Python for Equity Reports
I've scoured a couple of threads regarding this but have seen a few answers AOTP. Basically, I have some experience with Python and NLP. I've created a few workflows which take SEC filings (10-Ks, 10-Qs, proxies and 8-Ks mainly), and output certain things such as a positivity/negativity score, highest weighted keywords which the score is based on, summary of the risks disclosures, and of course all the financial statements (which I can then run any algos on to output whatever analysis I require).
Now this provides me a good glance into a company at a far more efficient rate than my first read of any statement. I still do my personal go-through of the statement after I do this preliminary search so that I'm confident nothing is missed. I am fairly certain that after a few months of tweaking, I can automate this algorithm to generate a summary report in a set template for my use.
My question is, as an incoming analyst, should I keep this for my personal use? Should I let my MD or team know I use this or CAN potentially use this? I don't want to step on anyone's toes or give the impression that I'm already stepping beyond an already demanding and exhausting role - especially at the most junior level.
Edit: 0 responses, damn. Okay let me reframe that question - as an associate and up, how would you feel if a new analyst brought in this weird programming algorithm that was seemingly useful. I see two schools of thought currently,
1) wow this is a great value-add and potential time-saver. Props to this kid.
2) who does this kid think he is. hasn't even been on the job a few months and already thinks he can change how things are done. what even are these outputs.
Can you respond with an account/burner so I can DM you?
This is cool. Honestly, I'd keep it to myself to ease my workload up first. If an opportunity arises for improving your groups efficiency then I'd introduce it to others. You also want to make sure it's very effective and accurate before introducing it to the group. Use it in your work and see how that work is received by your associate, MD, and go from there. You have to also think strategically, this algorithm is essentially proprietary to you. Think long and hard before introducing to others for free. Not saying you should or shouldn't but it's definitely a moment to think.
How are these programs different from what factset does?
OP here - Factset is a very formal tool that provides a much richer and deeper set of solutions for what it costs. It also is simpler to use, more standardized and works with large sets of data in a way that is tried and tested.
I'm not sure how much algorithmic experience you have but my algorithms have been built on top of EDGAR and Sec.api. This means that I understand exactly how my program functions and can tweak it for exactly what I need. The greater flexibility does require a solid foundation in program writing which is why I'm not expecting people on my team to contribute to the functionality of it, rather just the outputs. For example, I can input a dictionary of keywords which I think categorize any statement as a plus or a minus, and using my program I can identify how many times said keywords appear and in what contexts (setting ranges for adjectives around) to understand what sort of outlook the filing portrays and why. Furthermore, I can input two filings for different years for the same company and exactly see how much they vary and where the real material differences lie. The financial analysis portion of my program is pretty subpar to software that exists out there already -- exactly because I don't need to create something very powerful for that, at least that is what my thoughts about that are right now. I think the real value add is that I created something that works as a scraper from a filing and have full flexibility on what to do with said scraped data. So the more I continue to learn, the more I can shape the program to output using the scraped data.
Had a similar experience with a bit of my own NLP/Python experience and automating things at an internship (obviously where stakes were significantly lower). I used it to just lower my own workload for a while and brought it up to my boss who I got rather close to. He ended up passing it along and it got integrated into the work flow, but have also heard of other cases where people got mad for automating things. Think its rather dumb but was something along the lines of superiors thinking they were taking shortcuts (even though time decreased and quality of work increased). Something about it pissed off the old-legacy finance dudes. Honestly would keep it to yourself at least while you're validating how useful it is and even when you are sure about it be careful. Further, in my case I ended up getting more work because they realized I had more capacity at that point. No bonus or pay for it either but I was an intern so everything was fixed from the get go.
Yeah, that has been my main worry. Being ill-received is not a look I am going for, especially at the start so I think I'll just stick to personally using it and trying to make my life easier. The thought of leveraging this to become top-bucket and overall just the top performer in my analyst class has crossed my mind and I think that was the only motivation for even speaking to my higher-ups about it. I suppose the next optimization I should make for is a Seamless scraper which gives me options to spend my dinner stipend to the exact cent including tax and tip. Read that in a book somewhere sounded cool....
The random projects sounds a lot like what I love doing too! In regards to the model, making things production takes a lot of work and personally regret offering it to the firm (especially without them paying me for it). I guess if your tool helps you work more efficiently than others you either are going to be top bucket (simply by outworking others) or save yourself some stress by lowering workload. Either way good to see people integrating tech into finance so that somewhere down the line it isn't so much busy work.
Keep for yourself. Did this in my old corp fin role. Took me about 30-45 minutes to tweak program for output I needed on projects. Expectation was output within 1-5 days. Sat on finished work but gave before boss expected all while playing FIFA. Being able to consistently produce error-free work faster than expectation will put you in the top bucket. If you show boss best case is its received well and say they even implement it, short memory wont give you more than couple months and one bonus of being top bucket. Worst case boss is old school and comes at you with we have a process/ you need to learn this stuff from scratch by doing it manually/ whatever and its viewed as negative.
This might be some of the dumbest advice I've ever read on WSO.
Nevermind, saw your other comment
SVP lurking on WSO to dunk on a 23 yr old advice lol. Why is it dumb?
It depends on your team honestly. I have had old school analysts that didn’t trust anything tech related and would prefer you to punch in everything even though it’s more prone to errors. I’ve had younger analysts that were all about efficiency and limiting busy work to get to more value add projects. If you have this type of boss, you can implement and try to find other projects to put your programming abilities to use. This way, it’s not a one and done type boost, but really thrusts you forward career-wise. If you go this route, brainstorm types of information that you could gather that would be useful to the teams’ analysis. Talk to associates/analysts on the team to see if this data would be beneficial to them. If so, make a list of projects and pitch it to your boss.
I had a friend who did something similar and he just kept it to himself to minimize workload and kickback. If old school boss (which he had), I would just do this.
Most MDs would not have the programming knowledge to understand what you're doing.
Honestly, we don't really care how you do things, as long as it is correctly done, and done on time.
If you're 100% certain that you can write algos that correctly automate processes, you have 2 ways IMO to improve your situation:
1) Keep it to yourself. Use the saved time to do other tasks to become the star analyst, and get staffed on the deals you want.
2) Share it with your Analyst class if you're certain your algos work as you say they do. However, if even one of them fucks something up, they can easily point fingers towards you, so I would lean towards option #1 if I were you.
Lastly, it's all well and good to try and automate processes, and save time. But remember: accuracy>>>>>>speed.
There are services/companies that do this. The one I have in mind gives a sentiment score at the report level, color codes sentiment at the sentence level, provides keywords/themes that are searchable across all documents (including equity research, SEC fillings and transcripts) and more.
Would you rather put in the effort to build your idea out or use what you have done to get the team onboard to pay for more "modern" ways of consuming research?
Ipsam porro praesentium qui voluptatem. Veniam eveniet et similique. Et aliquam ad fugit distinctio.
Ratione illo sed voluptatem recusandae ut quis. Voluptatem voluptatem doloremque ad quas iusto quasi. Consequatur veritatis saepe aut repudiandae. Id sint ut molestiae quae perspiciatis dolorem ratione. Magni est id nesciunt eius commodi quia.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...