As the ongoing evolution of allegations against NQ Mobile (NQ) leaps from one theory to another, one particular theory that has caught the attention of this reader is the belief that NQ is round-tripping its revenues through its acquisitions. Fellow SA contributor Robert Zangrilli has kindly illustrated this theory in his article: Research Report: What Baseball Can Teach Us About NQ Mobile. The rapid display of support and agreement with Mr. Zangrilli's analysis had this author wondering just how detail oriented the short position is if approval of this fairly analytical and quantitative analysis can be done so quickly.
Please note the time stamp: Dec 11, 7:44AM. I'm assuming Muddy Waters Research (MW) was tweeting from California at 7:44AM giving MW very little time to read, assess and profess their support of Mr. Zangrilli's analysis which was published on Dec 11, 9:30AM EST.
Given the serious claims and resulting repercussions to not only the Company but to public investors, I decided to kick the tires of this round-tripping theory to see how well it stood up to some basic scrutiny.
A Brief Overview of NQ's alleged Round-Tripping strategy
The general theory of how NQ round-trips revenues and perpetuates fraud is below:
- NQ acquires fictitious companies using stock, with the sole purpose of monetizing the stock by having the target owners (who are colluding with NQ) to pledge the stock to lenders as collateral for cash
- Cash raised in this method is brought back into NQ via fictitious sales through Asiainfo Linkage (ASIA) the billing software provider for all three major mobile carriers in China and for many carriers in South East Asia and who's former CEO/CTO and current Chairman, James Ding, sits on NQ's Board of Directors
- Deception of this nature is evidenced by the company's preference for using stock in acquisitions and minimal capital expenditures to support revenues and growth
- Despite the large cash hoards, the company continues to raise cash as evidenced via its convertible offering and private placement to Atlantis Investment Management, with the intent to further perpetuate fraud
- The VIE (Variable Interest Entity) structure in China facilitates NQ's alleged fraud by allowing NQ to retain cash in China while denying investors the right to those assets should the fraud be discovered
So let's dig into the round-tripping theory and see just how detail oriented the short thesis is these days.
Claim: Relative to its peers, it's impossible for NQ to generate its levels of revenues due to NQ's limited investments in computer equipment per headcount.
As noted by Mr. Zangrilli NQ Mobile appears to generate a fairly high ROI given its limited capex spend on computers when compared to the universe of 75 companies he profiled in the software and application sector. As a general rule of thumb when pulling together a comparable compset, a good compset is narrowed down to those companies that are as relevant as possible to the target asset (NQ being the target). Profiling 75 companies is what we in the finance world call a data dump. Lots of data and good for macro trends, not terribly helpful in terms of analyzing nuances relevant to a specific company and industry. Narrowing your focus eliminates noise from unrelated assets. Even when data is normalized on a same scale basis (i.e. as a percentage or ratio) it's difficult to compare a Netflix or Amazon to NQ Mobile for a multitude of reasons (subsector, stage of growth, size/scale, location, go to market strategy, competitive advantages/weaknesses, prodigious use of operating leases vs capital leases, etc.). This is specifically the reason why a comp set should be robust enough to reflect an industry, but small enough to retain focus on a particular asset categorization.
In the case of NQ Mobile, there are few good public comparables to benchmark against a company like NQ (china/global focus, high rate of growth, security focus, mobile technology, highly leverageable platform), which make Mr. Zangrilli's compset particularly problematic to use. As Mr. Zangrilli previously notes, NQ had computer equipment per person of around $3K vs. his universe of software application firms in the +$5-10k range. When digging into his numbers, it's interesting to note how varied computer equipment is disclosed in annual filings. For some companies like NQ, computer equipment is just that. For other companies, computer equipment includes furniture, software assets or any multitude of related capital assets. Mr. Zangrilli's data seems to lump all capital assets with any description that includes computer equipment as generalized computer equipment. This is what I would consider mixing apples and oranges on a macro scale. Neglecting to adjust for these accounting variances can significantly skew results, which I believe to be the case when profiling 75 companies as one comparable set. While I would love to be able to adjust out software and furniture related investments in the comparable set, that level of detail is just not available in public filings.
NQ's Computer Equipment Disclosure
YY, Inc. (YY) Computer, Servers, Equipment and Software Disclosures
For those of you who think NQ spent too much on its domain name, please pay particular attention to YY's domain name investment in 2010. 12M RMB = $1.8M USD at 6.63 RMB / USD. As a point of further clarification, Mr. Zangrilli surmised, incorrectly, that NQ purchased the domain name rights to kk.com for $2.5M. Its probably worth nothing that in Japan, KK is abbreviated version of Corporation or Incorporated.
Qihoo 360 Technology's (QIHU) Computer & Software Asset Disclosures(click to enlarge)
Phoenix New Media LTD (FENG) Computer, Equipment & Furniture Disclosures(click to enlarge)(click to enlarge)
To adjust for this, we can include NQ's software and electronic equipment with NQ's computer equipment so that all metrics are on a relatively even basis. When adjusted, NQ's per person investments in computers & application software is $4.5k and $8.3k per person in 2011 and 2012 respectively. In the table below, you can see how NQ compares with a few companies I performed similar adjustments for. Not the best set of comparables, but the comps reflect Chinese internet/mobile application companies at similar stages of size and development. I only adjusted for a few companies as you can see fairly quickly, that NQ is in fact, not an outlier.
(click to enlarge)
Note: FENG expenditures includes furniture in 2011
When you consider that >20% of NQ's security products are carrier deals and often pre-installed with customer support at the carrier level, NQ's capital spend makes even more sense.
Per NQ's 20-F statement:
"Wireless carriers and service providers. The Group, via SPs, cooperates with wireless carriers to provide consumer mobile security services to the customers. In China, SPs have the exclusive licenses to contract with wireless carriers in offering consumer mobile security services to the end users and they are mainly responsible for assisting in the billing of consumer mobile security services. Wireless carriers are mainly responsible for billing, collection and customer support relating to the end users."
Now that we've established that NQ is likely spending enough on computer equipment and software, what should we think about the revenue per employee and growth in revenue per employee metric so enjoyed by short analysts? Personally, I think they're creative metrics, but fairly useless when amalgamating 75 companies in disparate technology segments for the following reasons:
- Significant noise created when comparing high revenue / low margin business to low revenue / high margin businesses
- Difficulty normalizing for high growth companies and or companies with scale advantages and or companies growing by acquisitions. Growing revenues off a small denominator (such as NQ and one of the smallest companies in Mr. Zangrilli's compset) will result in outsized revenue multiple gains relative to larger established companies
- High revenue variability due to rapid product maturation cycles in technology and boom bust cycles
A more appropriate metric to use would be gross profit or operating profit per employee as that eliminates some of noise associated with high revenue, low margin businesses. However then you have issues with other expense line items, with stock based compensation in particular, much larger (on a relative basis) in small, fast growth companies than large companies. Narrowing your universe even further would adjust more of that noise, but still result in a fun but not terribly useful metric.
NQ apparently has the infrastructure to support its operating plans and that metrics are only as useful as the data used to generate them.
Claim: NQ uses stock and cash to acquire fake companies. The issued stock is then used as collateral to borrow cash from lenders for round-tripping purposes.
Pledging common stock as collateral for lenders is fairly common and not difficult to do. Many lenders will view publicly traded securities as highly liquid and adequate as collateral to borrow against. In the case of the short thesis, restricted shares and shares to be earned via earn-out appear to be fair game in terms of collateral to borrow against. Under the right situation, a lender might be persuaded to lend against restricted shares. However, lending against un-earned shares, shares not technically owned, would be a hard sell to any lender. The recent devaluation in NQ's share price further complicates things with any potential borrower likely struggling to maintain a minimum loan to value ratio or other covenants to maintain good standing with a lender. Additionally, if NQ is round-tripping by issuing shares, why make earning performance shares so difficult? From the latest 20-f statement regarding the shares issued for NationSky "The remaining 1,150,000 restricted shares are subject to both performance and service condition…As of December 31, 2012, we determine that it is not probable that the performance conditions will be achieved. As a result, the related compensation expense was not recognized."
As it stands, using stock to round-trip is a terribly inefficient and costly way for the company to defraud investors.
- The need for collusion at multiple operational levels eg. revenue / billing agents, channel partners, vendors, etc
- Share recipients must agree to repatriate funds
- Officers (both China and US) would need to be in the know
- Auditors must be fooled or complicit (share capital discovery, annual impairment tests, undisclosed liens against assets, etc.)
- Potential tax leakage for the recipient of the funds
- Difficult to accelerate restricted or un-earned shares without tipping off auditors / investor
- Would be much easier, more efficient and less scrutiny for the company to do a follow-on offering or issue convertible debt to commit round-trip revenues back into NQ
- Any capital borrowed would still need to be repaid to the lender, with interest
Its theoretically possible for NQ to be round-tripping by issuing shares, but not likely. At worst, shares issued and earned might be sold on the open market or borrowed against, but that would be the maximum theoretical extent of any round-tripping. What makes this theory particularly strange is the short thesis neglects to consider that any cash raised via borrowing against shares will ultimately need to be repaid, with interest to a lender. Given how the company has decided against any massive share repurchases and no officers are scrambling to flee China from creditors, I think at a minimum, it's safe to assume that restricted shares and un-earned shares have not been used as collateral.
Claim: Flat cash balances year over year is strange! The company must be manipulating cash.
There are 3 major issues confounding this erroneous observation that flat cash balances = fraud.
1. NQ's cash balances during Mr. Zangrilli's measurement period includes the IPO proceeds thus creating a large base to calculate off of relative to NQ's ability to affect cash balances from cash flow or corporate actions
2. NQ has historically been too small to access the capital markets thus necessitating the company to use existing cash and stock to acquire new technologies. When compared to Mr. Zangrilli's portfolio of 75 companies, most of which are significantly larger, generate much more cash flow relative to cash balances and have access to various financing alternatives, the flatness in NQ's cash balances become even more glaring
3. Mr. Zangrilli's analysis excludes NQ's convertible debt offering but includes the Atlantis cash proceeds when calculating change in cash for NQ. All 75 of his comparable companies are allowed to retain any and all historical capital raises. If you exclude capital raises for NQ, you should be excluding capital raises for all 75 companies to maintain parity
Notice how cash is flat in Mr. Zangrilli's 2013 estimate. Notice how change in cash is ignored from 2008 - 2011. Why is Atlantis cash raise included but Convertible debt cash excluded? Including both would vastly discredit Mr. Zangrilli's thesis.
What's essentially happening is that a large denominator (existing cash balances) + relatively small cash flow generation = only slightly larger numerator (existing cash + cash generation). The flat cash balances is not unusual for a small, barely cash flow positive company that just had an IPO and is now cash rich. Stacking the deck against NQ by deliberately ignoring cash raised from the convertible bond offering and fixating on only current periods completely invalidates the short analysis.
Claim: Asiainfo Linkage is the missing link that binds all fraud.
While James Ding is associated with Asiainfo Linkage, his involvement with the company on an operational level ended in 2010, when NQ Mobile was still a fledgling startup and pre-IPO. The supposition that Asiainfo Linkage's existing management would fraudulently bill revenues in NQ's name would imply Asiainfo Linkage, as a publicly traded company, was party to fraud and willing to jeopardize its own operations over a tiny private company generating less than $18M in revenues, globally. Even if such a relationship did transpire, it would have ended a while ago due to fears that a discovery of this nature would have been discovered as part of the audit Asiainfo Linkage underwent as part of its pending buyout.
This theory is no better than playing 6 Degrees of Kevin Bacon. If there ever was a case of guilt by association, this would be it.
Claim: In the event fraud is discovered, NQ's Variable Interest Entity (VIE) structure allows the company to keep assets in China and away from shareholders
While Mr. Zangrilli has an excellent point that the VIE structure allows for significant latitude in terms management's ability to defraud shareholders by ignoring the legal claims between the HoldCo owned by Shareholders vs. the Operating company controlled by management, this weakness in the VIE structure is not unique to NQ, but rather a common weakness experienced by most Chinese companies listed as ADS in the US. It's actually at this point where I will decline to opine and let the accountants and lawyers determine if NQ's acquisition deal structures are legitimate or not and whether or not shareholder interest are being upheld. I will however point out the logical flaw here in the short assumption that NQ is using the VIE to defraud investors. NQ does not need to do any acquisitions to perpetuate fraud utilizing the VIE structure. A simple share offering or convertible debt offering can be done, with funds repatriated back to China as discussed in Toro Investment Partners article NQ Mobile: Behind Smoke and Mirrors Lies The Truth Part 2. Not only is this alternative cleaner and easier to do, it also requires less collusion amongst channel partners and leaves a much smaller paper trail (no new minority shareholder partners, no restricted stock issues, no need for channel partner collusions, no complex accounting, no impairment tests).
If you are not a lawyer or an accountant, its best to leave the VIE hypothetical's to the professionals.
Keeping Score in the Round-Tripping theory
While the short thesis of round-tripping apparently rests on some fairly shaky foundations, let's see what results come out of Mr. Zangrilli's model once we adjust for some more realistic assumptions. The below table reflects the major assumptions within Mr. Zangrilli's model along with my commentary and adjustments.
Revenue or Cash You Count
Short Thesis - Why
Trade Star Commentary
Trade Star Model Adjustments
Revenue on cash & investments
Auditor verified cash and investments at the bank; money made on this cash is probably real
-Short thesis is double counting forex in model
-Forex proceeds adjusted
Cash from fundraising
We know the IPO and convertible bond happened and was real
-Short thesis excludes financing fees: Atlantis + IPO + convert
-Includes fees adjustment
-Convert remains excluded to stay consistent with short assumptions
Cash from stock sales of collaborators and loans against restricted stock in acquisitions (China only)
Bank in China is controlled by three founders; VIE structure unlikely to stand up if founders refuse to honor agreement; collaborators have much to gain and little to lose by depositing money in this account
-Agree that common stock can be used as collateral
-Trade Star rejects using un-earned shares as collateral
-Short thesis neglects to tax affect cash proceeds before round-tripping back into NQ
-Short thesis incorrectly accounts for share vesting
-Assumes un-earned stock not available to borrow against
-Assumes cash paid to Target tax affected at 20% before round-tripping
-Includes tax affected cash associated with equity investments
-Corrects model for proper vesting dates
Proceeds from stock option purchases of employees
The shares sold are likely verified by the auditor
-What about employees under cost of revenues?
-Assumes inclusion of 100% of Cost of Revenue expenses
Expenses you count
Payments to employees
Auditor likely verified payroll and tax payments; employees do not want to owe taxes without actually getting paid; unlikely executives included hundreds of employees on the fraud
Cash you ignore
Revenue from operations
You suspect this is the company's cash as it comes from third-party payment processors affiliated with the company
-Short thesis excludes ex-china revenues. Why? Round-tripping thesis is based on China acquisitions, not ex-china revenues
-Should include 50% of revenues (ex-china revenues), but given this is the short thesis, we'll continue to use their assumption of 0%
Expenses you ignore
Payments to companies
It is much easier to get a small number of companies to cooperate with your fraud; possible to have a small number of related people controlling a large number of companies
-Short thesis excludes Cost of Revenues which includes employees, user acquisition fees, and hardware. There is real product there and channel partners
-Assumes inclusion of 100% of Cost of Revenue expenses and inventory expenses
Adjustments you make
This is not an actual cash expense to the company
Same as above
Please note, this table was recreated from Mr. Zangrilli's original report and modified to include my comments.
So what do the model results tell us?
After re-running Mr. Zangrilli's model, using the adjustments I have listed above, I come up with a cash similarity of 97% right after IPO, dropping to 88% by 2q13. 3q13 cash similarity is at 91% due to NQ's multiple acquisitions in that period.
What does this mean? Well first off, the creation and use of a "cash similarity" metric is a real head scratcher. It's effectively a single period snap shot of the cash generated by round-tripping relative to reported cash balance at that same period of time. It's certainly a novel concept, but not terribly useful in terms of analyzing a round-tripping exercise. Fraud is not something you turn on and off each quarter to manage cash balances. Plus the cash similarity metric is heavily skewed by the large denominator / numerator issue previously mentioned. A more accurate analysis would look at cumulative cash generation over time using the round-tripping cash flow model to drive expected quarter end cash balances relative to reported cash balances. The net effect over time is what is going to tell you if there is real round-tripping here or just a fun math exercise. When you adjust Mr. Zangrilli's cash flow model to calculate cash cumulatively over time, it quickly becomes evident that the Round-Tripping theory does not work as NQ operates on a negative cash flow basis every quarter assuming the short analyst theories are correct. By 3q13, NQ's cash balances based on cumulative round-tripping since 2q11 results in an estimated cash balance of -$200k, or roughly 0% of expected balances. Essentially the short thesis does not work.
The modified model can be found here.
Note. Aside from my adjustments to Mr. Zangrilli's model, the model is substantially the same. I have fundamental issues with how it was originally structured and would not recommend its use for any purpose. The nature of this analysis ultimately drives very little information for an investor as the assumptions lack a consideration of revenues of any sort.
Mr. Zangrilli's model can be found on his article here.
So what's the conclusion?
- A lot of fanciful theories substantiated by incomplete math and a basic understanding of accounting can lead to some interesting, but misleading results
- Taking a broad swath of companies to compare against a single target is… unconventional and likely to result in a lot of work for very little useful data
- Any investor in NQ should dig a bit further into any analysis both long and short before singing praise. Handing out high fives within hours of reading the latest short report seems premature, especially if your firm's reputation is about uncovering fraud and business and accounting irregularities, three of the very core topics in Mr. Zangrilli's article. Recall my observation of Muddy Waters Research's praise about an hour after the release of Mr. Zangrilli's article
- If the short thesis relies on hypotheticals and innuendo, logical and well researched quantification needs to be provided to substantiate those claims
- NQ is likely not round-tripping as the Short thesis suggests as the cash generation from supposed round-tripping is not there to support it