- The video features office hours with startups during YC, where partners give advice to startups individually
- The first startup, Salary Fairy, helps users learn their value in the job market by crowdsourcing salary predictions
- Users are using the service out of curiosity, but the goal is to make it something people desperately need and act on
- The startup currently has 9,000 users with a 10% weekly growth rate
- The partners advise the startup to focus on making something that employees love and find valuable before expanding to other areas
- The startup should track user retention and focus on making a big difference for users
- The partners suggest finding a metric that shows how many people are using the service to make a real impact, rather than just for curiosity
- The startup should focus on building something that users become dependent on and love
- The startup currently has a 30% return rate for users, who are mainly checking results or making more predictions
- The accuracy of the predictions is determined by user feedback, with 40% finding them fair, 30% finding them low, and 30% finding them high
- The startup is acquiring users through platforms like Hacker News, Reddit, and Product Hunt, but needs to think about long-term scalability and growth strategies
- The video then moves on to another startup called Pair Up, which is like the Hotel Tonight for excess food inventory
- Excess food inventory refers to food that stores can't sell and would otherwise go to waste
- The startup aims to connect this excess food with people who need it
- The startup has conducted research and found that traditional food banks and pantries are not effectively handling excess food inventory
- The partners express interest in the concept of Pair Up and wish the startup luck.

- The company is building a marketplace where vendors can sell excess inventory at a discount to users who are willing to buy it
- They have conducted a successful email proof of concept with 6-7 vendors and had open rates of 48%
- They are currently developing a mobile app and hoping to have it coded and developed in the next 2-3 weeks
- They are targeting college students and early tech adopters as their main users
- They are currently manually approaching vendors to sign up, but are also receiving recommendations from food waste consultants
- The app focuses on providing users with access to great food at low cost, while also highlighting the environmental impact of food waste
- The average discount rate for listed items is 50%, with a range of 25% to 75%
- The company is planning to target specific areas and specific verticals to ensure a good user experience and high user satisfaction
- They are currently in private beta and plan to submit the app to the App Store in the next few weeks.

- The team made a mistake by trying to accommodate different user groups, resulting in a lack of focus.
- They decided to simplify the product and focus solely on couples.
- The downside of focusing on couples is that there is no inherent virality in the app.
- They were inspired by the success of the app Couple, which also focused on couples.
- During beta testing, most users were couples, which influenced their decision to focus on this group.
- The importance of discipline and differentiation in a crowded market is emphasized.
- The idea of simplifying the product and pulling transaction streams is praised.
- Acquiring users and achieving high activity rates are key challenges for startups.
- The team currently has 40 beta testers, with about 25 using the app regularly.
- The decision to launch on the App Store is discussed, with the suggestion to go ahead but not expect significant traction.
- The importance of continuous feedback and iteration is stressed.

- Salary Fairy is a startup that helps users determine their value in the job market through crowdsourced salary predictions.
- The startup is advised to focus on making something that employees find valuable and to track user retention.
- Pair Up is a startup that connects excess food inventory with people in need, targeting college students and early tech adopters.
- The startup conducted successful email proof of concept and is developing a mobile app.
- The app focuses on providing users with access to discounted food while highlighting the environmental impact of food waste.
- The startup made the decision to focus solely on couples and is inspired by the success of the app Couple.
- Acquiring users and achieving high activity rates are key challenges for startups.
- The importance of continuous feedback and iteration is emphasized.

Sam Altman and Garry Tan Hold Office Hours at Startup School NY 2014

So next up, Gary, Tan, and I are going to do onstage office hours.

The partners meet individually with startups and we give them advice about whatever problems they're facing.

But we're going to try and do 3 startups in 25 minutes.

And yeah, looking forward to meeting them.

So Kat, can you send out the first startup?

So I guess to start, you should just tell us what you do.

We help our users learn their value in the job market.

And we do that by crowdsourcing their seller predictions.

Do users mostly look at this when they're changing jobs?

They don't necessarily have to be changing or looking for jobs, but they, according to surveys, 40% of American professionals have doubt in mind that they might be underpaid.

So are users using this just to find out if they're underpaid or because they're using it in salary negotiations?

Is it just for curiosity or are they really like saying, I'm underpaid?

They are using it for curiosity right now.

So these services tend to be most valuable when you actually give people something, not just for their curiosity, but this critical thing they need to have for something they really care about.

So how can you make this something that people are not using just for curiosity, but they're using because they really desperately need this information and they're acting on it?

Yes, so our aim is to create an efficient market for the job market, right?

So basically make the job market behave more like the stock market.

So this is the first step for our goal to price salaries accurately, and then get 1 step closer to that stock market for jobs.

Surely you've found people who are radically underpaid.

We are creating the other part of the platform where employers will come and they are going to be able to see professional profiles along with predicted salaries.

And they will be able to make salary offers based on that information.

And before that, before we even built that, we have users who have reached us and told that their seller-favorite prediction helped us boost their courage to ask for more money.

It's really hard to try to do multiple things at once as a startup.

And if you're trying to sort of make this great for workers and for employers at the same time, that's usually more than 1 startup can handle.

I would just focus entirely on making something that employees really love and that really helps them sort of get their market value.

And you should track, like, how many employees come to your site, find that they're underpaid, and are able to use that to make sure they're getting a fair offer and market comp.

And that's, like, you really can only just, you have to just focus on 1 tight little thing.

But until you have users that are telling you, like, this is the best thing ever, I got a 20% raise, and I told all my friends to sign up for the service.

You definitely shouldn't expand to other areas.

And I would try to really find out, like, try and really find a metric that lets you focus on how many people are not just using you for idle curiosity, where they'll use 1 time, forget about it 5 minutes later, not tell their friends, not come back, but how many people you can actually sort of like make this big difference.

And if you can do that, then you'll have all these other high-class problems later, like how you get employers into the market and how you make this efficient market.

But 1 of the things that we always tell startups during YC is that it's way more important to build something a small number of users really love than something that a lot of users find a little bit interesting.

What you'd like to do is make something that a lot of people really love, but a startup can never do that.

And so you end up having to choose 1 of those 2.

And a narrow focus on something that users become really dependent on.

You want to get to the place where your users are telling you, like, I would be so bummed if this product went away.

I'm going to use this for every future negotiation.

So for everyone who's tried the service, 30% have come back at least once.

And what are they doing that second time?

They want to see if they have new predictions.

And the second group is just coming back to make more predictions.

If you have better predictions, you get more scores.

And we have users that have made hundreds of predictions.

Like, how do you know the predictions are accurate?

We ask users if they find their prediction fair.

How will you know when you have found something that users have become really dependent on?

What are you looking for in terms of behavior you'd like

When we first started, everybody in our platform were able to predict each other's salaries anywhere in the world.

And then we got enough users, we were able to divide into countries.

And then we got enough users so that we were able to divide into cities, and then industries.

Right now, for instance, my salary predictors will be technology folks in New York City Where we want to go from here is even further than that.

We want to be able to divide people into companies divide people into years of experience for instance say that you have got a job offer from Microsoft and you want to evaluate an offer, then you will be able to crowdsource your prediction from folks who have worked at Microsoft as the same title as you.

So And there is no way to validate if you are underpaid, right?

You need to find a job that pays you that amount.

Even if the statistics, if you look at the labor data, you cannot say that you are not underpaid because you are paid as the same because the market has moved and the landscape changed.

So only way to validate these predictions that we give back is to find them a job at that price point.

Yes, Hacker News, Reddit, Product Hunt, and we are pitching to reporters using Help Reporter.

So the thing to remember is that those are great ways to get initial users, but that does not scale forever.

There are a lot of different sites where people can come to get salary information.

And you guys need to be like 10 times better than any of these other sites.

And it's worth thinking about, A, what you're going to do to make yourself so much better that people will tell their friends use SalaryFerry and not any of these other salary sites, and B, if there's some way you can build growth into the product.

Because this is not, there's not any like inherent viral piece in this, but you could probably build something really cool where you send out an email saying, guess my salary to a bunch of your friends and then they sign up.

Building those, like a lot of people think about I'm going to build this product and I'm going to get my users on Hacker News and then it's just going to grow.

If you build a sufficiently great product, sometimes that happens.

But you don't want to have growth be an afterthought.

And if you can build growth into the product at this early stage, that can be really helpful.

And I would definitely think early about how to do that.

We're about to run out of time on this 1.

OK, well, I'm a believer in crowd prediction, so it sounds cool.

Pair Up is basically the hotel tonight for excess food inventory.

So excess food inventory is often, you know, stuff that stores will pull from shelves 2 to 3 days in advance of sell-by.

It's stuff that stores can't sell through at the end of the day or that they're throwing away, turning over every 4 to 6 hours because they have a, like, fresh daily policy or, like, a fresh, you know, every 4 to 6 hour policy.

And it's food that doesn't necessarily have a place to go.

So if you're thinking about, oh, well, you know, what I thought growing up, you know, food banks and food pantries are taking care of this excess.

We found in our research that that's actually not the case.

Because of their own budgetary limitations, they're not as nimble in terms of pickup.

And so City Harvest, for example, has a 50-pound minimum when it comes to pickup.

And A lot of small to medium-sized businesses in New York City just aren't going to have that on a daily basis.

And so what they've got now is basically like you can throw it away, and you can pay action hauling to pick it up.

And So we found that to be highly inefficient, very wasteful.

And so what we're actually building is a marketplace that allows vendors to sell through that excess inventory at a discount to users who are willing to buy that excess inventory at a discount.

Why do you use Hotel Tonight as the metaphor?

Well, so Hotel Tonight is basically kind of what we're doing only in the travel and hospitality industry, because they sell excess hotel rooms, hotel rooms that the hotels themselves can't sell through whatever services they use.

So the target demographic of what we're looking at are like people like us or like college students, those types of folks who are interested in kind of early tech adoption but still, you know, don't have or want to have kind of food at a discount.

And what's the kind of typical, like, so they would buy an individual meal for them and their roommate or something

Yeah, or like a half dozen donuts, or eventually we're hoping the boxed soup from Trader Joe's, or something from the Prepared Foods counter at Whole Foods.

So we launched a kind of email proof of concept about 2 months ago.

Basically we just wanted to test a hypothesis of if people would actually go into the store if they got that information.

And so we had a subscriber base of maybe like 350 people.

So we had open rates of about 48% on our email.

Per listing was maybe like 4 or 5 people per listing.

So that's actually pretty decent, given that the quantities were fairly low anyway.

Did you get everything that was listed sold?

But some stores were better than other stores based on the items that they were selling.

What were the things that sold well and what didn't?

Yeah, so the things that sold well were like sandwiches and croissants and those types of things.

The things that didn't sell well was kind of like, there was this 1 bakery that had kind of an evergreen bread deal.

So people knew that this was kind of running every single day.

So they weren't kind of incentivized to go in today versus tomorrow versus the next day, because they knew that it was available always.

And it was also in a more, like, residential neighborhood.

They're, like, all at work somewhere else.

So we found the list to be fairly restrictive.

So it was successful in what we wanted to test.

But it wasn't successful in the sense that users only wanted 1 email a day.

So they all had different closing hours, and they all had different abilities to estimate their inventory.

So actually, I just finished wireframing them this morning.

And we're hoping to have it coded and developed in the next 2 or 3 weeks, which, yeah, it's kind of soon.

And kind of what the move is, is we are building this web app now to kind of test the hypotheses around live listings and kind of location-based searches and food searches.

And then the mobile app will hopefully be able to take that even further with alerts and being able to do location-based searches, kind of like when you're out and about in the city and stuff like

CHRIS RAMSDALELLER My 2 thoughts would be, I'd try to get the mobile app as quickly as possible.

That when people try to test these things on the web, if it's really the sort of a thing people want to use on mobile, it never works as well.

And the other thing would be hyper-focused when you start, not even a whole specific city, but very specific areas and, like, the specific verticals, like croissants and sandwiches, that you learned work.

Product matters, and density definitely matters as well.

And also, like, the types of vendors that we're working with.

We want to make sure they're a place that people want to go on a regular basis, so that if there is a discount, they're even more incentivized to walk in.

Yeah, but making sure that everyone that uses it has a good experience, and that they find something that they like, and that there's something nearby.

And no matter how much you have to limit it to make sure that those first 1,000 users really, really love it, it's almost always worth doing.

How are you going to get all of the people that have this excess food to sign up?

Are you just going around and banging on the doors?

See all Y Combinator transcripts on Youtube

Office Hours at Startup School NY 2014