.
• No fuzzy matching on chemical names. Two names that merely look similar are two different actives. There is no edit distance, no stemming, no “close enough”.
• Formulation type is compared as recorded text, not interpreted. Two products both described as an emulsifiable concentrate can still differ in the co-formulants and surfactants nobody publishes.

Whole-label coverage

Two products can be identical chemistry and both registered for the job you picked, while one is registered for a third of everything else you use the product for. That distinction decides whether an alternative is actually useful to you, so it is calculated and shown on every Tier 1 and Tier 2 result:

A worked example from the register: Boxer Gold is registered for 75 host and pest combinations. One product with exactly the same prosulfocarb and S-metolachlor chemistry covers all 75; another covers 55. Both are honest Tier 1 answers to “can I replace it for wheat and annual ryegrass?”. Only one is an honest answer to “can I replace it everywhere?”. That is why the two questions are a switch on the results, not a single blended score.

Tier 3 carries no coverage figure. Computing it would mean pulling the registered-use list of hundreds of products to rank a list the user has just been told is not a substitute, and a percentage sitting beside a different active constituent invites exactly the reading the tier exists to prevent. Tier 3 is alphabetical, with no implied ranking.

Who counts as a different supplier

By default, products from the same registration holder are hidden — presenting a company's own product back as an alternative to it is noise. One firm appears in the export as both “NUFARM AUSTRALIA LIMITED” and “NUFARM AUSTRALIA LTD”, so holders are compared after legal suffixes are stripped. That rule can only ever hide a result, never invent one: at worst two unrelated firms sharing a root name are treated as one company and a real alternative is withheld until you turn the filter off.

The word supplier is doing limited work here. PubCRIS records who holds the registration. It does not record who manufactures the product, who distributes it, which reseller has it on the shelf, or what any of them charge.

What is in scope

Currently registered agricultural products only — herbicides, insecticides, fungicides, miticides, molluscicides, nematicides, plant and growth regulators, algicides, vertebrate poisons, mixed-function pesticides and mixed-function seed treatments. Cancelled and expired registrations are excluded, as are approved active constituents, which are registration records rather than products anyone can buy.

Veterinary medicines, pool chemicals, dairy cleansers, disinfectants and household insecticides are in the same source file and are deliberately left out. The alternative-product question is real for them too, but “same actives, same strength, same formulation” carries very different weight for an injectable than for a knapsack spray, and each deserves its own treatment rather than being swept in.

Candidates are not filtered by product type

A miticide can be offered as an alternative to an insecticide, and vice versa, when the chemistry and the registered use both line up. PubCRIS product types are broad classifications, and the same abamectin product appears under different ones depending on the registration. The registered host and pest combination is the stronger constraint and is the one used; the product type is shown on every result so you can see it.