Imagine this situation:
- A bug generate wrong metrics.
- That metrics were register along several years.
- The bug was there since the begining. Everybody have lived a lie.
- Fix the bug has a huge impact into the business area. And the cost of fix it attempt to the client's confidence. It's not only the past involve in this problem, the future too.
- This wrong metrics starting to make some 'noise' because with the pass of the years we starting to have more information, more context and the bug started to look more and more evident.

What would we do?
We could fix incrementally following the 'natural' tendence of that metrics. For instance, if the true value is 10 and our current value is 2 and the tendence in the last medition was increase from 1 to 2: we increase the value from 1 to 3. The next year: 3 to 5 and so on. If the movement was the opposite: OK, we going down but less strongest. Soon or late the values arrives to the correct value.
Fix the bug with another bug that had less impact... and every month, every year, in a incrementally pace fixs the problem. The show goes on and nobody care about the Truth.

We cannot correct the metric suddenly or not?

A lot of ethical and professional implicancies are involved. I want to center on our side of the problem. The data team had to fix something and they knows the 'true' and participate on the 'confindence strategy' to take the bussines problem. Since that time, they learnt that at the end nothing care but sales. And that sales are the money in our accounts and the bread on our table. That day they left the data-kids things, and became a data-business-men.