Description
Estimates stage-to-stage promotion, repetition and drop-out rates from two aggregate snapshots, then simulates a notional cohort to show completion, time taken and where people leave.
Education planners needed to know how pupils move through grades when individual records were unreliable. Three approaches were compared: the true cohort method (follow real individuals: accurate but slow and record-dependent), the apparent cohort method (compare a stage's count with the next stage a period later, treating all loss as drop-out and ignoring repetition) and the reconstructed cohort method. The reconstructed method uses counts by stage for two consecutive periods plus the number repeating each stage in the second period to derive promotion, repetition and drop-out rates for every stage. A notional cohort of 1,000 is then simulated through the cycle under explicit assumptions, giving completers, leavers by stage, average time to complete and a coefficient of efficiency (ideal time divided by time actually spent). For products, the stages become levels, course modules, certification steps or plan tiers where users can progress, repeat or leave.
Attributes
- Discipline
- Product Analytics & Strategy
- Loss type
- Unmigrated
- Stage
- Post-launch, Strategy
- Data
- Quantitative
- Effort
- Days
Steps
- Assemble counts of users at each ordered stage at time t and t+1, plus the number at each stage in t+1 who were already there in t (repeaters).
- For each stage compute the repetition rate (repeaters in t+1 over count in t), the promotion rate (count in the next stage in t+1 minus its repeaters, over count in t) and the drop-out rate as the remainder.
- State the assumptions: no new entrants except the cohort, a fixed maximum number of repeats, constant rates across the cycle, and the same rates for repeaters and first-timers.
- Simulate a notional cohort of 1,000 through the stages to obtain completers, leavers by stage, average time to completion and the coefficient of efficiency.
- Check the data for over-counted stages, new entrants confused with repeaters and missing periods, each of which biases the rates in a known direction.
- Where individual logs exist, validate the reconstructed rates against a true cohort sample.
Sources
- Quantitative Social Research Methods (Singh, 2007)
Also known as
- UNESCO reconstructed cohort model
- Student flow analysis
- Apparent cohort method
- Grade-flow analysis
- Coefficient of efficiency
- Reconstructed cohort flow analysis (promotion, repetition and drop-out rates)
Categories
- Analytics & Behaviour
- Product Analytics & Strategy