The effective intervention point is session start, not mid-session

What it is

Where an intervention sits matters as much as how strong it is. Before a session starts, the user is in a cold state: goals, costs, and alternatives are all in view, and a single confirmation can change whether the session begins at all. Once use is underway, a hot state takes over: reward is near, time perception blunts, and any prompt that pops up competes head-on with the immersion machinery. The effective intervention point is therefore session start — change the conditions before the behavior launches, rather than asking for it to stop mid-flight.

Why it happens

The cold-hot empathy gap: people in a cold state cannot accurately predict themselves in a hot one — calm, they overestimate how obedient they will be to a prompt; impulsive, they no longer recognize the decisions their calm self made. A mid-session prompt faces exactly the hot state: imminent reward weights the judgment, interruption costs become immediate, and the prompt's message almost inevitably loses to situational pull — users endorse the reminder's value in retrospect (cold) and tap "continue" in the moment (hot). The session-start point sits right on the state boundary: the user is still on the cold side, and a launch confirmation — an intent question, a few seconds' delay, a today's-goal screen — acts on a decision that has not yet launched, where the same wording carries far more weight. For equal prompt intensity, the interception rate at the start point is systematically higher than mid-content.

Where it stops holding

This holds for use that has a clear onset. Forms without one — ambient background presence, fragmented glance-and-check — offer no start to intercept, and launch interventions have nowhere to stand. Habitual opening (unlock-and-tap on reflex) skips past the usable cold-hot transition window; the intervention must move earlier, to the icon layer, not inside the app. In sessions needed for work, launch friction misfires on legitimate use and needs pairing with purpose detection. Mid-session intervention is not worthless — a prompt hung on a natural stopping point fares better than one mid-content — but for equal investment the start point returns more. This is a priority-of-placement claim, not an all-or-nothing one.

Applying it

Spend the intervention budget on the launch path first: long-press instead of tap on the icon (turning opening from reflex back into decision), a launch delay of a few seconds showing today's intent, session quota presented at the start rather than discovered afterward. If mid-session prompts are kept, hang them on natural stopping points (content gaps, end-state cards), not mid-content. Keep the bypass route available: the bypass risk of enforcement is a separate problem, and placement choices should not be made in the name of tightening enforcement. Verification: place the same prompt copy before launch versus mid-use, and compare session-abandonment rates and rated annoyance — the launch position should win on both, judged less annoying as well as more effective.

Related

• Same group: P3.09.1 Time spent is not harm; purpose says more than volume · P3.09.2 Retrospective statistics have no power over use that already happened · P3.09.4 Friction must exceed impulse yet stay below genuine need
• Adjacent: P3.03.2 Prompting users to leave is a viable design choice · P3.04.2 Restriction tools are easily bypassed
• Search terms: cold-hot empathy gap · session initiation · friction placement