I work for a L/S fund that makes the analysts build models from scratch. To be honest I think is massive pain in the ass and a waste of time.
I do sympathize with the wax on/wax off philosophy behind the practice but when I have an Idea that is more macro driven/thematic it is just so much easier to pull financials from a SS model and flex assumptions as needed.
Anyway, this is just me venting but would appreciate any tips on speeding up the process (I already use bamsec to pull tables and fillings.)
For the veterans out there, do you think this is a necessary part of idea analysis and is critical in helping you understand businesses or is it just poor time allocation?
Which SS models do u often use? Do u need judgment in picking them? What if your asset class isn't traditional and the SS doesnt cover them?
Most SS models I come across aren't really that great or easily modified to new drivers, and there are basic assumptions driving most of the model. Do you just create a new revenue build (and/or cost build) and feed that into the rest of the model?
How can you have a cost build in public’s given such limited data and most of the time you don’t have more data than line item opex (eg SM, GA, RD)
In my experience, you do it 2 ways. You either talk to management and get as close as possible to understanding the inputs (they don't give you the exact details and numbers, but ballpark on fixed/variable, other helpful metrics or inputs, etc. to develop some understanding of unit economics). The other way is talking to industry insiders and building your own crude method and try to approximate it (which you can check vs, historicals) so you get as close as possible to forecasting unit economics in a way that makes sense. Precision is not the end goal, but you want to have some framework to understand the biz. What is that quote "better to be directionally correct than precisely wrong"
Often times it depends on your strategy as well. If it is critical to the thesis, then you gotta work hard to approximate it. If it is not, well...
The reason people create their own models is because, in their mind, they need to have a model revolve around something which is most likely revenue segment. SS reports vary based on what the senior analyst whats to focus on.
I use SS models just to create an outline for myself and then add anything else I see fit. To pull previous inputs I inputted a Bloomberg code which does that for me.
Hope this helps
Fully agree with you OP. Do you think that it is possible to build a BBG template that automatically extracts financials and balances it after inputting assumptions?
I've done both and they each have their place. When I was at a pod with intense focus on the quarter, the build by scratch model proved far superior. Modeling out every item I can to the greatest detail possible (SG&A - fixed vs variable) helped me see my revenue estimate percolate down to EPS, FCF - whether a beat/miss would be interpreted as high-quality. This PM was very modeling-intense and also had us put in quarterly financials for past 5 yrs, along with putting in excel comments on GM, topline drivers, etc. Just putting in these comments helped me to recognize patterns.
Before the pod, I was focused on more longer-term theses so less need to be perfect on every quarter. I usually did SS models but built from scratch for larger positions.
I agree, in a pod role it was very much worth it to build robust and detailed ground-up models of my coverage. I’m talking hand type in all the history, examine the restatements and resegmentations and disclosures in the text. In my sectors, Bloomberg pulls were unreliable and sell side models did not drill into segments or rev/cost builds that matter.
Now in a traditional model, it depends - it’s still better to build your own, but I balance it against 1) how big is the position going to be 2) does the variant view need a detailed model to express it? I.E. if A) if I think the end market is rolling over in a subsector and have a basket of names, lightly modified sell side models might be ok. If my variant view is more focused how certain segments are doing or how the unit economics/op-leverage is misunderstood, I would build my own model to express.
Adjusted numbers have become such BS these days, and sell-side parrots the headline figures - you’re probably better off at least using Canalyst or building your own model and then filling in historicals from sell side.
Above a certain size position u should always build the model imo
This. Definitely need to know how to build your own model and have the ability to play 3M and be able to tweak an existing model instead of sometimes reinventing the wheel ("We don't make things, we just make them better" if anyone remembers), but being able to create a one-off, to show-off, a unique scenario you see at least in abstract and can put pen to paper for.
I don't trust the BBG numbers and sell-side parrots either. I wouldn't even trust the IR folks' investor campaign figures either. May sound corny, but I'd suggest pulling the XBRL from the SEC filings for more resilient figures. Or if you somehow have access to the board books the c-suite/board of directors have since they're under SOX/Audit scrutiny.
Dead on about the sizing argument too. Regardless if it's a one-time block transaction or ongoing DCA.
It clearly is a waste of time - you should just be getting the models from sell side and then inputting your own assumptions. Pareto effect clearly at play here - 80% of the results are from 20% of the actions. Think about all the time lost from formatting, getting model to balance, etc.
My group has always built models from scratch, even though we are very macro heavy. We follow a "core plus" approach, which means always build out the last 2 years with quarters, plus LTM. So always two full years of historicals. The "plus" refers to using Bloomberg, visible alpha, or other sources to pull time series data for reference case and distributional analysis.
"Full" means integrated 3-statement model, d&a/capex model, corporate action model (public version of a debt sweep), share repurchase & dividend model, and then KPIs measured by the street and peers. We make our own custom valuation and earnings dashboard too, but that's a template that gets thrown in. It's usually a ~5 hour process to build the model from scratch (excluding the active parts).
I recommend using alphasense's look up similar table feature to grab all of the financials "in one go". Vastly superior to exporting tables into excel, and can use this for any financial line item (including capex, d&a, and such).
I don't know if I suck at modeling but 5 hours seems fast. Unless you just mean historicals and no detailed build on the projections. I am curious on time saving processes when building models and pulling in historicals - I have never had fancy tools so it has always been looking at 10ks and 10qs on my left screen and typing it in manually on my right. Sounds like you can copy and paste a few quarters of the full financial statements and then just modify it with your own equations in excel?
Second question is on your point re" "time series data for reference case and distributional analysis" - what does this mean? Is this just so you can look at revenue and margins over 10+ years and see the cycles?
Alphasense has a tool that lets you select a table on a filing and then see all historical examples of that table. I don’t export it— I manually input the data, but it’s very useful.
5 hours for the historicals and build, excluding the actual projection. All the formulas are set at this point, the growth rate is just a dummy variable.
If you’re doing it manually I assume you know the “one 10-k for two years, four 10-q’s for 8 quarters” rule.
Before we build a model on a company we run through a checklist to review the financials so we can design the builds.
Iure quae quisquam molestiae voluptatibus ad eum officia dolores. Quam incidunt et rerum a dolore est. Ut quidem ut impedit distinctio sit in. At corporis animi harum. Quia neque iste eveniet dolorum nostrum reiciendis.
Vero in at et est amet. Quia doloremque ipsum sed qui sed. Nobis quia quam non animi dolorem.
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...