Fine and coarse conversion values

What is the difference between fine and coarse?

The fine-grained value is the more detailed of the two. The coarse value gives a rough level. Apple defines them by where they appear.

So a single install can report in detail once, early, and then only in rough levels. The timing of each postback is laid out in SKAdNetwork postbacks and their three windows.

Why would the first postback carry a coarse value?

Because of the postback data tier. In lower tiers the system sends the coarse value instead of the fine one, even in the first postback.^[1]

Apple assigns the tier and relates it to crowd size.^[1] An install from a small crowd can report a coarse value where one from a larger crowd reports a fine value. Apple's documentation page gives no thresholds. The reasoning behind the tiers is on the crowd anonymity page, so the rest of this guide deals with what you control: the mapping.

How do you decide what the values mean?

Treat the mapping as a short list of questions you want each postback to answer. A workable order for a small team:

1. Pick one early signal of a good user. Choose an in-app event that happens in the first window, days 0 to 2, for a meaningful share of users.^[1] A trial start or a first purchase are examples. An event that usually happens in week three cannot inform the first postback.
2. Map the fine value to steps toward that signal. For example: opened the app, finished onboarding, started a trial, paid. Higher value, further along.
3. Map the coarse value to a simple question. With only a rough level to work with, use it for something like no key event, some activity, or paid.
4. Use the later windows for retention or revenue. The second and third postbacks cover days 3 to 7 and 8 to 35, and they carry coarse values.^[1] They suit questions such as "is this user still active" or "did they pay again".
5. Write the mapping down and keep it stable. If you change what a value means, postbacks from before and after the change cannot be compared.

Common mistakes

• Encoding too much. A mapping with many rare combinations spreads installs so thinly that few values ever appear in volume.
• Relying on the fine value alone. Lower tiers and later postbacks only give you the coarse one, so a mapping with no coarse plan leaves those postbacks empty of meaning.
• Mapping revenue you cannot see in time. If revenue arrives after the window closes, that window's value will not reflect it.
• Forgetting the ad channel. The postback goes to the ad network, so the network reads your values too. Check what each of your channels expects before you finalise the mapping.

Where to set it up

You can configure the mapping in your attribution tool and let the SDK update the value. Ask whether the tool supports both kinds of value and all three postbacks on your plan. Airbridge, which runs this site, lists a conversion value manager for SKAN 3.0 and 4.0 covering coarse, fine-grained and multiple postbacks.^[2] The details are on the Airbridge page.