How important is modelling to investing in reality
Is modelling really important? I always feel like 1) the numbers you read in filings might be not accurate in the first place 2) using those numbers to make overcomplicated assumptions in models is unnecessary 3) people generally are too obsessed with modelling every single line item and forget about the big picture.
Any hedge funds PMs can shed some light on how important is modelling and how simple / complicated a model it has to be? How do you think about what line items / numbers to care and more importantly NOT to care about?
Not super important. Rip a sell side model and overlay your view
Not important at all. But helps you think about what drivers move the needle. Can be fucking stupid if all people care about is yr on yr sales growth number. Might as well just project price and volume and call it a day lol
I find comps are more useful than DCF most of the time.
But I do study sell side DCF closely to see identifiable drivers of their models, I just don't take their target number/DCF valuation seriously or fussed about each line item.
Then I incorporate some useful elements into my own framework.
Depends on the process. I’ve worked for a PM where we primarily used sellside and overlaid our assumptions.
I’ve worked for another where every model was built from scratch and every driver was modeled.
Latter was at a MM so highly intense focus on the quarter and focus on where we thought the company would beat/miss sellside and where we thought guidance might be changed to after the quarter.
I disagree - I think it is important. Not being more complicated, but to the extent better modeling of key drivers or “what matters” allow you to be more accurate on prints, that is very important in MM funds and important on the short side in traditional SM funds. Timing and sizing is a source of alpha, just like hit rate and slugging pct. Like everything else, the real answer is “it depends” when it matters, it doesn’t. When it doesn’t matter it’s a waste of time. Variant views and alpha come in all shapes and sizes re: time horizon and ability to predict not just outcome but also ‘stock path’ can matter. Lots of ways to make money - sometimes a better model can be a way.
I think modelling is important in the overall process but there is a big difference between a 'Spreadsheet Buy' where the DCF says there is 50% upside and a strong thesis that leads the way to 50% being realised by identifiable (consensus upgrades/downgrades, guidance raises, earnings enhancing M&A).
I find a model helps understand the downside risks in a better way - i.e. my bullish/bearish assumptions are wrong and how much do I stand to lose on my position.
I think the alpha being in the model is reducing though - there are just so many teams at Citadel/Millennium/Baly/P72 looking for inefficiencies on a sector by sector basis through attending every conference, buying every dataset and modelling every Q with every bit of information available.
Removed for anonymity.
I'm still in college but I feel like the usefulness of the model is doing it backwards--playing with assumptions to back into current valuation and get a feel for investor expectations.
"1) the numbers you read in filings might be not accurate in the first place"
If this is true, this is literally fraud
Every valuation comes down to the same key variables: growth, reinvestment (capex/NWC), margins, and cost of capital. Companies have different sensitivities to each depending on their business model. For me, I have to model the company out to figure out a few things: which of these variables matters the most, what expectations are baked in to the most important variables, and where can I take a variant view? I generally will figure out which variables matter by building the model, looking for LT trends, getting a feel for what is “normal”, and running the sensitivity analysis. Once I know what really matters, I can build out the different scenarios to frame the range of outcomes. I also want to make sure that my model and/or consensus fits within the context of the company’s historical economic reality. For that I like to break down ROIC into IC turns and NOPAT margin. If consensus implies IC turns are going to inflect because their capex forecast is too low relative to their topline forecast, that might be an area where I can find disagreement and have a variant view. This is also generally where I will challenge the models of my colleagues. If your expectations imply an inflection point in the company’s economics, you have to have a really strong thesis pertaining to why that is realistic/of a high probability IMO. I wouldn’t be able to understand any of this if I didn’t build the model.
Now another thing I have learned from building detailed models and having my team start to use my model template is that a good framework in the wrong hands is worse than if the analyst hadn’t built a model at all. It is true that you can make a model say anything you want. You have to have true intellectual honesty for modeling to be of any value. I see so much bias in most models that often it would have been better not to go through the exercise at all.
The key variables are essentially bundled into multiples, and because of this if you defer to multiples or something like comps it essentially creates economic guardrails. I always check my models against the company’s historical multiples and see what my numbers imply as a sanity check for this reason. If my output implies something outside of the range of historical multiples I need to have a strong basis for such an output that ties into my overall thesis on the company.
I’ve also seen that most analysts just build really poorly constructed models. Here again it would be better to use multiples/comps than to build a shitty model.
So to answer the original question, is modelling really important? Yes, if you do it correctly, no if you do it incorrectly. How detailed should the model be to have value? I like this quote for answering that one: “Make everything as simple as possible, but not simpler” (Google says it is an Einstein quote). You really just need to be able to understand what is priced in and where you can take a variant view with confidence. A good model can help you do that, a bad model will result in bad conclusions and you’d be better off not building a model at all.
Eaque voluptatem sed saepe sed. Esse doloremque praesentium fugit necessitatibus cum doloribus. Omnis quibusdam sit accusamus reprehenderit modi impedit rerum. Deserunt expedita repellat voluptas quia magni dolor quae.
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...