Retention is how much of a delivery is still there after time passes: of 1,000 followers added this month, how many are still following next month. It's the quality metric that separates services which look identical at checkout.
Drops have exactly two engines. Platforms continuously remove accounts they classify as fake, which takes out the bottom grade of delivered stock in waves. And real people unfollow, which trims any audience, bought or organic.
The first engine is what retention measures against. The same 1,000 followers can mean very different things a month later, depending entirely on where they came from.
Cheap bot stock is built fast and detected fast: platforms sweep on the signatures of mass production, and bottom-tier deliveries produce the classic sawtooth: spike, then slide.
Higher-grade stock has nothing for the filter to catch: complete profiles, history, believable behaviour. It costs more per unit and survives, which usually makes it cheaper per KEPT unit. That's the only price comparison that matters.
One naming warning: video analytics use the same word. Audience retention is the share of a video people keep watching. Delivery retention is followers surviving. Same word, unrelated metrics; context tells you which is meant.
It ranges from terrible to near-total depending on stock grade. Bottom-price bot stock can lose most of itself; high-grade stock loses a small share. The service's grade and guarantee are your forecast.
Note the count when delivery completes, and check it at the same time weekly. Public chart tools do this automatically for any account.
Not honestly: platforms clean continuously and real accounts churn. Near-total retention over a window is the realistic top end, which is why guarantees exist for the gap.
Every definition lives in the full glossary, and the services behind the vocabulary are on the services page.