In the Locale Scorecard Lab we ask cohorts to clock one thing first: the interval between an agreed string freeze and the moment a locale is marked accepted by whoever actually has that authority — not the vendor’s “delivered” stamp, and not the TMS job closing itself at midnight.
That interval, time-to-locale, usually moves before volume metrics admit a problem. A language can sit at 12% remaining and still be healthy if last week’s comparable freeze cleared in 36 hours. Another can sit at 4% remaining and be doomed because review has stalled and nobody named an owner.
The measurement is fussy on purpose. Weekends need a rule. Vendor holidays need a rule. In-house linguistic review is often the hidden queue; if you only clock the supplier, you will praise a fast agency and miss a two-day wait on your own desk.
We treat a sudden lengthening of time-to-locale as a leading signal for store-date risk. It is not a law of physics. A one-off legal review can inflate the clock without predicting next sprint. That is what the exception log is for: name the inflation, date it, and refuse to let it become the new baseline by silence.
If your export cannot produce a freeze timestamp, you do not yet have Localization Performance Analytics. You have a status colour. The lab’s first fortnight is often spent getting that timestamp to exist. It is unglamorous and it is the whole job.