KEY TAKEAWAYS
- 01A campaign that bombed and one that worked
- 02What owning the P&L changes
- 03Owned channels and a brute measure of spend
- 04Agents in supply planning and SEO
- 05Why AI can't be the creative director
- 06Returns, referrals and the 100 day rule
THE CONVERSATION
A closer read
Agentic Marketing Podcast: SleepyCat's Kanishk Arya on why AI can't be your creative director
Kanishk built brands at Ola and Urban Ladder before SleepyCat. His biggest learning across all three is that consumers already know what they need. The marketer has to convince them the product solves it better, and make the trial risk free.
At Ola, that meant a good offer for the first ride. "Once they have a great experience, the product does the marketing, they'll come back," he says. At SleepyCat, the trial charges nothing, and an unhappy customer can give the mattress back.
A campaign that bombed and one that worked
After the second Covid wave, SleepyCat ran a campaign telling people to take care of themselves. It bombed. The brand told people something they already knew, without saying why SleepyCat was the answer. There is a thin line between inspiring and being preachy, Kanishk says.
The current campaign with Nimrat Kaur and other creators is sharper. Even a 30 year old now has back problems that once belonged to a 50 year old, and most people ignore the surface they sleep on. Pinpointing that problem is what gave the brand ROI.
What owning the P&L changes
Kanishk owns the P&L as well as marketing. Reviews and testimonials give ROI, but not the ROI of the 100 night trial, a business call because returns have a cost.
Warehousing works the same way. People take weeks to decide on a mattress, then want it immediately, used as they are to same day delivery. SleepyCat invests in warehousing and promises a new mattress in two days. He would rather be faster than spend more on advertising.
Owned channels and a brute measure of spend
Kanishk believes customers will buy where they want to buy, so SleepyCat makes its own channels more attractive, with more lenient trial policies on D2C and retail.
Attribution stays murky, so he tracks overall marketing cost against overall revenue, a trend line that has to come down. His models call Meta probably his worst ROI channel, yet it gets his maximum spend because that is where the customers are. After marketplaces grew faster for three or four years, own channels are growing faster this year as the retail footprint expands.
Agents in supply planning and SEO
SleepyCat's first agent took on supply planning, working out every day what needs to be produced and where it should go. Tools for this can cost five to ten lakhs. SleepyCat did it with a $100 subscription.
The second is a five agent system for SEO content, which needs volume, accuracy and a human sound. "It's not a one agent job. It's a multi-agent job that you try to do," Kanishk says. One agent humanises the content and another checks whether it will pass the AI test, so output is 99% there before a human reviews it. Rankings, he says, are improving.
Why AI can't be the creative director
For the Nimrat Kaur film, the team asked AI for a script drawing on her career and life. It did a decent job, but there was no reaction when they read it. Kanishk's view is that AI does an average job across every part of a creative and never takes a punt on one idea.
So the team wrote it themselves, with a nod to her villain role in Homeland and a line about not changing her mattress for ten years. AI earned its place in review. Given guardrails on the Meta algorithm and the points to deliver, it gave suggestions faster than 20 human interviews would. SleepyCat brainstorms with AI, "but we don't let AI be the creator," he says.
Returns, referrals and the 100 day rule
The biggest reason for returns is the wrong size, because even a king bed in India can have ten different dimensions. SleepyCat has built an AI agent that calls customers to confirm their bed size before dispatch. Human agents save more than 50% of size or feel related returns with a different product, and an AI agent for that is planned in the next 3 to 6 months.
About 30 to 40% of D2C mattress buyers are referred by a friend or family member, and 60 to 70% of mattresses sell with another product. About 23% of customers come back in the first 100 days, so SleepyCat now shows new buyers every category straight away through WhatsApp and Meta remarketing.
Changing teams and a single view of the customer
Kanishk does not expect AI to change the fundamentals of cost. Headcount will not change immediately, but its profile will, with less analyst work and possibly bigger content and creative teams.
His next project is a customer data platform that pulls together signals from stores, the AI chat agent, service calls and the website, so AI can pick the right products, offer, channel and timing for each customer. He wants it built in six months. "I think it's going to be a big value unlock," he says.
Episode 2 of the Agentic Marketing Podcast offers a candid look at where AI earns its place in a consumer brand and where it falls short. Through SleepyCat, Kanishk Arya is showing that taking the risk out of a purchase can do more than another ad, and that AI works best as a reviewer while humans hold the creative pen.
For more such episodes and learnings, stay tuned to the Agentic Marketing Podcast.
AUTHORED BY
MORE FROM THE PODCAST
IGP's Priyesh on sending fewer WhatsApp messages and earning more from each one
Explore
From cloud kitchens to brand love
Explore
AI in Banking: Vision to Impact
Explore
