Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: Crowdsourced salary prediction (salaryfairy.com)
36 points by cagriaksay on March 22, 2014 | hide | past | favorite | 57 comments



You need to offer a way for me to turn off my account. I logged in just to try it, found out it's not what I thought it was and am not that interested, and now want my account deleted and your access to my LinkedIn revoked.

Showing some peoples' profiles anonymously is not very anonymous at all. If you're the owner of a one man company then not showing your name is meaningless. You need to anonymize the actual companies too.

Don't get me wrong the site looks nice and you have done good work. But I want out, and you should offer a way out on day one.


We will offer self account deletion shortly. Meanwhile, you can send us an email to request removal.

We will also introduce new ways to anonomize your company and education info. Thank you very much for your feedback.


I wrote to info@salaryfairy.com and they removed me within 2 minutes.


This comment precisely reflects my feelings. One of my company names actually has my own name in it.


You can now delete your account and all user data from settings page.


It would be nice to see a sample so that I know what I'm getting into before I give you access to my LinkedIn.


Same thing I was thinking. Not too keen on providing all of my information without some sort of example(s).


Thank you both! You are definitely right, it is now the first thing on our list.


I would also like to know that this isn't going to share the results to my profile.


We won't, but we will make that promise explicitly on the homepage. Thanks for your feedback.


We have included screenshots of user profile, prediction, and score pages in our homepage. Thanks again for the feedback.


It's a bit difficult to tell what the experience will be before you jump in, and it's not very fun when you do. Screenshots of the experience would help a lot.

Also: After seeing it, it's not going to work very well for HMF-level people since most people don't know what an HMF looks like.

It might be better to randomly swap extracted features them around with other people so that you can tell how you're measuring each feature. You might find some features deserve drastically higher weights to correct for people's subjective "feelings" about how much someone should make.

I like that it tells me how other people guess (after I submit my own guess), but it only tells me the mean (and it says average). When it had collected guesses for my salary, I watched for a few guesses to pile in and it seems to be taking the mean of them. I don't think this is a good idea, for example someone guessing very low repeatedly will drag down everyone's score.

It might be useful to collect "what do I think this person makes" separately from "what do I think this person could be making".


We added screenshots of the experience to the homepage. We are working on weighting algorithms. Thank you very much for the feedback.


What's an HMF?


Head Mother Fucker.


How does SalaryFairy know what's correct? Just in general?

Someone with a statistics background might have a better idea, but I'm thinking might help to allow users to, in their profile (and have it be optional), enter their salary.

That way, you'll not only be able to show people what other people think they should be making, but you might be able to (once a certain threshold is met) show what people with similar experiences in a certain region are making. And you'd be able to see how close the guesses are to the real thing and maybe figure out a margin of error.

It might also be nice to allow people to put in descriptions of what they did at their workplaces. Someone doing embedded development might get a significantly higher salary than someone doing .NET, even if they have a similar # of years of experience.


Interesting, but I found it incredibly hard to come up with sensible estimates for profiles.

So much depends on their location -- do I assume they'll be staying where they are and try to estimate what salaries there should be like, or estimate based on my own location? -- and the sort of places they might like to work. If you're currently in a university or a small business you might be paid less than at a bank for the same work, but you might be choosing to work there for other reasons, so it's not that useful to point out that another employer would pay more.

I found my estimates were generally much lower than the averages from other users, perhaps because I currently work in a university myself.

Edit: and indeed, it turns out that the estimates I'm seeing for my own profile are about double my actual pay. Which isn't surprising, because that's about what I might expect in the private sector in the same location.


So far the prediction average of 2 people is less than half of what I currently make (and less than half for the average salary for my role in that city)

You need a LOT of people to make this work... so far it seems like a nice idea but not sure it can work relying only 100% on "crowd-smarts"


Agreed, thus far the predictions seem very low to me. With the exception of the occasional non-profit professional profiles, where the predictions are much higher than what my friends who work in the non-profit sector report.


My prediction average of 3 people is more than twice what I currently make.

The estimate I made for someone was 70kish before I went on to see my similar one. I wonder if that's just the safe number lots of guessers are settling on?


There's a major bias to guessing in the 70-90k range. People likely to earn far less or far more than that are being pushed toward the middle.


This is pretty cool, but I do have one question:

How do you account for the fact that many of your users may not be familiar with the salary ranges in a specific area?

I understand the goal is to get a large amount of estimates for each user, to compensate for that, but my guess is that locations where larger amounts of users are located would have far better estimates. I just saw a couple of Vancouver (where I live) people, in positions I have trained and hired people in, but the estimated salary seems to be quite off considering that Vancouver salaries are noticeably lower than San Fran in reality.

Edit: It seems most of the profiles for me are Vancouver/Toronto/Ottawa, so I'm guessing that's their way of getting locals to estimate salaries.


Thanks for your feedback. You are indeed seeing profiles only from Canada. You can change your settings to predict salaries from other countries. We will introduce filtering based on specific area in the future.


It'd be cool if you also showed the currency converted for where the person is located, maybe in brackets next to the dollar amount. I'm in the UK and I'm getting asked to predict in dollars the salary for someone else who is also in the UK.


Converted local currencies are now displayed besides $ amount. Thanks again for the feedback.


Excellent feedback! This is on our list now.


I like the concept. Interesting how it's a game and provides salary analytics at the same time. I just added my profile but I will check again after a few people have predicted my salary to see the results.

I was bouncing around a few ideas that involved using the Linkedin API a few months ago. I have a few questions about Salary Fairy that may help me with my research...

Does Salary Fairy download a user's profile to the database? If so, I thought Linkedin required the user to grant permission to store the data in addition to just obtaining access when logging in. Is this still the case?

What technology stack are you using?

What type of hosting?


Thanks! We added a privacy clause stating that we store LinkenIn data. We are using python/django on a linux server from DigitalOcean.


Thanks for the response.

I logged in today to find 3 predictions that were double my actual salary. Maybe if you include an option to add your real salary, the salary predictor bar only allows guesses within a specific range outside of the real salary. It might keep the predictions more realistic.

Also, I received an email confirmation from SalaryFairy. What is the purpose of this if I already logged in with Linkedin? Wouldn't my account already be verified if I have a Linkedin account?


Our site provides personalized salary predictions with wisdom of the crowds. We recently built a scoring feature where you get higher scores for good salary predictions. We would like to get your feedback.


Good luck guys. Will be really interesting to see how this aligns with actually salaries in the marketplace. Might be worth allowing users to give you their actual salary in the interest of tuning your understanding/algorithms.

i.e. The crowd thinks you should earn X. We think you should earn Y based on similar people. You currently earn Z.


Thank you very much, this is a great idea! We will definitely add a feature to opt-in to give your actual salary.


Agreed! My predicted salary from 10 predictions is about $24K lower than actual salary, and I'm happy to provide my salary information in-app.


Just connected my LinkedIn profile; how are you predicting salaries? Sharing my public profile with others and aggregating their salary estimates based on weighted criteria? I assume those with the closest estimates to actual salary have their estimates rated higher than others (based on your leaderboard here: http://salaryfairy.com/fairy/score)


Yes. We display anonymous profiles to our users and aggregate their predictions. Predictions based on weighted criteria is coming soon.


Excellent work. Something to consider would be to constrain salary estimates to the same industry based on LinkedIn profile endorsements my profile contains.

For example, I'm an IT professional. When I clicked "Predict salaries", a profile for someone in the financial industry was displayed. I have no idea what they should be making.


Thanks :) Scoring is one of the criteria we are looking into. We need to be careful not to bias predictions too much towards high scoring predictors. We are now filtering profiles by country. Once we get to a healthy user base, we will introduce more filtering options.


No problem! I've thought many times how to solve this problem, but all of the data is opaque :( Wish you guys well!


Salary is a weird number. For instance giving a startup founder who is an ex-Google engineer a prediction is awkward because she could maybe fetch a mid six figure salary if she wanted but she is probably is paying herself enough to get by or even nothing depending on the stage of a startup.

It's unclear whether I should be answering what salary does this person's skills merit? Or what salary do I think they are earning today?

I also wish I could refresh and/or edit my profile on salaryfairy.


It would be cool if you got recruiters on board, so they can see the crowd estimated salary of an employee - so they can see what people who actually know about technology think someone is worth.

I don't really like recruiters, in my experience they tell you you're aiming too high, but don't know anything about the technologies.

So my feedback is to try and aim to get recruiters and employers to use this to find new candidates.


Agreed! This is actually one of our goals: to create an opt-in program for our members to meet serious recruiters based on their salary prediction. We hope to give our members a fair starting point when talking to a recruiter. Thank you very much for the feedback.


It would be really interesting if profiles were not filtered by country but predictions were grouped by it. If I wanted to move abroad I could base my salary expectations on it.

This would also give more value to people from outside of the U.S. (there is only one other person from Singapore registered, so as it is I won't get many predictions).


Thank you for your feedback. We will work on features for members who are curious about other locations.


How is this different than Glassdoor.com?


Glassdoor gives salary ranges based on title, company, etc. Our aim is to provide personalized salary predictions based on your education, experience and skillset.


Apart from salary, is any indication provided of the volume of matching jobs/locations?


This is in our future plans.


It might be useful to show your estimated average salary, and your estimated average salary from people in your area. Most people probably don't have a good idea of what the going rates are outside of their geographic location.


They should offer A/B testing on your profile to maximize you estimated salary.


I think the best way for this to work is for you to pair it with a more professional algorithm (e.g. Payscale, Salary.com). I am not sure I would ever trust what a bunch of random unqualified people say my salary should be.


Some quick UI ideas http://i.imgur.com/7PDTIt9.jpg

Maybe also keep the scroller for when you absolutely have to have a higher resolution...


thank you very much! it is definitely a great idea to show time spent at a job more clearly. also we will work on the scroller as you suggested.


Why don't you take a guess at mine? http://salaryfairy.com/fairy/guess/420


This is a very interesting idea - so far with 7 predictions, the predicted # is more than $30k than what I currently make.

Goes to show how underpaid people can be.


I find it extraordinary how many people are willing to give full access to their LinkedIn account and get so little in exchange.


right off the bat, I'm not so comfortable about giving the site access to my LinkedIn account. No promise that it wont spam my account somehow or post things to my Linkedin feed.


We don't request the necessary permission from LinkedIn to send a message on behalf of you. We have no intention to spam, but we will put that promise explicitly on the homepage. Thanks for the feedback.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: