At Mobius Venture Capital, as in many venture firms, we don’t have
analysts (people whose primary responsibility is to run models, cap tables
and the like). As a result, each of us
does most of our own financial modeling. I actually like this set-up, because it makes
sure that I’m both directly responsible for my work and am up to speed on the
financials of each of the companies I work with. Reviewing financial models is not the largest part
of my job, but is an important part of what I do – for screening new
investments; tracking portfolio company performance as well as analyzing
follow-on investments into companies in which we already have a financial
interest.
In the course of reviewing
many many many such models, something rather counter-intuitive has struck
me: most
financial models are too detailed. That’s
right – most models have too much information in them; too many assumptions;
too many inputs; and are too hard to follow. Now, don’t get me wrong – there is definitely
a place and time for a detailed line item budget (say for a rolling 12 month
operating plan). That said, trying to
detail out line item projections over a 5 year period I think makes models less
rather than more useful.
When I was in college I
really enjoyed theoretical economics. One
of the classes that I liked the most was Econometrics. As a relatively green econ student, I remember
that my inclination (and that of my classmates) when building econometric
models was to put in as much data as possible – the theory being that more data
wouldn’t harm the model and would potentially help it. Our professor, Gary Krueger, pounded into us that
this was in fact not the case – weak data hurt your model and taking out
mediocre variables actually strengthened the veracity of the output (the
garbage in/garbage out theory – although he had more colorful way of describing
it at the time).
I think a lot of modelers
fall into a similar trap as me and my classmates first did – instead of
simplifying their business to a reasonable and manageable number of inputs and
variables, they attempt to put every complexity of their company into the
model.
In mathematical terms here’s
what I’m referring to:
Take one variable V that you have 80% confidence in.
Break that variable into 3 sub-variables – A, B, C – each
of which you have 90% confidence in.
Since your confidence in your original variable (V = 80%)
is greater than the product of the three sub variables (A*B*C = 73%) you are
actually better off sticking to the simpler variable even though you have less
confidence in it than in the sub variables individually.
There’s a balance here that
is important to strive for because financial models need to be sufficiently
detailed as to accurately reflect the business, be able to run realistic sensitivity
analysis on, etc. However if you end up
with a 10MB model for your start-up (and I’ve seen these), you’ve probably gone
too far.
Here are a couple of specific
thoughts:
- Before you start
modeling list out the key drivers of your business – really distill what the
key assumptions are and make sure you call these out in your model
- Add detail where
it really helps – a lag from bookings to revenue reflect what is really going
on in your business – that’s good; deciding on an employee by employee basis
what various raises are to be in year 4 doesn’t add much (simplify this
assumption)
- Break out your
assumptions – be explicit about the drivers of the business and group them
together (perhaps at the top of each page) so that a reviewer can easily see
what each of the drivers are
- Don’t hide assumptions
within formulas – formulas should be driven off of numbers that are exposed,
not contained within the formula cell
- Be clear (by
color coding or some other mechanism) what cells are assumptions (i.e., you
input a specific number) vs. derived from other cells
- Don’t be afraid
to make general assumptions where the detail doesn’t really add value to your
model – for instance on T&E load for employees (I’ve seen many models with
3 tabs to try to calculate things such as year 6 cell phone use per employee)
I could keep going, but I
think you get the picture. Perhaps more
important than anything else, don’t forget to step back from the model when you’re
done and look at the macro trends that you are predicting. Does the revenue ramp make sense? Do the revenue and expense totals per
employee seem reasonable and do they grow (or shrink) logically? Are variables such as your days receivables
and payables in the ball-park? Are your
working capital assumptions generally reasonable?
More variables and
assumptions are perhaps not the key to better modeling – smarter and more well thought out ones are.
sjl
Recent Comments