Manas J. SaloiModify

2025-05-15 08:00:00

Zepto is now suggest swapping expensive items in your cart for cheaper alternatives.

This isn’t your typical up-sell or cross-sell. It’s actually a “down-sell” through cart swaps. They’re actively encouraging you to spend less money on certain items. Seems counterintuitive.

The UX is super simple too. You see an item, they suggest a cheaper alternative, one tap and it’s swapped. They’re showing this after you’ve built your cart because that’s when price sensitivity kicks in - when you see the final amount.

Why I think they might be doing it:

• Testing price sensitivity - They can gather valuable data on how willing users are to switch brands based on price. This is gold for their algorithms.
• Moving unsold inventory - Got items approaching expiry date? Get them into carts through swaps.
• Clearing overstocked inventory - Rather than discounting everything, target specific overstock items through strategic swaps.
• Boosting preferred sellers - If you’re a new challenger brand launching on Zepto, there’s a clear opportunity to accelerate your growth. Zepto can offer you higher sales in exchange for better commission terms. As a preferred seller, you’ll benefit from stronger visibility and improved placement. Additionally, Zepto can leverage Swap to drive up your daily orders, helping you build early momentum and gain traction faster.
• Private label - Tomorrow Zepto will nudge price sensitive users towards their own private label.
• Each line item matters - Back at GoMart, Gojek’s groceries delivery product, we used to see drop-offs when packaging fees felt bigger as a proportion to the cart. Users judge the entire order at each line item.
• The switched product could be a sponsored product - Zepto reduces AOV by ~5%, impacting their margin on the order by 5%* 10% (0.5%), while the ads revenue might be ~1%. A friend told me about a payment switch in another on demand company, which essentially acted as a sponsored payment product, where if you switched to another bank and did more transactions through the payment method of the bank the on demand company would make more. Brands can bid for these sponsored spots.
• Increasing margin, Margin over AOV - Sometimes a lower-priced item might actually have better margins. In grocery especially, small margin improvements compound significantly at scale.
• Building trust - This is perhaps the most underrated benefit. When a company actively helps you save money, it creates tremendous goodwill. Users feel like the app is on their side. This also helps counter the narrative around the dark patterns they are accused of deploying across the app.
• Avoiding you cross some price threshold - Psychological thresholds matter a lot in commerce. When your cart hits certain price points (₹500, ₹1000), conversion drops. If a single expensive item pushes you over that threshold, you might abandon the entire cart. By suggesting a swap to keep you under that threshold, Zepto potentially saves the entire order.
• Conversion to single cart - I don’t use Zepto, so not sure this is relevant, but on Swiggy I sometimes end up with two deliveries because one item lives in a different store. If the small second cart is only there for a pricey cone, I drop it. A swap keeps everything in one basket so the order feels worth the delivery fee. This can be one a way to solve it.

I believe this feature helps with retention. When users feel an app is helping them make smarter purchases rather than just extracting maximum value, they’re more likely to return.

The beauty is in the simplicity. It doesn’t feel intrusive or like a hard sell - just a helpful suggestion that the user can easily ignore.

[Edited by Claude, my editing buddy. It’s time it starts delivering ROI on my $20 monthly spend. It can’t just be a personality hire anymore.]

2025-05-10 08:00:00

I have been talking to friends who have started companies, I am an unpaid intern in half a dozen of these companies, and I also meet quite a few VCs thanks to “lets exchange notes” catchups, and I can see a stark difference in how VCs used to evaluate founders and markets now compared to a few years back. Previously in SaaS, people would assume that because in India, the team sizes would be lower and costs would be lower, you could price your software lower. This would enable you to compete with foreign SaaS players, capture significant Indian market share, and grow. Maybe you could even compete in the US like Freshworks did.

But now that narrative is over, or at least dying, and many VCs want Indian founders to explain how they will establish a successful US GTM strategy.

SaaS investing is at its lowest (?) globally. The narrative has shifted to AI. And AI agents. Earlier you could argue that your opex will be lower due to lower salaries in India, even though infra cost would have been the same (AWS), but in today’s time the narrative is not lower salaries, but having the fewest employees. Especially in product development. As much so that people are actually showing fewer employees in Linkedin because VCs have started measuring their portfolio companies’ productivity. And sending them memos on how to be AI first. India was always a hard place to sell SaaS vs the US.

The value of an Indian’s time is probably 1/40th of that of a US person. Of course, you can argue this isn’t true for high-value workers. An Indian software engineer might earn around 30 lakhs, which is approximately 35,000 US dollars, while in the US, it might be 150,000, so it’s only ~5 times higher. But people value their time very differently.

For example, if you’re selling some QA automation test suite to someone in the US, they will say, “Sure, this product is saving time for me, and I can’t hire an infinite number of people. I can’t solve problems by throwing people at it. As long as it’s giving me ROI, I would definitely buy it.” Meanwhile, Indians will be thinking, “I can just assign more people to this task.” This approach is very common in India. That’s why many Indians are tired of selling to Indian companies, especially SaaS products. Yes, your sales team will not cost you what a US Sales team would cost, but in India time to close deals will be higher and your ticket size would be lower too. India is a cost advantage market and a breeding ground for lean SaaS is no longer sufficient when thinking about SaaS.

I know some people will get upset when I mention this, including some of my friends. But this is what most VCs want to hear: What gives Indian founders the right to win in the US versus the average Stanford graduate?

If you eventually wanted to sell in the US, earlier you would say, “There is Freshdesk, which is cheaper and almost as good, so why pay for Zendesk.”

But when it comes to AI, if you are an Indian founder wanting to figure out US go-to-market, compete in the US, and have some AI solution for customer support or similar applications, the VCs will ask, “What sustainable repeatable customer acquisition playbook do you have that makes you win in this market?” What’s your GTM wedge? Will you do cold outbound? content? partnerships? What unfair insight do you have? Are you just another feature or do you replace a budget line?

Previously, at the pre-seed stage, VCs would simply back founders who were targeting a large market. Founder pedigree is all that mattered. At seed stage, they might look for some signs of product-market fit. But now, people want to see a detailed GTM playbook and much more depth. Investors are concerned about Indian founders not being able to figure out how to sell in the US, thereby capping the their growth.

In the US, AI companies (in every domain) have probably raised millions of dollars. YC invests in dozens of them every batch. So how do you compete in overheated markets?

I know a founder who is building a product in vibe coding that is better than Lovable. Anyone I have asked to test the product says it is probably the best vibe coding product available. But the competition is Lovable, which has scaled quickly to 40 million in annual recurring revenue.

Again, the question becomes, “How can you compete with Lovable and Bolt?” You need to run paid marketing campaigns, create a compelling narrative, and do far more as a later entrant. Your product being 15-20% better probably does not help enough. If you say the market is big, and even 10% of that market can lead to meaningful ARR, it won’t work too, because the VC will ask for every incremental user evaluating Lovable vs your product, how would they know you even exist. Second why are you in consideration. And finally why would they choose you over Lovable or Bolt. The only option is to raise big. Be as unhinged as the Cluely founders. Become generational marketers. And learn how to get cheap distribution and always be top of mind. Most founders are introverted. They won’t even put out their funding PR. I know, I try my best to change them. But again I am an unpaid intern after all.

You have to tell a story of how you will be able to win against Lovable. Just being incrementally better does not help in today’s competitive landscape. US GTM is hard. A VC friend was joking that you have to turn yourself into someone you are not. Embed yourself in networks in the US. Hustle till you can talk US football (not Soccer) with your potential clients. Say the right things. How can you do that as a founder who probably has not even set your foot in the US? Clay took years to reach their revenue. So did Figma. No Indian VC will wait for you like people did with Clay and Figma in the US.

I know so many founders sitting with 200K USD ARR who are trying to figure out the next steps. VCs will say “Hey look at this US company that grew to 20 million ARR with 10 people in 12 months”, and ask you to spend money, and 3 months later you will realise they got cold feet because of the sector heating up, change their thesis, and say good luck to you and stop responding to your mails. What is worse is the borrowed conviction. A lot of these VCs won’t even use the products of their portfolio companies, so they are just FOMO investing. It is an extremely tough environment out there. And I sympathise with both my founder friends, but also VCs who will have to figure out their investment strategy for the future. Get ready to answer what is your repeatable US GTM playbook in your seed pitch itself. One off contracts, even in the US, won’t help much. You have to show how you can reproduce your sales. ZIRP is over. The only thing we can do is grind.

2025-05-09 08:00:00

People are discussing why OpenAI acquired Windsurf on x dot com. It’s clearly not for the revenue. A $100 million revenue stream won’t add much value to OpenAI, which is already valued at $300B. While the Windsurf team is certainly talented and acquiring them is beneficial, I believe the main reason is that while OpenAI has become the default application for general chat, they’re lagging behind Claude and the rapidly improving Gemini models when it comes to coding capabilities.

How does OpenAI catch up? When building better models, they typically conduct Reinforcement Learning from Human Feedback (RLHF) through partners like Mercor and ScaleAI who help with labeling tasks and providing feedback on outputs of these models. Imagine you are trying to generate iOS code, you would need a lot of iOS experts from Mercor, Scale AI, and Turing AI to give you feedback on the code generated. There is not enough iOS code out in the wild for foundational models to train on. This becomes expensive and consists of repeated transactions with the human labellers who are not even in a direct relationship with the foundation models and sourced from these third parties who source these experts. OpenAI of course don’t want these operational expenses (additional head counts of experts) on their balance sheets, nor do they want to continually go through third parties that will cut into their margins.

The best approach is to acquire software that enables an ongoing feedback loop. Currently, when people use ChatGPT to generate code, OpenAI receives no feedback about whether users applied the code, used it, what iterations they made, or if they even pushed the recommended code to GitHub. There’s no feedback loop - OpenAI provides output on ChatGPT client and there is no accept, reject button, no response by the user on whether it served the purpose. Users might not even say “thank you, this worked” because that costs additional tokens. And Sam Altman himself has been complaining about how much money OpenAI burns because of these “Thank you”s in general conversations.

With Windsurf, users interact with agents that rely on foundation models and providing continuous feedback. The system tracks user satisfaction with responses and subsequent actions. More importantly, since Windsurf is a wrapper around multiple models, by acquiring Windsurf, OpenAI gains feedback not just on their own models but also on competitor models. This creates a permanent data and feedback loop that will help build better coding agents.

Amjad from Replit has suggested that the key to achieving AGI is having the perfect coding agent, as it can code whatever you want and eventually self replicate. Don’t remember the source though (where he said it, apologies if I misremember). Foundation model companies know this. Claude now has Claude Code, and OpenAI has launched a standalone coding product recently that you can run from your terminal called OpenAI Codex CLI.

However, these products might not be sufficient because they’re terminal based and targeted mostly at professional developers. Yes, Pro devs use IDEs too, but my point is that currently what foundational models have shipped are terminal based coding agents. Windsurf and Cursor have access to much more data because their user-friendly interfaces attract not just professional developers but also novice coders, “vibecoders”— essentially the entire spectrum of developers. This results in significantly higher volumes of data and feedback on response quality. I believe an IDE with the largest base of coders will always outperform the current iterations of Claude Code and OpenAI Codex when it comes to data generation and feedback loops. After reviewing dozens of transcripts from companies like Mercor and ScaleAI on Tegus, I’ve understood the costs these foundation models and application layer products incur to ensure optimal performance in specialised tasks like coding through human-in-the-loop feedback.

Coding is a P0 use case for foundation model companies, and it seems inevitable that Anthropic will also try to acquire a coding IDE. Sundar Pichai following the Cursor account on x dot com has led to speculation that Google might eventually acquire Cursor. There will be significant competition for Cursor though, with their impressive $300 million in revenue, and recent 900 million raise, they likely won’t sell cheaply anytime soon.

Google has recently released their own coding IDE, but I doubt people will adopt it given Google’s history of shutting down products that don’t achieve their desired scale.

2025-05-08 08:00:00

I don’t like the new OpenAI voice-to-text input.

The earlier version had a more thoughtful flow:

• Active listening indicator: You saw a timer and microphone animation, which gave feedback that it was capturing your voice.
• Transcription preview: After you spoke, the transcription was visible before sending. You could edit any mistakes.
• Explicit submit step: You were in control of when the message was finalised and sent.

Now:

• There’s no clear indication of how long you’ve been speaking. It just says listening, then the state changes to transcribing, and then it just blasts off the query.
• It auto sends the transcript without confirmation. And once the response comes you realise it could not understand your accent and wrongly transcribed like 5 critical words that changes the entire context of the query.
• If the model mishears you, you can’t correct it unless you manually redo the entire message. Or correct after ChatGPT responds to your first message with typos.
• And God forbid it throws errors like “I didn’t catch that”, especially during longer recordings, there’s no option to retry, and you’ll have to repeat everything you just said.

This “faster but dumber” flow works if everything goes perfectly, but when it doesn’t, it’s frustrating. Claude does something similar. No intermediate correction step, does not let you correct the transcription. Thankfully it still shows the timer.

Some PM at OpenAI probably optimised for speed and simplicity (1 tap voice interface), without realising voice messages are different from voice to text query sent to an LLM. The new UX assumes high transcription accuracy and fast LLM turnaround, which isn’t always true, especially with accents, ambient noise, and longer thoughts.

2025-05-07 08:00:00

Wander’s Counter-Positioning Against Airbnb & Its Self-Imposed Constraints.

Disclaimer: Never stayed in a Wander, but have been following the company for years ever since Kyle Tibbitts (a marketer I respect) joined as the CMO.

If you listen to TBPN, you must have heard Jordi Hays and John Coogan shout “find your happy place” enough times to google Wander and see what it is. Now they are raising a Series B and the size going would be hit soon on TBPN I guess. 

Let’s get back to the real topic: Why I love Wander and why it is another example of counter-positioning done well.

The vacation rental market is interesting. Let’s talk Airbnb. On one hand, you have Airbnb, which has become a verb at this point. On the other hand, you have thousands of horror stories of people showing up to Airbnbs that look nothing like the pictures. Surprise cleaning fees. Hosts who don’t respond. Hard to find locations. Low touch experience at the price of a hotel.

Wander has been deliberately going after this problem. They are a counterposition to Airbnb. They have created a ‘hotelified’ alternative to Airbnb for the top 1%. They’re curating the top 1 percent of vacation homes, professionally managing each one to guarantee consistent, luxury-grade experiences. No surprises, no mismatched listings, no unreliable hosts. This is what Oyo was supposed to be in India. But couldn’t.

Wander is something Oyo could have been if it was a luxury product. “70% of Wander guests are affluent, with net worths exceeding $1M,” according to Wander’s website. “Wander’s audience is primarily composed of CEOs, founders, executives, software engineers, and individuals working predominantly in the tech sector.”

What makes them interesting is their vertical integration. Unlike Airbnb, which is just a marketplace where anyone can list their property, Wander owns and operates everything end to end. Again, like how OYO promised. They buy the properties. They renovate them. They furnish them. They handle the bookings. They manage the properties with their own team. This creates a moat that Airbnb can’t easily cross without completely changing their asset-light approach. Airbnb’s strength is also its weakness: with millions of properties, they can’t control the experience.

[Check sentiment on the Airbnb experience over the last few years on Twitter.]

There’s an obvious tradeoff. By purchasing, renovating, and operating each property itself, Wander takes on all the principal risk. This capital intensive model makes it harder to expand quickly and limits their selection compared to Airbnb.

It’s a classic case of self-imposed constraints. While Airbnb can add new hosts at internet speed with near zero cost, Wander has to buy and upgrade each property. They’re deliberately capping their total market size in exchange for higher unit economics, stronger brand loyalty, and way better guest satisfaction.

Every property has the same hotel-grade amenities: super-fast Wi-Fi, private pools or hot tubs, gyms, and 24/7 concierge service. No wonder Wander’s guest satisfaction is high (see the testimonials on their site). They’ve built something called WanderOS, which centralises all property controls into a single app. Door locks, lighting, temperature, even Tesla access. Yes, they provide a Tesla at all their vacation homes.

This appeals to travellers who want the freedom of a private home but demand the reliability, transparency, and service levels of a luxury hotel.

They’ve also got an AI-assisted concierge team that handles guest inquiries with site-specific data, automatically creating maintenance tickets or escalating to humans when needed. This tech investment lets them manage dozens of homes with fewer property managers, delivering consistent quality with higher margins.

Wander isn’t doing the 1% improvements to Airbnb’s model. They’re creating a step function change by completely rethinking the approach. They’ll never match Airbnb’s scale, but they don’t need to. They’re building a profitable, defensible niche in the high-end segment of the $1.3 trillion lodging industry.

It’s a classic example of counter-positioning. They’re building something that their bigger competitor can’t copy without undermining their own business model. 

And they are also capping their growth by constraining their growth.

And I love these kinds of bets. Let the big platforms scale to infinity with lower margins while carving out the premium segment with better economics and happier customers. Not every business needs to be a winner-take-all marketplace.

(To deal with the capital constraints, they launched Atlas in 2022, which they call the world’s first vacation rental REIT. This moves property acquisition costs off their balance sheet while letting them keep operational control. Smart way to unlock investor capital without dilutive funding. You can read the details in an amazing Deep Dive done by Packy McCormick way back in 2023.)

[Post created by thought dump on ChatGPT and then verifying some of the details on Packy’s post and Wander’s own website. Then edited with Claude because Narayana knows I don’t have the patience for it.]

2025-05-06 08:00:00

Why Rapido moving into food delivery make sense:

• 2W is still their core revenue generating business, note revenue not GMV, and we all know the uncertainties around 2W regulations. It makes sense to not be dependent only on 2W mobility and have more revenue lines
• When it comes to pricing strategy they say that you should gradually raise your price until your target customer literally laughs — that “laughter” marks the upper bound of what the market will bear. Food delivery businesses have been doing the same. They have been layering more and more fees over time. And now it has reached a state where users are unhappy, restaurants are cribbing about commission, and order growth has slowed. So the WHY NOW is clear for Rapido. They have executed this playbook in 3W and 4W ride hailing businesses and have got meaningful marketshare.
• Food delivery business has less regulatory uncertainty. Far fewer regulations too. Restaurants have been doing food delivery for ages. On demand platforms have made it a scalable VC backed business. This means that you have it easy in a lot of things. Driver onboarding becomes easier. Food delivery is moving goods, and not people. So fewer steps in onboarding drivers.
• Fewer steps, easier onboarding, faster time to get onboarded to the platform and start earning means that driver supply can be increased and later the same fleet with additional onboarding steps can start doing ride hailing too.
• Every company is struggling with driver acquisition and retention. This means the more you help your drivers be occupied, the more you help them earn, the higher the likelihood of retention. You don’t want your drivers to be busy only during peak office hours. You want them to be utilised during lunch hours too. A lot other drivers multi platform anyway. So a driver churning midway during the day because they get busy with other delivery apps might mean they not coming back to your app later. This means supply retention for the entire day becomes harder.
• More use cases for your users. Once you acquire them, you can spread the CAC across more products. CAC payback time decreases. You can also afford to pay more CAC because your LTV becomes higher, as users’ usage frequency becomes higher as they now are using multiple products.
• Rapido has raised a big round recently and they have money in the bank. While Swiggy and Zomato are both public companies and won’t be able to cut their take rate because they will want to show revenue growth over the next few quarters, especially Swiggy which is quite new to the public markets. Another reason for WHY NOW.

Why it won’t be as easy for Rapido to win marketshare in Food delivery as 3W, 4W ride hailing:

• 3 sided marketplaces are far far harder to run than 2 sided. Operational complexity increases by a lot.
• Onboarding restaurants require a sales team. You need more Ops people. Your Opex increases by a lot.
• Most food delivery businesses anyway don’t make money in the food delivery part. 25% take rate leaves extremely low profit margin (if any). So how do they make money? Ads. For Ads you need enough demand. Enough users. You need a merchant ads team, need to build your ads platform. Ads can only be run at scale. So it will take sometime for Rapido to make money from this business.
• Delivery time matters. Reliability matters. When Ola acquired food panda they were giving ice cream in 9 Rs. But what is the point of ice cream for 9 Rs if it comes in 1 hour and in a melted state. If you don’t get a rickshaw you cancel. Booking conversion rate for most ride hailing companies in India will be close to 50%. Like you reach your destination 1/2 of the times you press the ‘Book Ride’ button. Here you need 99.99% reliability. You need the food to arrive whenever you order. Else you churn.
• You can’t pool orders when there is not enough demand. The only way to cut delivery cost is by pooling orders. Yes, it is not fun to see your delivery person deliver two orders, and yours being the 2nd one, but without that cost of delivery will be even higher. Running a reliable profitable food delivery business is playing the on demand marketplace game at Hard difficulty.
• This means you need to have enough margin to run the business well. Subscription just does not let you do that. Remember not 2 sided, this is not just about drivers, you have 3 stakeholders now: Customer, Driver, and the restaurant. So you can’t just change your business model, attack the take rate of the incumbents, and hope this is enough to gain meaningful market share.

It will be interesting to see if Rapido can make a dent in the food delivery space. Even global behemoths like Amazon have tried and failed in the past.