Just rewrote the UI of Listen Notes (a podcast search engine); feedback appreciated
Hi IH,
I’m Wenbin, a software engineer in San Francisco. I’ve been working on some side projects on and off for several years. However, I had never worked on these projects full-time (e.g., several work days in a row). Things changed recently. I just left the company I cofounded on Sep 15, so I got a chance to work on my side projects full-time.
I spent most of last week to rewrite web UI for one of my side projects: Listen Notes ( https://www.listennotes.com/ ) . I’d love to get some feedback & I’m happy to answer questions! 🙏
Listen Notes is a podcast search engine. You can search ~350k podcasts & ~18 million episodes by people, places, or topics. For example: https://www.listennotes.com/search?q=Dwyane Wade&sort_by_date=1
Listen Notes is a very simple “I can build this in a weekend”-ish project. It crawls podcasts RSS from iTunes, indexes podcasts meta data in ElasticSearch, and provides a simple web UI. No AI. No speech-to-text. No audio search.
Despite of the simplicity of this project and the little development effort, I find that this project is actually good enough for my personal use — some of my friends feel the same way as well.
Let me know if you have any feedback or questions :)
Thanks!
This is very cool and saves me the trouble of building the same thing! :) Have you thought about a "suggestion" feature? For example: I listen to a lot of history podcasts, but when I search for this term, the podcasts top my list aren't shown (likely because of keyword relevance). "The Fall of the Roman Empire" and "Tides of History" are two good examples. I'd love to be able to suggest these two so that I can take advantage of others' similar suggestions. Suggested podcasts could actually help to make your search results more relevant, too, by using suggestions (or something similar) in your sort algorithm.
Yes, there are some TODO items in the following weeks, including autocomplete while typing queries, related queries, podcasts similar to this, semantic search rather than just relying on keywords ...
Keep it up, though. I have been looking for something like this and it seems like your product would fit the bill.
Cool project! Seems quite useful. One challenge here is competition, since people might just default to iTunes search, and iTunes brings extra information to the table to improve results (e.g. download numbers to sort by popularity). Any long-term plan to provide extra benefits of your own?
Good question. I do have some thoughts on competition.
There are forces of competitions: 1. Big companies; 2. small startups.
1.1. Apple
Apple is in a very very good position to make podcast great again, i.e., biggest podcast player app + biggest podcast directory. But, Apple simply doesn't care about podcast. If we were running Apple as a company, what can convince us to invest in the podcast area? Can we make big money from podcasts? Or should we invest more in hardware (e.g., iPhone, Watch) and other online services (e.g., Music) that can bring in real revenue? The definitions of "big money" in the eyes of Apple and us (indie hacker / small business) are different in several orders of magnitudes :)
1.2. Google and others
Google has image search and video search. Can it easily build an audio search? Maybe. If Google sees real value / $$ in this podcast search thing, they may build it. This won't be a bad thing actually. The competition from Google help validate the market. Let's allow Google to own 80% of the market, and other small players (like me? haha) own other 20% :)
People always wonder why Instagram could make it, given that Flickr and Picasa were already big in 2010 -- how hard for Yahoo & Google to throw together an iOS app? http://www.businessinsider.com/why-google-cant-build-instagram-2010-11
I'm worried more about small startups than big companies.
The selling point of those startups is (almost) always AI / audio-search / speech-to-text, which looks good from surface (e.g., easily go to top of Hacker News or Product Hunt) and makes fund raising easy. But they can index only a small subset of podcast episodes, due to the slowness / accuracy of their AI / audio indexing ability. Imagine this, it won't be too useful if Google indexes only 2,000 web pages.
Listen Notes is way too simple for a big company or a well-funded startup to build. There's always ego that gets in the way of these companies, which oftentimes makes their solution over complex. Instead of "depth-first" approach (i.e., indexing a small set of podcast episodes using high-tech AI/audio-search), Listen Notes is "breadth-first" approach (i.e., indexing just text description + title meta data for the whole internet's episodes). In the following weeks, I'll try to improve the core search experience + search quality. Listen Notes can use the help of AI/audio-search techniques in the future, when these techniques become mature enough -- I doubt that the first version of Google Image & Video search were indexing just text meta data (or primarily).
That was a very thorough answer :-D I agree that you shouldn't worry about large companies going out of their way to compete with you. For one, it's unlikely (although podcasting is growing in popular so that may not always be true), and two, there isn't much you'd be able to do about it anyway.
What I was actually more interested in is the value proposition for users. Why does someone choose to search Listen Notes instead of just searching iTunes? What makes it better?
Three things.
Search quality. My friends and I ran some tests and compared search results among Listen Notes, the default iOS podcast app, and other popular podcast player apps. For most queries, Listen Notes returned more results (and more relevant). I think the "breadth first" approach makes Listen Notes better, for now.
Listen Notes is a web app. A web experience is good for casual listeners or non-listeners to explore what's out there, e.g., no need to install app, cross platform, friends can easily share URLs... There are more non-podcast-listeners than listeners at this moment (way more...), just like there are more non-Internet users than internet users on the earth. iTunes doesn't provide search function on web.
Listen Notes is episode-centric, instead of podcast-centric. Most podcast players focus on podcasts, instead of episodes. It's like in early days of web, when people bookmarked a lot of websites. Then Google becomes a verb, and people oftentimes search for just web pages, instead of websites. I rarely subscribe to new podcasts nowadays. I just binge listen to a bunch of episodes related to same topic -- it's still inconvenient though; I have to subscribe to those podcasts, finish listening those episodes that I'm interested, then unsubscribe in my podcast player app (Overcast) :(
A fun comparison: there are 18+ million episodes in Listen Notes search index; there were 25 million web pages in Google's index in early 1998. Get the volume and make it just good enough to use. Then iterate from here :)
I still remember when I first heard about Google back in 1999 or 2000 (in China). Back then, people told each other this search engine was quite good and you could always find relevant things. I hope Listen Notes could have such reputation in the "podcast search" world :) This requires a lot of focus and relentlessly improving the search quality.
Very cool, that explains quite a lot. Now that I understand all of that, here's some feedback I have:
Becoming useful on mobile.
I'd guess that the vast majority of podcast listeners listen on mobile. I listen while I walk to the office. Many fans of the IH podcast listen during their commutes as well. The web is a big differentiator for you, because it's a far better platform for doing any sort of heavy lifting (searching and filtering through hundreds of episodes to find what to listen to), but after I do my filtering I'm going to want the episodes I've chosen to be available on my phone instantly via an app where I can listen to them.
Remembering my decisions.
Unlike webpages and Google search, I don't really want to see the same podcast episodes over and over when I do successive searches. If I see something in the search results and make a decision about it, I want that decision to be permanently saved: (1) I've already listened to this. (2) I'm currently listening to this, or I'd like to in the future. (3) I'm never going to listen to this, so please never show me again.
Factoring in popularity.
I think the reason a lot of people subscribe to shows rather than hunt for episodes is because, well, most content is pretty crappy. The best signal for quality is whether or not you've liked a particular show/host before. The next best signal is probably how many other people are listening. If the #1 thing you are competing on is search quality, I'd find a way to include these factors.
Very good feedback!
Mobile: Listen Notes used to have iOS & android apps that allow to search & add to play queue & listen. The app was poor-designed. So I took it down several weeks ago to refocus on the web experience for now. I'll relaunch the app in the future. Most podcast players are podcast-centric that you have to subscribe the entire podcast to listen just one of its episodes. I still need to figure out how to streamline the process of "search on web then listen individual episodes in a podcast player app". Maybe I can provide api and podcast player developers can use to easily import & play single episodes from listen notes?
Personalization: need to use cookie or provide login experience to remember personal preference. Only ~45% visitors are from USA. People from other countries may want to search for podcasts in their language.
More signals to improve search result quality: like what you suggested in "Factoring in popularity" :)