We read an interesting case from our colleagues at Habr. In the article, PM Ira Belitsa and engineering manager Svyatoslav Sychev told how the Mindbox team dealt with overloaded technical debt using proper prioritization.

Context: The Mindbox team is working on a highly loaded mailing list product: more than 850 customers generate over 20,000 RPS. In this regard, dozens of tasks for adding new features and fixing bugs regularly "fall" on developers.

Problem: due to the high load and lack of resources, the team has prioritized work on reliability and maintaining the level of service (in Mindbox, this area is called Toil). Because of this, a queue of unrealized client tasks has accumulated. 

Solution: Reorganizing the backlog using a RICE-based framework. Each task was evaluated according to the following parameters:

• Impact — the impact of the problem on the team and/or the client (for example, loss of money);

• Frequency — frequency of problem recurrence;

• Blast radius — how many customers are affected, how massive is the problem;

• Estimate — an approximate estimate of the time frame.

And then we consider the priority according to the formula (in the attached picture). To keep it relevant, we also introduce two weekly syncs between leads.

Result: the number of completed client tasks has increased almost 2 times, and the volume has decreased to comfortable values for the first time in six months.