App startup time has a number on it: the Android vitals thresholds to budget against
Android vitals publishes explicit excessive-startup thresholds - 5, 2 and 1.5 seconds for cold, warm and hot starts to first frame. What TTID measures, and how to run a startup budget like a release gate.
Users judge an app in the first two seconds. Long before your onboarding copy or your feature set gets a vote, the cold start has already cast one. Android vitals publishes explicit "excessive" thresholds for startup time - the store telling you, in numbers, where it stops considering your app's startup healthy. Startup performance is no longer a nice-to-have buried in a tech-debt list; it is a measured behavior with a published bar.
The store's bar, in numbers
Android vitals considers these startup times excessive: a cold start of 5 seconds or longer, a warm start of 2 seconds or longer, and a hot start of 1.5 seconds or longer. Cold, warm and hot describe how much of your app the system had to reconstruct: a cold start begins with no process at all, a warm start reuses some state, and a hot start is a resume. The metric behind all three is time to initial display (TTID) - the moment the first frame of your UI is on screen, including process initialization. Play evaluates quality over a rolling 28-day window, so a bad release does not haunt you forever, but it also means improvements take weeks to wash through the numbers. For the metrics Play designates as core vitals, crossing a bad-behavior threshold can reduce your app's visibility on Play and put a warning on your store listing - treat every published vitals threshold, startup included, as a line with consequences behind it, not a suggestion.
What the thresholds tell you about users
Read the thresholds as Google telling you where patience ends. Five seconds is an eternity for a cold start, yet that is the line for "excessive" rather than "good" - a competent budget aims well under it. A practical internal budget many teams adopt: cold start under 2 seconds on a mid-range device, warm under 1, hot effectively instant. The store threshold is the floor you must not fall through, not the target to aim at.
Where cold-start time actually goes
Profile before you optimize, because the usual suspects are boring. Application-class initialization doing network or disk work on the main thread; SDKs that front-load their setup because nobody questioned the default; heavyweight splash layouts that inflate view hierarchies the user never sees; and lazy-loading that turned out to be eager. On the Apple side there is no published store threshold to anchor to, so teams carry the same budget across platforms and enforce it themselves - a launch-time regression test in CI is worth more than a quarterly performance panic.
A working routine
- Measure TTID on real mid-range hardware, not the flagship on your desk. Startup cost is dominated by the slowest devices your users actually own.
- Watch the 28-day vitals trend after every release. A regression caught in the window is a fix; caught after the window, it is a reputation.
- Move work out of the startup path. Defer SDK init until first use, make the first frame cheap, and load the rest asynchronously.
- Gate releases on a startup budget. A CI check that fails a build over your internal TTID budget turns a value into a habit.
The honest summary
Google has put numbers on the table: 5, 2 and 1.5 seconds for cold, warm and hot starts, measured to first frame over a 28-day window. Treat those as the floor, set a tighter internal budget, and enforce it the way you enforce crash rates. The first frame is your app's handshake with every user - make it a firm one.
Sources
Prepared with AI assistance and reviewed before publication. Thresholds and metric definitions are quoted from Android Developers pages checked on 19 November 2026; the internal-budget suggestions are editorial practice, not a store requirement.