Inventory the live states
Screenshot every paywall, plan sheet, and restore-purchase path currently shipping. Name them. If two screens share a name in your schema, rename one.
Upgrade Signals
Intent leaves traces. Most schemas throw them away.
Upgrade Signals is the studio’s working name for the markers we teach teams to keep: not “clicked upgrade,” but the quieter evidence that a person is considering a higher plan.
What counts as a signal
A signal is any repeatable behaviour that changes your belief about upgrade intent. Comparing two plan cards for more than a few seconds can be a signal. Opening billing history after a feature gate can be a signal. Hitting a storage limit three times in a week can be a signal.
A vanity metric is a count that moves when intent does not. Paywall impressions often belong here. So do “upgrade CTA visible” events fired on every home screen render. If the number can inflate while nobody is considering a new plan, it is not an Upgrade Signal.
How we classify them
In the lab we sort markers into three shelves: capacity (the product is pushing against a limit), comparison (the person is studying plans), and commitment (they have started a billing or family-share flow). Mixing the shelves into one “intent score” is how teams fool themselves.
Thailand adds a fourth shelf we keep separate: instrument friction — PromptPay vs card, storefront currency mismatch, VAT surprise. Those events look like drop-off. They are often not a no.
A short field sequence
Attach one honest window
Pick the billing cycle that actually governs the product. For many Thai subscription apps that is monthly plus a festival shoulder. Do not copy a 7-day SaaS window.
Write the disconfirming test
Before you celebrate a lift, state what evidence would prove the upgrade was accidental, refunded, or family-plan noise. Keep that sentence next to the chart.
Where this method is taught
Upgrade Signals is not a separate product. It is the spine of the Plan Conversion Analytics Lab and the shorter Paywall State Taxonomy workshop. Desk Reviews apply a compressed version to a single surface.
If you only want the booklet without a seat, we do not sell it that way. The lexicon is easy to misuse without the limitation practice that sits around it.
See the coursesQuestions about applying the method to a live app: write to the studio.