Which source of Company Financial Data should I use for my model?

If we want to create a model of a company's future performance, we need a starting point, a set of inputs which we can examine, adjust and extrapolate our best estimate of how we anticipate the company will perform in the future.

So what are our choices?

1. Lifting values straight from the Financial Reports filed by a company
2. Data Vendor - Bloomberg, Thomson Reuters, Capital IQ, FactSet etc
3. XBRL

I believe XBRL provides the best possible place to start. Why?

Well, as my old Physics teacher was never short of telling us, lets go back to first principles.

Our ideal starting point would be a perfectly accurate picture of how a company is performing right now. We can’t get this however for two reasons:

1. We don’t have real time access to companies accounting systems so we are constrained behind the curve by the reporting calendar.
2. We can only see the published external numbers, the numbers that the company allows us to see, subject of course to any legal disclosure requirements or the opinion of their auditors.

So even this, our best source, is inherently flawed, so we must be ready at all times to make adjustments/correct the figures put in front of us. Despite these caveats, the company still has to be our first port of call because only they have the closest and best view of the current operating performance.

So why would we use a data vendor?

Well their starting point is exactly the same but what they do is prepare the accounts for financial analysis. If you start from the Financial Reports filed by a company, preparing every single company this way is expensive, which is why they will charge you thousands of dollars for the privilege. Criticisms levelled at data vendors in the past have been that they don’t always get the figures right and that its not always clear how and what adjustments have been made.

Also a standardised approach can be problematic as the Corporate Finance Institute note:

“Companies such as Bloomberg, Capital IQ, and Thompson Reuters provide powerful databases of financial data. However, financial statements retrieved from these databases tend to be in a standardized format. Thus, if the company uses an accounting value unique to its business operations, you will not grasp it from data retrieved and it will affect your analysis.”

This is why in critical decision making, when an investment decision or M&A deal could be worth millions of dollars, vendor data would never be used as a starting point for a single entity centric model. Of course this data has value in screening and modelling whole markets or sectors but even here, for the reasons stated, they are potentially flawed inputs.

So what does XBRL bring to the party?

XBRL takes the effort out of lifting the financial values from the report and provides a first pass at standardization. Not normalization mind, but a first step in that process if your goal is vertical analysis against a company’s peers. As I discuss in this article, unfettered XBRL, as filed with the SEC and made available through Edgar or alternatively for a fee via the XBRL.US API, does not, nor indeed intend to, provide a perfect set of standardized values.

By harnessing the XBRL tagging, your models can be automatically derived from genuinely as reported values, the closest view of past operating performance direct from the company and untampered by any third parties. The hard graft of lifting values, monotonous, error prone and time consuming, is removed. But as I underlined above, this is not quite the finishing point. Most of the hard work is done but we must always be ready and prepared to make a few adjustments*. They will undoubtedly be required.

*How you can easily make these adjustments is discussed here and in the following video. If you want to read more about the need for adjustments in XBRL, then check out the next post in this series.

You can also read about totaliZd from Fundamental X here, our complete solution for preparing XBRL derived inputs to financial models in Excel.