Any "creative" ways to detect mistakes in your models before the client sees them?
Hi all. I was wondering if any of you have found clever or creative ways to find mistakes in your models before it goes out to the client? What I have been doing is checking the source of every number entered and almost all formulas to see if the calculations were done correctly. It's the safetest way but gets inpracticle when I'm time-pressed with multiple projects at hand. Any advice? Thanks!!!
First, there's no fast way to QC a model. However, the more familiar you become with a particular model, the quicker you'll be able to spot "irregularities" in the model. Mistakes happen all the time, especially when working with "version 15" of a model. Most important is to have someone else QC a model vs. you. The reason being someone else's naked eye would be more critical.
Impossible to properly check a model on screen.
Print it out and go through all the calculations by hand.
I think you were meant for audit!
it's next to impossible - sometimes you look at figures and you have to see if it looks right - that's the best guess. If you're a monkey and not understanding what you're putting in, i'd say you're more apt to make errors.
well excel has that little green thing on the right side that tells you if there are irregularities
...put in something ridiculous and see if the model still balances. If it does I just control-z out. If not I have some work to do.
It's always worth the time to "stress-test" your models. I've definitely had a model blow up in front of a client and felt like an idiot in front of a VP and CFO.
I would never go through it by hand. You are more likely to screw that up than your spreadsheet. (unless it's something VERY basic.)
that is what i do too, i put in ridiculous numbers and see how it flows. i think it works pretty well.
You guys are scaring me...
I know, shit happens and especially in banking.
The "does the shit make sense" test works wonders. If it makes sense to an anal-retentive associate, 75% of the time it makes sense to a 9 to 5 client.
Man, for an LBO model, that's ALL i use.
Disclaimer: I have no finance training, I am still in college, and this advice just comes from a model that I spent 12 weeks creating for a merger department of an engineering company last summer.
I created some sensitivity analysis into my model. Very basically, I made a second model that multiplied the inputs of the original model by a percentage. Put in zeros, and your second model should be zero (if not, you haven't linked all your inputs correctly). Put in 100%, and your two models should be identical (again, if they're not, you've lost a link somewhere because of your last change). I just ran this after version 15 or so, when it got too complicated to see if I'd lost a link or not automated an input I thought was automated.
This is just a method I came up with, so I can't say whether it's a "good" method. But I was able to debug the model really quickly, and it worked for me. I can't really explain much better than that, sorry.
I usually QC my models in two ways: 1. Stress test like everyone else mentioned (e.g. put $1 purchase price, see if everything fucks up and put $99 billion and see what happens). 2. Lots of checks that sum into a master check (using absolute for the checks). This way, your master check will be non-zero if any of the other checks are off. I set my model title to say "Error" if the master check is off so that I don't forget - I printed models a few times and realized title said Error before distributing. Avoided a few embarrassments that way in rushed situations. 3. See if #'s make sense (you can usually tell if something is really out of line).
I usually have the following checks: 1. B/S balances; 2. All B/S non-negative (apart from accum. dep.); 3. Debt schedule sources and uses balances; 4. Sources and uses in lookup section balances; 5. Sources and uses on cover page balances; 6. DCF sensitivity table balances with current case DCF; 7. Warning if average interest is not turned on
One thing I have found to be invaluable is to be able to turn avg interest off. The iterations for avg. interest are the most common way to ref your model...easy to turn off and back on and everything is ok again.
you're a tool for posting this -- for anyone who isn't familiar with DealMaven, everything here (error in title, etc.) is taken directly off of the program
not to say it doesn't work
I agree that my post was for beginners, but I have seen many models sent to us by lots of banks that do not have comprehensive checks in them.
Also, I don't think everybody knows DealMaven. I think most ppl on this forum are early stage and knowing 3 basic ways to check your model should be helpful to some.
I agree
.
Why are you guys posting about LBO models? The original poster didn't identify any one specific model.
I've just spent the past 2.5 hours reviewing an analysts' model and HE clearly did what you guys are advocating, ergo didn't check his work thoroughly and now I'M shot to shit. I'm giving this crap back to him.
Folks, don't do as these ding dongs say. Do as I TELL YOU!!!
My new thing is if I see even one thing incorrect on the first worksheet, I'm giving the entire shit back. Not worth my time to trace the problem.
Chill out..."shot to shit" at 8 o clock? Come on oh great associate.
Chill out? A mere moment doesn't freak me out. But when someone wastes my time, it's aggravating.
Aliquid dolor doloremque praesentium nam similique numquam ipsa. Maiores cum et nulla totam est beatae. Sint dignissimos nulla quasi placeat. Molestiae aut aut iusto unde voluptatem. Aliquam quis et necessitatibus neque animi quia alias. Sunt odit rerum ducimus. Qui aspernatur placeat laboriosam nemo provident omnis delectus.
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...
Ut asperiores autem mollitia rem et quidem. Qui maiores id non quaerat placeat iure. Exercitationem ex in vel magnam omnis eos ea.
Non at magni iste illo aliquid sit. Soluta odit quos enim nulla. Ullam corporis adipisci totam dolore ea non in.