32% Retention but Stuck at ₹30 Cr? How to Use AI to Hit a ₹500 Cr Goal | Ft. The Little Farm Co.

ELI5/TLDR

Two siblings run a pickle and chutney brand called The Little Farm Co. They sell about 70,000 jars a month, mostly through Blinkit/Zepto/Instamart, and are knocking on a ₹30 Cr run rate. Their dream is ₹500 Cr. They asked Arjun Vaidya and Shantanu Deshpande three questions: when do we go to physical retail, how do we stop screwing up inventory, and should we go wide into more categories or deeper into pickles. The advice was uncharacteristically restrained: don’t go retail yet, fix your stockouts on quick commerce with AI demand planning, and use a 70-20-10 split between your core, your experiments, and your wild bets. The recurring theme — most “good deals” in offline retail are actually bad deals, and your real 10x is hiding inside the channel you’re already winning at.

The Full Story

The setup

The Little Farm Co. was started by Aditi after her father passed away and she found herself reviving family pickle recipes from her bua’s kitchen and her dad’s old notebook. She studied digital marketing at Cass Business School, did D2C consulting on the side, and turned the pickle thing into a website around the start of COVID. Her younger brother Aditya joined later as CEO after stints at EY and Jimmy’s Cocktails, where he learned modern trade and watched the e-commerce-to-quick-commerce transition up close. Today: 32% repeat rate, EBITDA positive, cash flow positive, ~₹30 Cr ARR, ~70,000 jars a month. Channel mix is 70% quick commerce, 15% marketplaces, 10% own Shopify, 5% HoReCa (a few five-star hotels in NCR carry their pickle bar). Stated goal: ₹500 Cr in nine years, with milestones at ₹24, ₹40, and ₹100 Cr along the way.

The retail question, and why the answer is “not yet”

Aditi asks: when do we go offline? Vaidya flips it back: why go offline at all right now? He walks them through a framework he likes — every FMCG channel can be scored on three things, awareness, affordability, and accessibility. Pre-internet, awareness meant TV reach percentages, affordability meant getting the right rupee coinage, and accessibility meant distribution. The internet upended all three. Awareness now lives on Meta, Google, and increasingly answer-engine optimization (AEO/AIO). Affordability stopped meaning price points because incomes rose and the audience that discovers you on Meta can already afford ₹235. Accessibility became a fight over dark store SKU slots, not corner-shop shelves.

He has them score themselves channel by channel. D2C: tick, tick, tick. Quick commerce: dotted-tick on awareness, full ticks on affordability and accessibility. Marketplaces: weaker, opportunistic. Retail: zero, zero, zero. The point isn’t to embarrass them — it’s to ask whether their current proposition can win in retail at all. A ₹235 pickle in a thin plastic jar that may not survive 30 days of sun in a kirana store is a pickle designed for dark stores, not for shelves. Going retail without redesigning the product, the packaging, and the price-point arithmetic is what Vaidya calls “a very expensive 0-to-1 journey when you’ve already started 1-to-10.”

The math he sketches is the real punch. India has roughly 35 crore households, of which urban is maybe 15 crore. Their realistic affordability segment is 2-5 crore households. If they get to 1 crore households (half the conservative target), each buying 2 jars a year at ~₹170 net realisation, that’s ₹350 Cr — just from quick commerce, where they already have a structural advantage. Today they’re at ₹30 Cr.

So you have a 10x potential today without even touching retail.

The frame he keeps returning to: deploy capital where you have an advantage, where others (Meta, the dark store networks, the legacy brands’ lack of digital instincts) are doing the heavy lifting for you. Going where you’re weak just because the market is “big” is the same mistake MNCs made in the 1990s — multiply 1 billion people by USA per-capita and watch the spreadsheet make you broke.

Why “good deals” in offline retail aren’t good deals

Aditya keeps closing offline distribution deals he thinks are great. Vaidya defines what “good” actually means in retail:

A good deal is they buy the inventory with no return and the cash. You’re not responsible for working capital, you have no liability. That’s a good deal. But even that is not such a good deal — because if it doesn’t sell, you blow your reputation. They’ll never buy again, and they’ll tell people.

In other words: most deals founders take in MT/GT are inventory pushed onto the brand’s books, sitting in a warehouse the brand still owns, in a channel that may never produce sell-through. The reputational tail risk is asymmetric. Quick commerce, by contrast, is the new kirana — fewer brands per category (Blinkit shows you five mango pickles, Amazon shows you 35), so once you’re in the top five, you don’t get displaced. Quick commerce momentum is moving back toward brands with strong market share, the way it did during the Amazon/Flipkart era for D2C startups.

The stockout problem is actually a supply chain problem

Aditya admits he’s still doing inventory planning on a spreadsheet that hasn’t evolved since the company was doing ₹5 lakh a month. They keep falling short on some SKUs and overstocking on others. Vaidya’s diagnosis is sharp: quick commerce isn’t a branding battle, it’s a supply chain battle. The best supply chain operators in India right now are the quick commerce companies themselves — go ask them which brands at your scale fulfill at 90-95% and learn from those brands directly.

If you just remain in stock, you will win the algo by itself. Out of stock damages your algorithm and it takes time to build back.

Then the actionable bit, which is the part Aditya gets visibly excited about. Stop running supply on a spreadsheet. Feed your daily MIS, your demand plan, and your sell-through into a custom GPT that you tell to act like a supply chain expert. Ask it which SKU on Instamart in which city is going to stock out next week. The math itself isn’t hard — it’s the cognitive load of running it daily across SKUs and platforms that breaks founders. Vaidya offers to share the GPT his team trained for this.

The framing he leaves them with:

You’d rather spend your time with customers in the market or with a category manager at Blinkit, not with the purchase order inventory.

The Be Big, Be Fast, Be Bold framework

This is the spine of the conversation and the bit most worth memorizing. Vaidya’s universal capital allocation framework, which he claims works at every scale he’s used it:

Be Big — your core business. The thing you understand. The thing that pays the bills. For Little Farm, this is pickles. The cycle here is plan → execute → review → adjust, and the game is to run that cycle faster and faster. In retail it takes months, in marketplaces it takes weeks, in quick commerce it takes hours. Allocate ~70% of capital, attention, and energy here. This will end up generating ~85-90% of the business.
Be Fast — adjacent experiments where you don’t know what you don’t know. For Little Farm this is chutneys. The game here is not to execute, it’s to A/B test. Don’t blow money executing — keep cycling through versions until something sticks. Allocate ~20%. Will generate maybe 7-8% of the business.
Be Bold — out-of-domain bets. For Little Farm this is ketchup, a foreign condiment with no obvious adjacency to home-style Indian pickle. Allocate ~10% of time. Will generate ~2% of the business. The principle here: make a little, sell a little, learn a lot. Don’t get inventory and cash stuck. The point of Be Bold isn’t the revenue — it’s that occasionally a Be Bold winner graduates to Be Fast, and a Be Fast winner graduates to Be Big.

The reason this matters: founders are zero-to-one DNA. They get bored of the boring core that funds everything and want to spend 70% of their time on chili oil. The framework gives them permission to be excited about chili oil while disciplining the capital allocation to stay sane. Or as Deshpande puts it — founder focus and founder energy are two different things, and you have to optimize both.

Right to win, decided by the consumer

The other discipline he lays on them — for every potential extension, ask whether you have a “right to win,” and decide it from the consumer’s seat, not yours. Bombay Shaving Company had a right to win in trimmers (same problem space), a partial right to win in face washes (same shelf, similar fragrances), but no right to win in shampoos despite both products living in a bathroom.

For Little Farm, the consumer’s mental model is “tastes just like home.” So spices in isolation might fail the test (they’re an input, not an output). Sauces might pass if they extend the idea of a condiment that finishes a meal. Korean and Chinese sauces could plausibly work because Veeba and similar brands have proven the path. The framework isn’t go-deep-vs-go-wide — it’s a six-box filter the founders themselves have already articulated, applied honestly to every new SKU.

The corollary on tail SKUs: don’t keep dragging along a tail just because it sells a little. Cut it and use the freed-up working capital and mental space to try a new Be Bold bet. Asked how many SKUs is the right number — “as few as possible.” Every SKU adds inventory, complexity, and decision load.

The grandmother on the jar

A small but interesting detour at the end. Deshpande and Vaidya both push the founders to put their grandmother — the actual person whose recipes started this — on the brand. Reference example: Sweet Karam Coffee names its grandmother mascot after a real person. Amazon’s “empty chair for the customer” trick gets reframed: in your meetings, leave a chair for your grandmother. Every product decision goes through that filter. There’s a moment where Aditi mentions she has a 60-year-old lemon pickle her dad made still preserved at her bua’s house, and Vaidya basically tells her to put it in a glass case in the middle of the office. Bar set, decisions clarified.

Sampling ideas worth stealing

Two concrete tactical ideas worth noting:

Matri-achar sampling packs — small two-piece matri sachets bundled with a sample of pickle, given out at quick commerce delivery (a la the Cheez-It cracker pack). Solves the “sample sachet gets lost in the bag and never opened” problem.
Newspaper-style apartment drops — partner with Country Delight / Milk Basket to drop sampling packs at apartments around the dark stores already fulfilling them. Each Blinkit store serves 10-15K homes; sampling at the same geographic footprint compounds.

Key Takeaways

Don’t go retail because the market is big — go where you have a structural advantage and where someone else is funding your distribution
Quick commerce is the new kirana — fewer brands per slot, top-five lock-in, and a supply chain battle dressed up as a branding battle
Out-of-stock damages the algorithm — the cheapest growth lever for a brand at this scale is just being in stock 95% of the time
70-20-10 capital allocation — Be Big (core), Be Fast (adjacent A/B tests), Be Bold (foreign bets, make a little / sell a little / learn a lot)
Founder energy and founder focus are different variables — the framework lets you indulge excitement without blowing capital
Right to win is a consumer question, not a founder question — bathroom adjacency doesn’t mean shampoo adjacency
Custom GPT for daily demand planning — feed it your MIS, your demand plan, and ask it to flag stockouts before they happen
Cut the tail to fund the next Be Bold — every SKU costs working capital and decision bandwidth
The 10x is usually inside the channel you already understand — Little Farm has ₹350 Cr of headroom inside quick commerce alone before retail becomes worth thinking about

Claude’s Take

This is one of the more useful BarberShop episodes because Vaidya does the unglamorous thing — talks two ambitious founders out of the exciting move and into the boring one. The “don’t go offline yet” advice is the kind of thing investors and mentors usually whisper privately while publicly cheerleading omnichannel. Hearing it stated plainly with the math attached is rare.

The Be Big / Be Fast / Be Bold framework is the most portable idea here. It’s a tidier version of the Three Horizons framework McKinsey beat into a generation of MBA students, but with the bandwidth-allocation honesty that Three Horizons usually lacks — i.e., you’ll be tempted to flip the ratios because the new stuff is fun, and the framework’s job is to stop you. The 85-90% / 7-8% / 2% revenue split he claims emerges from this is suspiciously clean, but the directional point holds: a Be Bold experiment that succeeds gets to graduate, and that graduation pipeline is the real reason to fund it, not the experiment’s standalone revenue.

The AI demand planning bit is the most actionable specific takeaway and the most underwhelming as an “AI use case.” It’s a structured prompt over a spreadsheet. But that’s also why it’s the right entry point — most operators don’t need an AI agent, they need to stop running their core supply decisions on Excel intuition. Calling it AI gets it built; calling it “a better dashboard” doesn’t.

The weakest part is the title. “How to Use AI to Hit a ₹500 Cr Goal” is YouTube SEO bait — there’s maybe four minutes of AI talk in a 66-minute conversation, and the rest is classical FMCG strategy that would have been valid in 2005. Which is fine, because the classical FMCG strategy is the actually useful content. Score: 8/10. Strong founder-conversation episode, well-structured advice, with one genuinely transferable framework and a few tactical ideas worth stealing.