“For two consecutive quarters”: why a bad fortnight costs six months
Restaurant operators read this phrase as a waiting period and it is not one, it is a reset rule. Just Eat requires its four Local Legend criteria to be held “for two consecutive quarters”, and the word doing the work is consecutive. A restaurant that satisfies everything for five months and then loses a weekend does not lose a weekend of progress. It starts again. That turns a short outage into a six month commercial consequence, which is a different order of cost from the orders missed on the night.
Why does a consecutive window behave so differently from an average?
Averages absorb bad days. Consecutive windows do not.
Under an average, a quarter with one bad weekend and eleven good ones still looks fine. Under a consecutive requirement, the bad weekend is not diluted by anything. It is a break in a chain, and the chain has to be rebuilt from zero.
The practical translation is that variance matters more than the mean. A restaurant with steady, unremarkable availability beats one with excellent availability punctuated by occasional collapses, even where the second has fewer lost hours in total.
What does that do to the value of a single incident?
It changes it from an hourly loss into a multi-month one, but only sometimes, and you cannot tell which at the time.
An incident in the middle of a qualifying run is expensive out of all proportion to its length. The same incident a week after a reset costs almost nothing extra, because the count was already at zero.
Nobody inside the restaurant knows which situation they are in, because the count is held by the platform. Just Eat publishes the requirement and not the counting rule, so the reset described on this page follows from the word consecutive rather than from anything the platform has stated about what happens to a broken run. Either way the argument is the same: treat every avoidable outage as though it were the expensive kind, since you have no way of establishing that it is not.
How does this interact with the offline criterion?
Directly, and it is the pairing that catches operators.
The availability criterion is “Time offline at 10% or below”. The window rule then says it has to hold twice in a row. So the question is not whether you were under ten percent on average across the year, it is whether you were under ten percent in each of two adjacent quarters.
A restaurant with an annual figure of eight percent can easily contain one quarter at twelve, and that quarter breaks the chain regardless of how good the others were.
Can I see where I am in the count?
Not from a partner report, which is the structural gap in the whole arrangement.
You can be told the outcome, that you qualified or that you did not, but the running position is held on the platform’s side. So the restaurant is playing a game whose score it cannot see, on a metric it does not measure, over a window it cannot audit.
The part you can fix is the metric. Recording whether the listing was orderable during published hours, day after day, gives you your own version of the number, and it is the same behaviour the platform is scoring. With a Kitchain (kitchain.co) series behind you, a chain that has just been broken is something you learn in week two of the quarter and not in a letter six months later.
Does this pattern appear on other platforms?
The window rule is unusual in being published. The behaviour it describes is not.
Platforms across markets adjust ranking, promotional access and advertising eligibility according to availability, and several do it on rolling windows of their own. What is different about this one is that the period and the threshold are written down, which lets a restaurant reason about it instead of guessing.
Treat the published rule as a model for the unpublished ones. Where a platform will not say how it measures you, assuming a rolling window and a reset is closer to the truth than assuming a forgiving average.
What should change in how a group operates?
Move availability from a monthly review to a weekly one, and look at it per site rather than per estate.
Monthly is too slow for a rule that resets on a single bad period, and an estate average hides exactly the sites that break the chain. The two changes cost almost nothing and they are the difference between finding a problem inside the quarter and finding it in the letter that says you did not qualify.