Hacking Product Hunt
A data-driven approach to win at Product Hunt
The rise of Product Hunt
In less than two years, Product Hunt (PH) has become one of the best ways to spread your newborn product/startup to the world.
Why? Simple, if you have ever made a startup (I did it and I failed) or built a product you should know that the hardest part is not to have your super cool Ruby on Rails back-end working.
The hardest part is to get those bloody first 1000 customers.
PH gives you a user base of tech-addicted at no cost. A win-win situation for every player: founders, users, investors and, obviously, PH.
PH is kicking asses: this is a fact, not just an opinion. Data talks: this study from Mattermak shows that 206 startups have raised funds after being added to PH. 13 of them have raised money in less than 2 months after being added to PH.
Let’s hack it!
It seems that everyone is applying a data-driven approach to maximize sales or to give people what they really want. So why not using the same approach to get the best out of PH?
The time when you choose a random description of your product and just post it on PH when you are not too busy, it’s over. Or at least it’s over if you want to succeed.
I studied this powerful tool for you and that’s what I found out.
Analysis
The data set
- 14,549 posts
- All posts from 2013–11–24 to 2015–05–23
Top voted posts
PH users tend to be PH aficionados (like me) and to use PH website/app at least one time a day, every day. Therefore, it’s not surprising that among the top 10 posts there are version 1.0 and 2.0 of the official PH iOS app.
PH is also very focused on startups by its nature, so I’m not surprised that the most voted product is a curated directory of resources and tools for startups (it did better than Periscope and Google Inbox).
- Startup Stash - A curated directory of 400 resources & tools for startups - 2520 votes
- Periscope - Explore the world through someone else's eyes - 2109 votes
- Product Hunt for iOS - The best new products, every day, in your pocket - 2044 votes
- Tesla Powerwall - Tesla home battery - 1876 votes
- Product Hunt 2.0 for iOS - The best new products, every day, in your pocket - 1696 votes
- Be My Eyes - Lend your eyes to the blind - 1561 votes
- Inbox by Google - Next generation inbox - 1553 votes
- Lily - A drone with a camera that follows you. Throw it in the air. - 1532 votes
- Sunrise Meet - The fastest way to schedule a one-to-one meeting - 1490 votes
- Email Hunter - Find all the emails related to a domain - 1481 votes
Top commented posts
What about the most commented post? Is it about…Google? iOS? UX? No, PH again! ☺
This time I’m surprised to see that the second most commented post is Snoop Dogg’s latest album. This makes me think about how PH will evolve in the next months (I have my thoughts about it but this is not the right place to talk about it).
- The Product Hunt Effect - A book about the impact of community-driven startups - 231 comments
- BUSH - Snoop Dogg's latest album, produced by Pharrell - 190 comments
- Product Hunt for iOS - The best new products, every day, in your pocket - 175 comments
- Periscope - Explore the world through someone else's eyes - 167 comments
- Mailbox for Mac (Beta) - Fly through your email. Now on Mac - 149 comments
- Density - See how busy popular places are in real time. - 135 comments
- Inbox by Google - Next generation inbox - 130 comments
- Point - Awesome link sharing/commenting with friends. - 126 comments
- Pablo by Buffer - Create engaging images for your social media posts in 30s - 120 comments
- Startup Stash - A curated directory of 400 resources & tools for startups - 116 comments
Who are the best product hunters?
PH’s team has always been good at building a great community. I can remember when only a narrow circle of ambassador could post products. Now being a Product Hunter has become a real job.
Here you can see the hunters who hunted products that received more votes on average.
With 10+ products hunted:
With 50+ products hunted
So if you are going to post on PH in the next days you have to start following Ben Lang, Violeta Nedkova and Robleh Jama on Twitter (even if you don’t know who they are!).
When do people post on PH?
Timing is everything. You already know that if you write your super interesting philosophical essay on Facebook at 4AM you will get zero likes.
But what about the best time to post on PH?
(EDIT: timezone is PST and GMT-7)
Here is the total number of posts for weekday subdivided by hours (from 0 to 23). The image has very small details but on the top left there is hour 0:00 of Monday, on the top right there is hour 23:00 of Sunday. The highest value is for Wednesday at 15:00 (about 260 posts) and the lowest is Saturday at 3:00 (less than 5 posts).
Here is another visualization of the same data:
When is the best time to post?
Weekday
On average a product posted on Friday receives 73.3 votes (this is the min) while the same post posted on Thursday receives 82.6 votes (this is the max).
If your goal is to get to the first position, maybe you should post on Friday because you have less competition.
Assuming a correlation between the number of upvotes and the number of visits, if you want to maximize the reach maybe you should post on Thursday or Sunday. These are the days when posts receive more votes on average.
Hour
This represents how many votes, on average, a post gets when published in different daytimes.
Let’s correlate the previous graph with the total number of posts per hour.
And let’s have a look at the total number of posts with 300+ votes.
We can notice that:
- People tend to post more between 13:00–16:00 but posts posted in the morning get more votes (best time is between 7:00 and 8:00)
- Mean number of votes between 2:00–6:00 are not statistically significant because less than 100 posts have been posted for that hours
- Some great products have been posted at 2:00 or at 4:00 (don’t you have something better to do in that hour?! :-P)
- For posts with 300+ votes the best hours are 8:00, 13:00, 14:00. This can be related to the fact that people who post very good products are more careful about choosing the right timing (according to this study the best time to post on Linkedin is 7:00 to 8:00 and the best hour to get max shares on Facebook is 13:00)
Conclusion: if you want to maximize your visibility, you should post between 7:00 and 9:00 in the morning where posts receive more votes (maybe because fewer posts are posted at that hours than if compared to those posted between 13:00 and 16:00).
Weekday + hour
Is there any single combination of day and hour where posts receive more votes?
Here are the time slots where posts (posted on that time slot) receive more votes on average. Not all possible combinations of weekday-hour are plotted because for some of them there weren’t enough posts to have statistically significant means. In this graph only the slots with 200+ posts have been shown.
It can seem strange that no morning slots are present here. The conclusion one can get from this graph is:
“Wow, don’t post during the morning!”
If we look at the number of posts posted by hour, we can see that morning posts are far less than evening posts, so morning slots have been cut off by the “200+ posts” threshold.
If you don’t know what standard deviation is: when standard deviation is high, values are distant from the mean (and so mean is less meaningful).
What are the most used tagline words?
In PH we see a lot of apps, a lot of free stuff and many iPhone-related products. The list of the most used word reflects this fact.
The word pre, used 286 times, stands for pre-launch products and it’s very common in the last months.
word - #taglines
app - 802
new - 491
way - 462
create - 404
free - 393
simple - 384
ios - 374
share - 356
mobile - 353
web - 339
find - 332
time - 325
world - 324
get - 320
one - 315
best - 303
pre - 286
online - 279
friends - 272
email - 265
launch - 265
people - 261
game - 257
beautiful - 254
real - 254
make - 251
video - 245
iphone - 235
social - 234
apps - 234A bad description can cost you up to 60 votes
No-one should judge a book by its cover or by its title. But hey, everyone does it many times every day and I’ve never bought a book with an ugly cover.
Also in PH, titles and descriptions do matter.
Game, local, marketplace, social, platform, management: please don’t describe your product with those words!
This seems an advice from a VC, but it’s not. In fact those are some of the most “hated” words commonly used in the taglines.
PH users seem also not to like these words:
- Enterprise: we are startup guys, we are not interested in B2B or B2E
- Communication: we are geeks!
- Security: I said geeks, not nerds
- Addictive: no more addictive products please, we are already addicted to facebook, instagram, twitter, slack, PH
- Friendly: ???
- Wallet: the bitcoin protocol it’s too hard to understand
And here are the worst 100 words (explanation comes later):
word AvgDif occ CV
---- ---- ---- ----
bird -57.7 23 1.7
enterprise -57.0 24 0.88
reader -56.7 26 0.95
flappy -56.7 21 1.71
decisions -54.0 23 1.05
bring -53.7 24 0.92
gifts -53.2 23 1.32
communication -51.0 22 0.99
case -49.1 25 0.82
sports -47.4 45 1.14
security -46.2 31 0.94
rate -45.4 23 0.97
portable -44.5 33 0.94
baby -43.9 25 1.44
market -43.4 24 1.1
peer -42.7 22 1.16
addictive -42.6 24 0.78
recommendations -42.5 28 0.99
friendly -41.9 22 0.77
bluetooth -41.9 25 1.02
anytime -41.7 28 0.94
moments -41.5 25 1.41
bike -40.6 34 1.56
virtual -40.2 43 1.1
shopping -40.1 29 1.01
lets -40.0 31 0.76
wallet -39.8 26 0.86
event -39.2 39 1.08
health -38.9 48 1.07
feed -38.9 22 1.55
sync -38.6 28 1.38
publishing -38.6 30 1.3
game -38.6 258 1.36
helping -38.6 33 1.2
automation -38.6 28 0.91
visually -38.3 23 0.96
art -38.2 92 1.0
system -38.1 68 1.59
local -38.1 102 1.19
adventure -37.2 21 1.31
intelligent -37.1 26 1.14
3d -36.7 85 1.37
city -36.3 24 0.95
events -36.1 69 1.37
monitoring -35.7 40 1.36
reality -35.7 22 1.32
information -35.5 32 1.57
clock -35.3 22 0.77
kids -35.0 81 1.09
secure -33.7 40 1.45
emoji -33.6 39 1.02
marketplace -33.2 112 1.11
bitcoin -32.9 71 1.34
easier -32.7 31 1.45
connected -32.6 52 1.2
point -32.3 23 1.76
others -32.3 22 1.61
unique -32.2 37 1.47
anonymously -32.2 33 0.76
monthly -31.8 50 0.77
platform -31.8 198 1.54
shirt -31.7 22 0.99
fun -31.6 94 1.16
door -30.6 33 1.11
alarm -30.6 23 0.87
social -30.4 235 1.96
brain -29.9 29 1.27
tell -29.9 21 0.75
group -29.9 42 1.33
weather -29.5 42 1.13
management -28.9 118 1.23
college -28.9 22 1.3
payments -28.7 27 1.62
makes -28.6 52 1.49
solution -28.5 35 1.36
shop -28.3 42 1.03
random -28.3 22 0.96
plan -28.1 26 0.81
fashion -28.1 32 1.13
tracking -28.0 89 1.36
human -28.0 32 1.03
room -27.9 27 1.63
word -27.6 25 1.38
network -27.6 102 1.26
parking -27.6 21 1.0
space -27.5 42 0.9
drive -27.5 35 1.05
making -27.1 51 1.34
activity -27.0 33 1.55
subscription -26.9 49 1.06
based -26.5 179 1.38
change -26.4 21 0.99
presentations -26.4 27 0.85
collaborative -26.4 37 1.35
wearable -26.2 43 1.02
message -26.2 33 1.47
go -26.1 72 1.11
smartphone -26.1 56 1.49
cash -26.0 21 1.33
let -25.8 36 1.06Techniques used (this is for geeks)
A naive analysis was made to identify the most/least useful words used in the taglines.
For every word W, products were split into two groups:
- WithWord: posts with tagline containing the word W
- WithoutWord: posts with tagline not containing the word W
Then the average number of votes received by posts in the two groups was calculated:
- WithWord_AVG: average number of votes for posts in group WithWord
- WithoutWord_AVG: average number of votes for posts in group WithoutWord
Then the difference WithWord_avg - WithoutWord_avg was used to get the “strength” of a word.
Legend:
- AvgDif: WithWord_avg - WithoutWord_avg
- occ: number of posts containing the word
- CV: coefficient of variation of the votes for posts in WithWord — the lower it is the lower variability there is among the votes in the WithWord group (TIP: the lower the better)
What words should you use to describe your product?
PH users seem to love:
- Slack related products
- Material design products
- Products about UX
- Products for startup founders
- Products about prototyping
- Products with 5 in the description (yes, seems that 5 performs very good and you can jump to the next paragraph to see why)
Here you can see the list of the 100 most useful words to be used as description:
word AvgDif occ CV
---- ---- ---- ----
resources 207.1 26 1.71
generator 163.8 22 1.28
slack 141.2 53 0.87
startups 136.2 94 1.56
movies 120.0 35 1.29
material 119.9 21 0.94
startup 118.1 125 1.05
ux 111.0 23 0.68
stock 109.1 34 1.45
paid 108.2 21 1.75
founders 106.9 27 0.95
sketch 104.9 31 0.81
menu 99.3 21 1.08
everywhere 98.4 24 1.48
interface 98.2 32 1.2
collection 94.7 48 1.32
prototyping 90.3 27 0.91
busy 89.3 22 1.54
domain 88.6 30 1.56
hunt 88.3 105 0.88
schedule 87.8 26 1.72
mac 87.6 216 1.18
icons 87.5 49 0.94
programming 84.9 29 1.21
entrepreneurs 84.1 37 1.25
course 83.1 33 0.71
product 81.6 165 1.06
someone 79.7 26 2.41
templates 78.8 26 1.15
ui 78.0 41 0.95
day 76.7 113 1.83
remote 75.2 29 1.62
largest 74.5 26 1.51
note 74.4 23 1.59
pocket 73.7 43 2.46
css 71.5 38 1.15
tools 71.0 79 2.03
images 70.8 72 1.59
code 70.7 138 1.46
native 69.4 34 1.29
5 68.8 46 1.0
minutes 68.7 76 1.32
fastest 67.9 53 1.63
button 66.3 29 1.41
growth 66.2 27 0.96
generate 65.9 26 1.14
designers 65.2 82 1.34
bar 65.2 29 1.22
messenger 64.3 41 1.17
companies 63.0 55 1.34
tiny 62.7 28 1.38
gmail 61.5 42 1.44
podcast 60.9 61 0.93
explore 59.3 39 2.32
available 58.5 32 1.31
products 58.3 118 2.02
transform 57.9 27 1.63
cool 57.8 27 1.85
amazing 57.4 45 1.4
database 56.2 34 1.04
list 55.3 73 1.19
jobs 55.2 27 1.89
behind 55.1 28 1.12
interviews 55.1 21 1.04
user 55.0 58 1.26
ever 54.2 44 1.31
free 54.2 394 1.49
learn 53.8 140 1.36
care 53.6 26 1.47
press 53.4 21 2.13
display 53.2 30 1.73
write 51.8 42 1.48
tweet 51.5 23 1.83
build 51.3 169 1.44
page 51.2 65 1.41
calendar 51.2 53 1.41
old 50.7 22 1.13
every 49.3 145 2.0
html 48.9 32 1.54
websites 48.8 68 1.19
design 48.0 222 1.18
quick 47.6 35 1.47
sms 47.2 34 1.27
awesome 47.2 91 1.65
name 47.2 25 1.22
generation 46.5 35 2.15
writing 46.1 47 1.42
inside 46.0 26 1.32
7 45.8 23 1.29
without 45.3 117 1.46
best 45.0 304 1.8
inspiration 43.2 29 1.0
unlimited 42.9 34 1.31
building 42.7 51 1.68
places 41.6 46 1.49
push 41.3 35 1.25
goals 41.0 21 1.2
grow 40.1 29 1.36
desktop 40.0 38 1.16
board 39.2 34 1.42Why does the word ‘5’ work so good?
Here are the 10 top voted products with ‘5’ in the tagline:
Highbrow Expand Your Knowledge Universe in just 5 minutes a day (574 votes)Picnic Register a new website with one click. Live in 5 mins. (513 votes)8^8 Find your soulmate. Take 5 mins to answer these questions. (472 votes)Gigster Hire a contract developer in 5 minutes (432 votes)Ethanifier Build your own Ethan" app in under 5 minutes" (431 votes)Xcode for Designers Learn how to create native iOS apps in 5 days (413 votes)DeepLink Kit Add deep links to your app in 5 minutes. (368 votes)RefreshBox Create and get subscribers to your weekly 5-link-newsletter (320 votes)Sidebar 5 curated design links, every day (303 votes)Pocket Casts Powerful podcast manager for iOS & Android. New v.5 release. (257 votes)
It seems that ‘5 minutes’ is very appealing.
This is how good words look like on a graph
I hope you can enjoy the graphical visualization for words ‘slack’, ‘hunt’ and ‘5’.
Bonus: how is PH growing?
(I know, professor Rafael Irizarry from the Harvard Data Science course would not be happy with those blue lines. ☺ )
From this graph we can see that after a fast growth between May 2014 and July 2014 the number of posts per day is almost stable.
Here is the growth of total votes per day (+2 months prediction for linear regression e and polynomial regression with degrees 2, 3, 4 and 5)
What regression function will predict better the growth for the next 2 months?
Y-Combinator effect?
On Jul 17, 2014 TechCrunch announced:
Product Hunt, The Popular Tech Product Discovery Site, Is In Current Y Combinator Batch
What does it mean in terms of growth?
What happened on August 2014?? Why that fast growth?
Here’s the answer:
Product Hunt Raises $1 Million, Launches iOS App For Hunting On The Go
Posted Aug 21, 2014
So maybe here we have a “iOS app effect” (or a “1 million dollar effect”) rather than a YC-effect.
Further investigations
I’d like to answer/investigate to some of these questions/topics:
- Is there any temporal pattern that correlates the amount of votes in the first N minutes with the total votes (example: if you can get at least 10 votes in the very first hour you have a higher chance of getting more than 200 votes no matter if your product sucks)?
- Is there a real correlation between words used and votes (example: is the product description more important than the product itself)?
- Is it possible to create an accurate estimator of the number of votes (Ryan Hoover what about a Kaggle competition?)?
- I’d like to analyze taglines using bigrams, n-grams and giving a weight to words based on tf-ids or something similar
Tools used
I used the official PH API and this wrapper for python to get the data set. I don’t think I can share the data set but here it is the code I used. Tools used:
- Python
- Analysis: pandas, sklearn
- Plotting: matplotlib, pandas
- Text analysis: nltk
Want to hire me as a data analyst/scientist?
Drop me a line:
- Email: edoardo.venturini@gmail.com
- Online CV: http://edoven.github.io/
- Linkedin: https://www.linkedin.com/in/edoardoventurini
Want to be the first to read the second part of the analysis?
Follow me on twitter.
**SPECIAL THANKS**
Thanks to Nicola Greco, Alessia Anniballo, Luca Ruggeri, Giovanni Gaglione and Claudio Perlini for reading drafts of this.
