Modern trade trial, kirana repeat

Research summary

The analyst team triangulated three independent signals before the CMO readout: laddered occasion maps from calls, conjoint uptake on ₹10 vs ₹20 tiers, and photographic shelf audits from ethnography. None of the three alone would have survived finance scrutiny; convergence is what moved the decision from debate to signed pilot.

Challenge

Haldiram's saw strong trial numbers from a modern-trade blitz but flat repeat in neighbourhood kiranas — the channel that still drives 70% of category volume in their West markets.

What made this brief hard was not sample size — the brand already had sell-out dashboards and distributor heat maps. The gap was causal: modern-trade trial and kirana repeat live in different mental accounts for the same household. Until those occasions were separated in the interview guide, every aggregate repeat rate looked like a flavour problem. Finance’s MRP model and product’s reformulation roadmap were both optimising the wrong lever because they shared the same blurred cohort definition.

We treated the board’s festive deadline as a forcing function for decision clarity, not speed for its own sake. The engagement succeeded only if we could show which bet — value pack, reformulation, or kirana architecture — was still alive after disconfirming evidence. That meant documenting rejected paths with the same rigour as the recommendation, so the CMO could defend the pilot to the CFO without hand-waving.

Brand context

Haldiram's sits in the ₹18,000 crore Indian namkeen category with heritage in Gujarat and Maharashtra. Modern trade had become the hero metric for innovation teams, while kirana remained the profit engine for distributors.

Competitors had flooded ₹10–₹20 price ladders with colour-coded sachets. Leadership feared a race to the bottom without understanding which occasion they were losing.

Approach

Forward-Deployed Researchers mapped the decision tree: repeat vs. one-time trial vs. competitive switch by occasion. Bundl AI Caller ran adaptive interviews in Marathi, Hindi, and Gujarati with laddering on the last three shop trips.

Quota design was deliberately uncomfortable for the client: we overweighted modern-trade trialists who had not repurchased in kirana within 21 days, because that cell held the repeat mystery. Sales initially wanted more ‘loyalist’ interviews that would confirm the brand story; we kept loyalists as a contrast arm, not the centre of inference. Adaptive probing on the last three shop trips surfaced occasion language early — before conjoint tasks anchored respondents on pack sizes they did not actually buy weekly.

Pairing qual with conjoint was non-negotiable. Interviews alone would have produced rich stories finance could dismiss as anecdotal; conjoint alone would have missed the shelf-search failure mode shop-alongs later quantified. The FDR team ran member checking every third day so two internal flavour narratives died in week one instead of resurfacing in the exec deck as ‘consumer said spice’.

Field design

Quotas balanced modern-trade trialists, kirana loyalists, and competitive switchers by SEC and city tier. Shop-alongs captured shelf photography and time-to-find at 120 high-velocity outlets.

Shop-alongs were timed for late afternoon kirana peaks, when chai sachet missions dominate and visibility is worst. Photographers captured competitor colour bands at eye level versus the client’s packs — the 41% miss rate was not abstract; it was measured in seconds-to-find with the shopper narrating aloud. Distributor display compliance varied by city tier more than the client expected; Pune outperformed Mumbai on facings despite lower total volume, which told us the issue was architecture per outlet type, not regional taste.

Conjoint holdouts deliberately included a ₹10 tier the client had not yet manufactured, to test demand before capex. That choice created internal tension — product worried about signalling — but it was the only clean read on whether repeat intent existed at the occasion price point. Field teams logged every quota miss and rebalanced toward competitive switchers when modern-trade cells filled too fast.

Methods & pedagogy

Sentiment scoring was used as a triangulation input, not a replacement for interpretation. High frustration on shelf search co-occurring with low negative taste emotion was the pattern that unlocked the recommendation — if we had led with flavour lexicon, we would have over-built spice storylines in the codebook. Pedagogy here means making the method visible: leadership could see which instruments carried which claims.

Graded confidence on the ₹12Cr run-rate reflected pilot geography and distributor execution risk, not qual uncertainty on the mechanism. That distinction mattered for the board vote: everyone agreed on what to fix; finance debated how fast it would pay back. We separated those questions in the written recommendation so accountability stayed traceable.

Sentiment analysis

Synthesis

Triangulation pointed to architecture, not reformulation: trialists liked the product in large packs but repeated on a ₹10 occasion the brand had under-served in kirana.

Analyst readout · triangulation

By day 14, the forward-deployed team published a graded evidence table linking each board claim to instrument, sample, and confidence. Material claims required two converging sources; flavour-led explanations never cleared the bar.

“The kirana uncle is not rejecting you on taste. He never saw the pack you expect him to repeat on.”

Result

The brand launched a ₹10 kirana-forward SKU with a distinct colour band (not a “premium refresh”) and a display playbook for 2,400 high-volume outlets. No blanket MRP reduction.

“We stopped debating spice levels in the boardroom. The evidence was pack, price point, and whether the kirana uncle could see us.”