Research studio
Choose a research question, inspect how the data would be collected, and decide what the result supports.
11 results
- Does upfront fee disclosure improve checkout?Randomised experiment
Assign 4,000 eligible app users evenly before they see the checkout. Keep assignment stable across sessions and analyse users in their original group.
- Who is missing from the merchant survey?Stratified validation
Draw a stratified sample of storefronts by ward and opening hours. Retain a contact-attempt log for every sampled merchant.
- Do personal reminders increase attendance?Blocked experiment
Randomly assign students within each timetable block. Choose attendance per eligible session as the outcome and keep the study to one full teaching cycle.
- Can refill invitations increase repeat orders?Randomised encouragement
Randomly invite eligible shoppers who are due for a refill. Compare all invited and uninvited customers, including people who ignore the invitation.
- Why do farmers express interest but fail to book?Qualitative interviews
Sample growers across landholding size, route distance and booking history. Include people who declined and people who booked but cancelled.
- Did the new shift improve finishing quality?Matched comparison
Match line-weeks on fabric family, order size and operator experience. Alternate shift patterns where staffing permits.
- Who loses access when meals move to preorder?Mixed methods
Observe both shifts at paired sites, and interview workers who did not order. Track meal availability, waste and workers who could not use the ordering channel.
- Do shorter itineraries reduce cancellations?Blocked field test
Assign routes within comparable departure time blocks. Log temperature and severe-weather cancellations separately from ordinary customer cancellations.
- Did easier registration improve warranty records?Process audit
Sample registrations from both form designs. Independently reconcile each record to an invoice and device serial number.
- Can peer coaching improve first-contact resolution?Stratified experiment
Assign teams within comparable queue and tenure blocks. Keep difficult queues in the analysis when staffing changes.
- Will a turnaround guarantee pay for itself?Counter-day experiment
Randomise the guarantee by counter-day. Include jobs rejected after inspection when recording conversion and analyse repair complexity by arm.
Starter lab: repeat demand in Indore
Can the Indore pilot distinguish repeat demand from initial stocking?
- Define eligibilityRecord when each store received stock and how long it could sell.
- Measure outcomesSeparate sales, repeat orders and missing observations.
- Review the evidenceDecide whether the measured reorder rate supports expanding the pilot.
Design the study
Sample 30 stocking retailers across three wards. The retailer is the unit of analysis. Record pack sales and reorders over six weeks. Ask each owner when they first sold through a case, whether they reordered and why they stopped.
Read the result
In a simulated tally, 12 of 30 enrolled stores reorder within six weeks (40%). Two closed during follow-up and remain in the denominator. Record store ID, ward, opening inventory and first reorder date. Track missing visits separately from observed zero reorders.
Decide what follows
The estimate describes the pilot stores. Store recruitment and promotions limit extrapolation, and the tally cannot identify a causal effect of distribution. Inspect reorder volume by ward before forecasting annual demand.
Apply the evidence to the Indore case