Client studies · Templar

Food delivery and quick-service restaurants · India (Delhi NCR and Jaipur)

Two changes, one month

486 CONVERSATIONS 7 WORKING DAYS HINDI · ENGLISH CONSUMER RESEARCH

THE CLIENT

A quick-service restaurant chain with outlets across Delhi NCR and Jaipur.

THE DECISION

Whether to pay the switch-back cost and return to the old delivery partner, whose win-back offer expired at the end of the quarter, or reverse the roughly ₹30 menu price rise instead. The board wanted an answer before its next meeting, three weeks out.

THE SITUATION

Two changes went live in the same month: a menu price rise, about ₹30 on a typical order, and a new logistics partner for the chain's own-channel deliveries. Over the next two quarters, repeat orders fell steadily. Inside the company the argument split cleanly. Operations blamed the price rise, marketing blamed the new riders, and the old partner had tabled an expensive win-back offer with an expiry date.

WHAT WE DID

We designed the study around both hypotheses instead of testing either one directly. Boli held 486 recorded conversations in Hindi and English with two groups drawn from the chain's own consented customer list: 262 lapsed customers and 224 still-active lookalikes matched on city, order frequency and average spend. Nobody was asked about price or delivery until they had walked through their last three orders in their own words, so respondents could not simply agree with either camp. Because both groups lived through the same two changes, any difference between them had to come from experience rather than opinion. The study ran in seven working days; the extra day beyond our standard six went to matching the active cohort.

WHAT WE FOUND

68%

Asked why they stopped ordering, 68% of lapsed customers named price first. But 61% of the still-active group also called the new prices too high, and they were still ordering. The stated reason did not separate the two groups.

44%

The last order did. 44% of lapsed customers (95% CI 38 to 50, n=262) described a specific failed final delivery: food arriving cold, very late, or not at all. Among active lookalikes the figure was 9%.

n=84

The damage was concentrated. In Jaipur, 57% of lapsed customers (n=84) described a failed last delivery, against 38% in Delhi NCR, and most of those accounts pointed at weekend dinner orders. That pattern matched where the new partner ran its smallest rider fleet.

71%

Return was conditional on delivery, not price. 71% of lapsed customers said they would order again if delivery worked the way it used to. Only 24% conditioned their return on prices coming back down.

WHAT THEY DECIDED

The chain kept the price rise and declined the old partner's win-back offer. It took the findings into a renegotiation with the current partner instead: guaranteed rider capacity in Jaipur and a service floor at dinner peak, with an exit trigger if the floor was missed. We advised against reversing the price rise; it was the reason people gave, not the reason they left.

Everyone disliked the new prices, but only the people with a bad last delivery left.

Bring us a decision like this

Direct-to-consumer skincare · United Kingdom

Three stories, one cheque

512 CONVERSATIONS 7 WORKING DAYS ENGLISH · POLISH · PUNJABI CONCEPT AND CREATIVE TESTING

THE CLIENT

A direct-to-consumer skincare brand selling across the United Kingdom.

THE DECISION

Which of three repositioning routes would carry a media spend of roughly £800,000, with the creative agency briefing booked three weeks out.

THE SITUATION

The brand had grown for five years on simple formulations at honest prices, then growth flattened. Three repositioning routes were drafted: a clinical, lab-led story, an everyday-routine story built on simplicity, and a visible-results story. The founder favoured the clinical route and believed it would justify higher prices. The rebrand and the media plan were waiting on the choice.

WHAT WE DID

We fixed the decision and its date first: one route, chosen before the agency briefing. Because the founder had a favourite, the study was designed so respondents could not simply agree with it. Each person described the brand in their own words before any concept appeared, then heard all three routes unlabelled and in rotated order. We spoke with current buyers and with lapsed buyers as the comparison group, in English, Polish and Punjabi to match where the customer file lives. Every count in the report links to a recorded voice, and a researcher signed the recommendation.

WHAT WE FOUND

61%

Before hearing any concept, 61% of current buyers (n=320) described the brand in the language of the simplicity route: plain, honest, no fuss. They had not heard that route yet.

19%

Heard unlabelled and in rotated order, the clinical route was picked first by 19% of respondents (95% CI ±3.4 points, n=512). The simplicity route took 44%, the visible-results route 37%.

31%

Among lapsed buyers (n=192), 31% compared the clinical route, unprompted, to the prestige brands they had left. It read to them as a step towards the thing they had already walked away from.

£38

Asked to guess the price of a serum presented under each route, respondents put the clinical version at a median £38. The same serum sells at £19. Under the simplicity route, the median guess was £22.

WHAT THEY DECIDED

The client briefed its agency on the simplicity route and moved the full spend behind it. The clinical work was kept as substantiation on product pages rather than as the lead message. Templar advised against blending the three routes into one; nobody in 512 conversations had asked for a blend.

The people who pay £19 did not want it to feel like £38.

Bring us a decision like this

Packaged snacks · India (Uttar Pradesh and Bihar)

The ten-rupee one

520 CONVERSATIONS 12 WORKING DAYS HINDI · BHOJPURI · MAITHILI PRICING AND REVENUE ANALYTICS

THE CLIENT

A packaged snacks maker with kirana distribution across Uttar Pradesh and Bihar.

THE DECISION

Kill the ten-rupee pack, cut its grammage a third time, or raise the price to fifteen rupees. The next packaging film order was due in eight weeks, and the film carries the printed price.

THE SITUATION

Edible oil and packaging film costs had risen for three straight quarters, and the ten-rupee pack, 58% of unit volume, was selling below cost on some routes. Leadership leaned toward a third grammage cut, on the belief that the first two had gone unnoticed. The pack had built the brand's distribution. Now it was quietly eating the margin that distribution earned.

WHAT WE DID

Templar wrote the decision and its date down first, then designed around the client's hypothesis: no respondent was ever asked whether they would notice a smaller pack. Buyers described the pack from memory, in their own words, in Hindi, Bhojpuri and Maithili, over a link and a voice, no app. A comparison group of 208 lapsed buyers, matched on town size and outlet type, sat alongside the 312 current buyers, so the study could see who had already walked away and why. In parallel, the model measured what actually followed the two earlier grammage cuts in 104 weeks of outlet-level sales, then simulated all three moves on the table. Every claim in the report links to the recording behind it.

WHAT WE FOUND

61%

Of current ten-rupee buyers, 61% said unprompted that the pack holds less than it used to (n=312). The client was certain the two earlier grammage cuts had passed unnoticed. They had not.

44%

Of current buyers, 44% ask the shopkeeper for 'the ten-rupee one', often without naming a brand (n=312). The price point holds the shelf, and the kirana owner decides which pack lands in the hand.

29%

In two years of distributor sales, 29% of ten-rupee unit volume moved in transactions of two packs or more. A large share of the cheapest pack's buyers were not spending ten rupees at all.

-22%

The elasticity model put first-year volume loss at 22% (95% CI: 16% to 28%) if the pack moved to fifteen rupees. At a kirana counter, fifteen is not a price. It is an argument about change.

WHAT THEY DECIDED

The client kept the ten-rupee pack at its current grammage and moved the margin question to a new twenty-rupee pack aimed at the buyers already purchasing two at a time. Templar advised against the third grammage cut; the second one was still visible in repeat-purchase rates two years on. The film order went ahead unchanged.

Buyers were not loyal to the brand at ten rupees; they were loyal to ten rupees, and the shopkeeper decided the rest.

Bring us a decision like this

Consumer fintech · India

The ₹100 round trip

480 CONVERSATIONS 7 WORKING DAYS HINDI · ENGLISH · MARATHI · TAMIL CUSTOMER ANALYTICS

THE CLIENT

A consumer savings app with UPI-first onboarding, used across metro and tier-2 India.

THE DECISION

Whether to spend the next engineering quarter rebuilding KYC verification, with the roadmap due to lock in three weeks.

THE SITUATION

Registrations were growing. Second deposits were not: a large share of new users funded their account once and went quiet. The funnel showed a visible dip at the KYC step, and the team had settled on KYC as the culprit. A rebuild was already being scoped when Templar was brought in.

WHAT WE DID

The first session was about the decision, not the study: what the client would do differently if KYC turned out to be innocent. Boli then ran 480 conversations in Hindi, English, Marathi and Tamil, recruited by link from the client's consented customer list under the DPDP Act. The guide was built around the KYC hypothesis rather than toward it; respondents walked through their first two weeks in their own words, and KYC was never named unless they named it. Each consented conversation was matched to the client's transaction logs, so accounts of waiting could be checked against actual settlement times. A funnel chart had shown where people stopped; the conversations were there to say why.

WHAT WE FOUND

7%

Asked to retell onboarding in their own words, only 7% of drop-offs raised KYC as a difficulty (n=320, 95% CI 4% to 10%). Most described it as tedious but expected, the way people describe airport security.

58%

58% of drop-offs described their first deposit as a test, not as savings. The median first amount was ₹100. They were not starting a habit. They were checking whether money put in could come back out.

n=187

Of the 320 drop-offs, 187 had tried to withdraw their first deposit. 71% of them either could not find the withdrawal option or waited longer than they expected for the money to land, and read the delay as a warning.

2.4x

The matched transaction data told the same story. Users whose first withdrawal settled within a day were 2.4 times as likely to be in the second-deposit group as users who waited longer.

WHAT THEY DECIDED

The client shelved the KYC rebuild and pointed the quarter at the withdrawal path instead: a visible withdraw option on the home screen, the settlement time stated before the user confirms, and a notification the moment money lands in the bank. Templar advised against the KYC-lite tier already under discussion; the evidence would not carry its cost.

The first deposit was a question, and the app was slow to answer it.

Bring us a decision like this

Video streaming · Indonesia

The wrong week, not the wrong price

520 CONVERSATIONS 13 WORKING DAYS BAHASA INDONESIA · JAVANESE CHURN AND RETENTION MODELLING

THE CLIENT

A subscription video streaming service operating across Indonesia, several million subscribers, strongest in local serials.

THE DECISION

Whether to roll the monthly price back to IDR 49,000 across the entire base, with a board meeting set four weeks out and the next billing cycle running before it.

THE SITUATION

Last March the service raised its monthly plan from IDR 49,000 to IDR 59,000. Within two billing cycles monthly churn climbed from 3.9% to 6.8%, and exit surveys pointed at price, which was also the answer the survey made easiest to give. Leadership drafted a full rollback and wanted a verdict before it went to the board.

WHAT WE DID

The decision came first: keep the new price or return it, settled before the next billing cycle. Templar ran survival modelling on fourteen months of anonymised subscriber records to establish who left and exactly when, then put Boli in front of 304 recent leavers and 216 matched stayers, recruited from the client's own list where consent allowed. The conversations were designed around the price hypothesis rather than toward it: leavers reconstructed their final fortnight as subscribers, what they watched, how they paid, what they saw at renewal, before money was raised at all. Javanese ran alongside Bahasa Indonesia because a large share of the manual, e-wallet renewal base lives in Central and East Java. A human researcher wrote and signed the recommendation; every claim links to the conversation it came from.

WHAT WE FOUND

68%

Sixty-eight percent of leavers called the new price fair for what the service offered (n=304, plus or minus 5 points at 95% confidence). Price was the box most had ticked on the exit survey. In conversation, the number itself drew almost no heat.

3.4x

After the rise, subscribers renewing manually through e-wallets or mobile credit churned at 3.4 times the rate of auto-renewing subscribers. Before it, the gap was 1.4x. The new price turned a passive charge into an active decision, made monthly, at a screen.

57%

The model placed 57% of the spike's extra cancellations inside nine days of the finale of the platform's flagship local serial. The higher price reached renewal screens in the one week many subscribers had just run out of reasons to stay.

n=104

Among stayers who renewed manually at the new price (n=104), the reason named most often was an unfinished series, not the amount. Where something was still running, people paid.

WHAT THEY DECIDED

The client kept the price at IDR 59,000. It decided to time future price changes and manual renewal prompts away from the quiet weeks of the content calendar, and to rebuild the renewal screen so next month's slate appears beside the amount. Templar advised against the full rollback and against a blanket win-back discount, since most leavers had never objected to the number.

Most subscribers could live with the new price; they could not be asked for it in the week the reason to stay had just ended.

Bring us a decision like this

Grocery retail · India (tier-2 cities)

Whose name goes on the atta

480 CONVERSATIONS 12 WORKING DAYS HINDI · MARATHI RETAIL AND CATEGORY INTELLIGENCE

THE CLIENT

A regional grocery chain with 40-plus stores across tier-2 cities in Madhya Pradesh and Maharashtra.

THE DECISION

Whether to put the store's name on atta, dal and edible oil. Milling and packaging contracts had to be signed before the October shelf reset, which left one board meeting to decide.

THE SITUATION

A national discounter had opened in two of the chain's cities and staple margins were thinning. The board had debated a private-label line for two years and stalled each time on the same fear, that shoppers who buy branded atta would not accept a store's name on food. Marketing had budgeted a launch discount of eight to ten rupees a kilo to buy trial.

WHAT WE DID

The decision and its date came first; everything else was designed backwards from the October reset. Auditors measured share of shelf in 22 stores while fourteen months of loyalty baskets were modelled for switching behaviour in staples, which is why this ran twelve working days rather than the usual six. Boli then held 480 conversations in Hindi and Marathi across four cities: 320 with the chain's own shoppers, recruited from its loyalty list under consent, and 160 with similar shoppers who buy their staples elsewhere. The conversations were designed around the client's hypothesis, not toward it. People first described how they judge the quality of atta, dal and oil and whose judgement they borrow; the store's own brand was raised only once, at the end of each conversation, so nothing earlier could lean toward it.

WHAT WE FOUND

43%

Mid-tier national brands held 43 per cent of staple shelf space across the 22 audited stores. The same brands accounted for 19 per cent of staple revenue in fourteen months of loyalty baskets.

68%

Asked who they trust to judge the quality of atta, 68 per cent of the chain's shoppers named the store itself, unprompted (n=320). The most-stocked mid-tier brand was named by 9 per cent.

57%

Asked at the end of the conversation, 57 per cent of chain shoppers (n=320; 95% CI: 52 to 62) said a store-name atta would go into their next basket at the same price as their current brand. The eight to ten rupees a kilo budgeted to buy trial would have paid for something shoppers were already offering.

n=160

In the comparison group of similar shoppers who buy staples elsewhere, 54 per cent already buy loose atta or dal on a shopkeeper's recommendation. Trusting a retailer's judgement of staples is the incumbent habit in these cities, not the innovation.

WHAT THEY DECIDED

The board approved store-name atta, dal and oil for the October reset, priced at parity with the mid-tier brands and taking over their facings rather than new shelf space. Templar advised against the launch discount marketing had budgeted; in the conversations, a cheaper staple raised doubts about quality instead of interest in trying it.

The chain spent two years asking whether shoppers would trust its name while renting 43 per cent of its staple shelf to brands they trusted less.

Bring us a decision like this

Consumer durables · India

Betting the plant on a heatwave

128 CONVERSATIONS 15 WORKING DAYS HINDI · MARATHI · GUJARATI FORECASTING AND PREDICTIVE MODELLING

THE CLIENT

An air cooler manufacturer selling through dealer networks across north, west and central India.

THE DECISION

How many units to commit the factory to for next summer. Vendor contracts and line bookings locked in mid October; after that date, capacity could not be raised until the season was half over.

THE SITUATION

Last summer's heatwave was the longest in years, and sell-through jumped 31%. The planning team read the jump as a new baseline and drafted capacity for 7.6 lakh units. The board wanted one independent read before the vendor contracts were signed.

WHAT WE DID

We started from the October date and worked backwards. The model separated weather from demand: five summers of district-level sell-through matched to daily temperature records, with the districts the heatwave skipped acting as a natural control. Boli ran 128 dealer conversations in Hindi, Marathi and Gujarati, asking about carryover stock, order timing and what shoppers cross-priced. The interview was designed around the client's hypothesis, so no dealer was ever asked whether demand had stepped up. The output was three weather scenarios with a stated confidence band, not one number, delivered in fifteen working days on the longer track we use for modelling-heavy work.

WHAT WE FOUND

8%

Headline sell-through grew 31% last summer. In the fourteen districts the heatwave largely missed, it grew 8%. Most of the jump the planning team had treated as a new baseline was weather.

58%

74 of 128 dealers were still holding unsold units from last season and said they would order later and lighter this year. The trade was planning for a normal summer even while the factory was not.

5.2-7.3 lakh

Across the three weather scenarios, the 90% band for season demand ran 5.2 to 7.3 lakh units, with a central case near 6.1 lakh. The draft plan of 7.6 lakh sat above the hottest scenario.

n=61

Among dealers in tier-2 towns (n=61), 44% said shoppers who priced a premium cooler last season came back asking about entry-level ACs. The ceiling on the premium range was not a rival cooler.

WHAT THEY DECIDED

The client committed capacity at the central scenario, near 6.1 lakh units, with a lighter premium mix, and negotiated component options that could raise output if December bookings tracked the hot scenario. Templar advised against building to the internal 7.6 lakh figure, which sat above the band under every weather case. The gap between the two plans, roughly ₹80 crore in component and line spend, stayed conditional instead of committed.

A confident forecast is often just a weather bet with a decimal point.

Bring us a decision like this

Telecom · Kenya

Paying to be remembered

520 CONVERSATIONS 12 WORKING DAYS SWAHILI · ENGLISH BRAND AND MARKETING MEASUREMENT

THE CLIENT

A prepaid mobile operator serving customers across Kenya.

THE DECISION

Approve, cut, or reallocate the media budget for the coming financial year. The marketing team had asked for an increase, finance had proposed a cut of nearly a third, and the planning cycle closed in seven weeks.

THE SITUATION

Advertising spend had risen for three consecutive quarters while prepaid top-up revenue stayed flat. The marketing team read this as a creative problem: the campaigns were tired, and a bigger, fresher push would move the number. Finance read the same flat line as proof the budget was waste. Both sides were arguing from the same chart.

WHAT WE DID

We started with the decision and its date: one budget line, due when the planning cycle closed. The model and the conversations were built to answer different halves of the same question: what the spend was buying, and why. The conversation guide was designed around the marketing team's hypothesis rather than toward it: nobody was asked about the campaigns until they had explained, in their own words, where their last five top-ups went and what decided each one. Conversations ran in Swahili and English, matched to where respondents lived. Twelve working days rather than the usual six, because the model needed 30 months of clean weekly data before the fieldwork could be read against it.

WHAT WE FOUND

71%

Of the modelled advertising effect on top-up revenue, 71% came from existing subscribers shifting the timing of top-ups they would have made anyway. New SIM activations showed no detectable response to media weight in any of the three quarters.

92%

In the comparison group of 200 dual-SIM users who spent mostly with a competitor, spontaneous awareness of the client's brand stood at 92%. The campaigns were reminding people who had never forgotten.

n=200

In the dual-SIM comparison group, 63% said their second SIM took the larger share of their data money because of night bundle pricing at the KES 20 and KES 50 level (95% confidence interval, plus or minus 7 points). What moved them was denomination pricing, not anything an advertisement said.

2.4x

Recall of the latest campaign was 2.4 times higher among the client's own heavy top-up users (n=134 of the 320 subscriber conversations) than among the dual-SIM users the spend was meant to win. The media reached the people who needed no persuading.

WHAT THEY DECIDED

The client held total media spend flat and reallocated roughly a third of it, away from mass reminder campaigns and into acquisition media in four counties where the model showed headroom, plus clearer offer communication at the KES 20 and KES 50 top-up levels. Templar advised against the requested increase, and also against the deep cut, because the model showed the existing spend still defended top-up timing during competitor bursts. The budget question was settled before the planning cycle closed.

The advertising had been paying people to remember a brand nobody had forgotten.

Bring us a decision like this

EdTech · India

The customer who never pays

560 CONVERSATIONS 11 WORKING DAYS HINDI · ENGLISH · MARATHI · TELUGU CUSTOMER ANALYTICS

THE CLIENT

An online tutoring provider for school-age students across India.

THE DECISION

Whether to launch a 20% renewal discount on the ₹36,000 annual plan. Pricing had to be locked within six weeks, before the renewal window opened for the new school year.

THE SITUATION

Acquisition costs meant a household only turned profitable in its second year, so first-year renewals were the business model. They had fallen for three straight quarters. Exit surveys put price at the top of every list, and leadership had a renewal discount drafted, costed and ready. The board wanted it confirmed or killed before the new school year.

WHAT WE DID

Templar interviewed each household twice: the parent and the student, separately, through Boli, and analysed every household as a pair. The sample came from the client's own first-year customer list: lapsed and at-risk households, with similar renewed households as the comparison group, and the client's session logs joined to each pair where consent allowed. Because the client's hypothesis was price, the conversations were designed around it: fees were never raised unless the respondent raised them first. Conversations ran in Hindi, English, Marathi and Telugu, matched to the client's enrolment geography. Driver modelling on the paired transcripts took the engagement to eleven working days rather than the usual six.

WHAT WE FOUND

64%

64% of parents in lapsed households (n=125) named the fee as their reason for leaving. In those same households, the child's session attendance had fallen by roughly half in the two months before cancellation. The fee was the explanation, not the cause.

2.6x

In the driver model, the student's answer to one question, would you keep the classes if it were your choice, predicted the household's renewal intent 2.6 times more strongly than the parent's overall satisfaction score.

31%

Only 31% of parents (95% CI: 26 to 37, n=280) could name one specific thing their child had learned in the previous month. Parents were paying for progress they could not see, and a discount would not make it visible.

n=131

Students in Classes 6 to 8 were the least attached. 57% described the sessions as a second school day, and their most common request was the same tutor every week, something the client's scheduling system did not promise.

WHAT THEY DECIDED

The client shelved the discount and put the budget into a tutor continuity guarantee, one tutor per student for the full school year, plus a short monthly note from that tutor naming one thing the child had learned. Renewal outreach moved forward to month four, where the attendance drift began. Templar's written recommendation advised against any price move: the conversations located no price-sensitive segment large enough to pay for one.

By the time the parent said no, the child had been saying it for months, one skipped session at a time.

Bring us a decision like this

Nine decisions,nine answers.

Two changes, one month

Three stories, one cheque

The ten-rupee one

The ₹100 round trip

The wrong week, not the wrong price

Whose name goes on the atta

Betting the plant on a heatwave

Paying to be remembered

The customer who never pays

Nine decisions,
nine answers.