Sub-IDs for Split Testing: Ten Slots, Used Properly

Affiliates who do not tag their traffic are not running a business. They are running a mood. Clicks went up, deposits went down, and the argument that follows is religion: the bonus, the GEO, the landing, the day of the week. A sub-id turns that argument into a table.

AFFILIFY gives you ten slots on the tracking link. They are appended when you copy it, they survive the click, and they show up on the conversion. This post is what to put in them, what not to, and how to read the result without lying to yourself.

What a sub-id actually is

A sub-id is a label you choose. It is not a secret tracking technology. It is a field on the click that you fill in, so that when an FTD lands you can still see the label. Source, creative, landing, campaign week, a creator's handle: anything you might want to slice on later.

The link still works if you fill in none of them. That is the trap. The empty link is easier to copy, so it is the one that goes into the bio, the article, the paid campaign and the Telegram channel, and three months later you have one number.

Ten slots is enough to run real experiments and few enough that you can remember what each one means. The failure is not running out of slots. The failure is using slot 1 for four different ideas over four weeks.

A working map, not a religion

You will see ten-slot maps online that pretend there is one correct assignment. There is not. There is a map you write down once and do not change in the middle of a test. Here is a default that works for a site, a paid campaign and a creator programme. Steal it, then lock it.

Slot 5 exists because intended GEO and actual GEO disagree more often than people admit. A campaign you labelled DE that arrives from somewhere else is a different campaign. You will not see that if the only GEO you have is the one on the click.

On AFFILIFY the ten slots are on the link builder. They are appended when you copy. You do not have to hand-edit query strings, and you should not, because a hand-edited parameter is how QA failures start.

How to run a test that means something

One change at a time. If you swap the headline and the landing and the bonus in the same week, the winning sub-id tells you that the bundle won. It does not tell you which part to keep.

Give it enough volume. Five FTDs on hook-a and three on hook-b is weather. Decide the sample before you start: a number of FTDs, or a number of days, and do not peek in a way that lets you stop when you are ahead.

Keep the meaning of a slot stable for the life of the test. Renaming slot 3 from "creative" to "author" halfway through a month produces a report that looks like data and is not.

Read FTDs, not clicks. Clicks tell you which creative was loud. FTDs tell you which creative was paid. If you are on RevShare, read NGR on the same slice after a few weeks, because the creative that deposits fastest is not always the one that stays. CPA versus RevShare is that argument in full.

Five ways to make the report useless

One link for everything. The empty-slot problem. You will feel busy and know nothing.

Overloading slot 1. `telegram-review-v3-br-mobile` as a single string. You cannot slice it. That is what the other nine slots are for.

Changing the code and keeping the same name. Hook-a last week was a screenshot. Hook-a this week is a video. The report will tell you hook-a improved. It did not.

Testing on a broken tag. If the click id does not round-trip, every sub-id in the world still produces zero FTDs. QA first, test second.

Reading the first day. Paid traffic and matchday traffic both spike. A Tuesday morning on a sportsbook book is not a sample of Saturday.

What the stats actually show you

When it is working, a month of tagged traffic answers questions you would otherwise argue about.

• Which Telegram room deposits, and which one only clicks.
• Whether the prelander is earning its extra second of load time.
• Whether the cashier landing beats the homepage for the same GEO.
• Whether the creator you like is cheaper than the creator you think is too expensive.

It will not tell you why. A sub-id is a label, not a psychologist. If a landing loses, you still have to look at the landing. The label's job is to point you at the right one.

How AFFILIFY works is the path the click takes. The first thirty days is when to start tagging (day one). And if you are splitting sportsbook landings, prematch versus in-play is the test worth using slots 3 and 4 on.