The Loyalty Card Memo: A Causal Claim
Memo from the marketing head of a supermarket chain: "In the quarter after we launched the Silver Card loyalty programme in our four Pune stores, sales of packaged snacks in those stores rose 18 per cent over the same quarter last year, while sales in our Nagpur stores, which did not run the programme, rose only 4 per cent. The card clearly caused the increase. I recommend extending it to all 40 stores next quarter at an estimated cost of ₹60 lakh, which the Pune results show will be recovered within one quarter." Facts on file: a large competing supermarket in Pune closed two months into the quarter; the quarter included Diwali, when snack sales typically rise across the country; and the Pune stores simultaneously ran a "price-drop" promotion on 300 snack lines. The chief executive has asked for a judgement on the proposal before signing the cheque.
1. State the causal claim in the memo and the assumption it rests on, and explain why the Nagpur comparison does not by itself establish causation.
Read the worked answer
Claim: the Silver Card caused an 18 per cent rise in snack sales in Pune. Assumption: apart from the card, Pune and Nagpur were comparable, and no other factor affected Pune snack sales in that quarter.
That assumption fails on the facts on file. Two of the listed factors are specific to Pune: a large competitor closed two months into the quarter, pushing its customers to the Pune stores, and the Pune stores ran a price-drop promotion on 300 snack lines. Either could have raised snack sales without any help from the card.
Diwali is a weaker explanation for the gap, because it falls in both cities; it may explain part of Nagpur's 4 per cent rise, but it cannot explain why Pune rose 18 per cent while Nagpur rose 4. This distinction matters: a factor that affects both groups cannot explain a difference between them, while a factor that affects only one group can.
The memo treats a comparison of two stores groups as a controlled experiment. It is not - the groups differ in location, promotion activity and competitive conditions, and the card is only one of several differences.
2. Rank the three listed facts by how seriously each weakens the recommendation, and justify the ranking.
Read the worked answer
Strongest: the competitor closure in Pune. It shifts an entire catchment area to the Pune stores and affects every category, not just snacks. It is entirely Pune-specific, and nothing in the memo controls for it. The card's estimated effect is likely to be inflated by this alone.
Second: the price-drop promotion on 300 snack lines. Also Pune-specific, and directly targeted at the very category measured, which makes it a close competitor to the card as an explanation. It is placed second only because it is a deliberate action of the chain and could, in principle, be costed and separated.
Weakest: Diwali. It raises snack sales everywhere, so it explains part of the 4 per cent in Nagpur and part of the 18 per cent in Pune, but it does not explain the gap between them. It weakens the absolute figures, not the comparison.
The ranking matters because it tells the chief executive where to look: the Pune-specific events, not the festival, are what threaten the causal claim.
3. What evidence would you require before approving the ₹60 lakh, and what check could be run before the full rollout?
Read the worked answer
First, separate the card from the other Pune events. Ask for category-level data across the whole store, not just snacks. If total Pune store sales also rose close to 18 per cent in the quarter, the gain came from the competitor's exit, not from the card, because a loyalty card tied to snack purchases would not lift every category equally.
Second, ask for a comparison that holds the confounders constant: Nagpur stores is one control, but the memo needs stores matched on size, catchment and competitive conditions, and it needs the pre-card trend. If Pune snacks were already growing at 10 per cent a year, the card's real contribution is far smaller than 18 points.
For the rollout, run a phased pilot: launch in roughly eight stores and hold back eight matched stores for one quarter, with the comparison fixed in advance. Cost the exercise properly as well - ₹60 lakh across 40 stores is ₹1.5 lakh per store, so the chain needs snack sales per store before it can claim a one-quarter payback. Approve the pilot, not the full rollout.