The story that won't go away
More detail emergesSo the Autonomy kerfuffle rumbles on. Nearly a week in the story still hasn't gone away.
Of course we still lack precise details of what has been alleged. After the initial rush of coverage subsided we've now had a crop of "inside Autonomy" articles - the NY Times on Monday, Reuters today long on colour, short of detail.
A few more hints came out in this WSJ article (if you don't subscribe Google the headline and then access it via Google News). It flagged some some interesting stuff about channel stuffing with reseller Tikit, and round-trip accounting with customer VMS.
(In interests of disclosure, I corresponded with one of the WSJ authors of the article before publication, but none of these allegations came from me.)
Mike strikes back
I think this post from AllThingsD sums up his PR strategy quite succinctly. The line of attack is to muddy the waters by focusing on HP's (undoubted failings). As I said before this is something of a red herring - HP being complete incompetents and Autonomy being dodgy are not mutually exclusive conclusions.
Time to take a breakMy suspicion is things now take a pause. Given HP has referred the matter to the authorities, I suspect the case is now sub judice. This means we are unlikely to get any more details soon, no matter how much Mike yells for them.
So given there's probably going to be a break in the newsflow I want to look at the broader issue of how software companies dress up their accounts. I suspect this will provide some useful context for what may or may not follow.
First I want to explain in simple terms how and why companies massage their numbers. Then a few thoughts on why it happens to software companies so often. And finally I want to present a rough guide so some of the more common wheezes.
Why software companies massage their numbers
Valuation ratios and perverse incentivesYou would have thought it was self-evident why companies want to boost profits. More profits = a higher share price right? Actually its slightly more complicated than that.
I wrote about this back in October. In summary the only way to properly value a company is to figure out the net present value of all the economic benefits that will accrue to you from owning it. In short either DCF or EVA the bastard - the maths come out at the same answer.
But as I wrote in September, people often use valuation ratios as a short-cut. The key point is that while valuation ratios provide a sense-check they should never be a substitute for an in-depth long-term valuation. If you do "value" a company based on a valuation multiple you are implicitly making one very big - and very dumb - assumption:
You are taking the value of one year's revenue / EBITDA / earnings and extrapolating the entire value of the company forever and ever and ever based on a single number.That to me is completely nuts. While this year's number is an indicator of how things are going at the moment, in most cases it says very little about how things will be going in two, ten or fifty years time. US quarterly reports are the most heinous example of this - Mr Market regularly marks the long-term sustainable value (i.e. the share price) of US corporates up or down five or ten percent based on three months trading data. Signal & Noise.
But back to the topic in hand, the other consequence of this is a perverse incentive for management. Because in order to maximise the value of their company they don't have to build a long-term sustainable business. The only need to pump up the near-term numbers until they can either cash in their options or attract an M&A bid. Agency Problem 101.
But why is it always Software companies?Okay so if Mr Market uses imperfect yardsticks like valuation multiples, all companies have an incentive to cook the books. But software companies seem particularly vulnerable.
In my time covering the European sector iSoft, FAST and Torex have come to trial, and I'm sure there are many more that have not. In the US Informix was the highest profile case (shock, horror CEO went to jail for a whole two months), but outside the courtroom Microsoft, Oracle, Computer Associates, Veritas, Sybase, Symantec, I2, Red Hat, Macromedia, Microstrategy and BMC have all had to restate their accounts.
There are two problems here. Means and Incentives.
Problem 1: The means are close at handMeans first. The basic problem is that a software companies don't sell physical stuff, they sell intellectual property (specifically software licences). As I've written previously, this has tremendous benefits with regard to creating a scalable, asset-light business model. But there are downsides. Because there's no physical product to be shipped it is very difficult to determine when you should recognise revenues and how to assess the costs.
Take revenue recognition as a case in point. The gold standard for this is US SOP 97-2, (in paraphrase: you need a contract, product must be delivered, the fee must be fixed and collection of it likely). However even under this regime much of the actual revenue recognition comes down to a matter of judgement. Let's give a practical example:
It's 11:50pm the 31st of December and a customer has just signed a $5m deal for your enterprise software package. At 11:55 you send them a download link with a copy of the installation files at the other end. However actually installing the damn stuff is going to take about 18 months and involve two dozen wage-slaves from Accenture. Testing it and getting staff trained up on the system is going to take much longer. Obviously you haven't received any cash yet - the customer's not that stupid.So when do you recognise revenues? You've got a deal, product (or at least a download link) "shipped", the price is in the contract and payment is likely (so long as you don't screw up). Do you recognise it all at 11:55? Or do you wait until Accenture have started, or Accenture have got so far the customer can't back out, or Accenture have finished and the customer has signed off, or the customer has trained their staff and they're all happy. And what if you find a show-stopping bug five months down the line? (Remember a v1 software package is probably the only product guaranteed to have defects on day of release!)
In the end, it comes down to your judgement. Useful...
Problem 2: The incentives are massiveSecondly on Incentives. Tech is intrinsically high growth (in a nutshell: Moore's Law). High growth companies attach high valuations. High valuations mean rich earnings multiples. This means the benefit from each additional penny of earnings when you capitalise it on 35x P/E is commensurately greater.
Coupled to that software companies have a culture of rewarding employees with stock, and you have ample to incentive to ramp the share price at least until your options vest.
In short the high rewards on offer to tech companies present ample motive to fix the numbers.
How software companies massage their numbers
Four ways to fiddle your numbersOkay that's the theory. How does it work in practice?
To start with a general framework. If you are trying to maximise profit you can do to things - either increase your reported revenues or decrease your reported costs.
In turn there are two ways you can do this - you can change the timing of revenues/costs (pull forward future revenues; push costs into the future), or you can artificially change the magnitude of revenues/costs (booking false revenues; not showing appropriate costs).
Stick these in a grid and this is what you get:
|What they didn't teach you in Accounting class...|
Let's go through these one at a time.
1) Changing the the timing of revenue
- Taking licences upfront: As I said earlier with software companies it is often a matter of judgement as to what proportion of a big licencing deal you book upfront. 30-50% is not unusual. 100% happens, but is probably imprudent - that would certainly inflate your revenues (and you all-important revenue growth rate).
- Misclassify types of revenues: Different types of revenue are accounted for in different ways - as I said above software licences might come up front while long-term consulting projects take longer. So if you mis-classify consulting as software licence (call it a "solution"... no one will know...) they you could take the revenues early.
- Channel stuffing: Often technology vendors don't sell to the end user, they sell to a reseller who holds a certain amount of inventory on hand to fill customer orders. So what you might do is sell more into the channel, book it as revenue. Except next time the reseller still has all of last quarter's kit lying around so doesn't need to buy more from you, and your revenues fall off a cliff. Let's be clear - variations between sell-in to the channel and sell-out to final customers is a normal part of doing business and very common in consumer hardware. But there remains scope for it to be done to an inappropriate magnitude (I think this is one of the things HP is pointing to with Autonomy).
2) Changing the timing of costs
- Merger provisions (and other provisions): When you do a big merger you typically take a whacking great cost upfront through the P&L which everyone ignores (because its seen as part of the capital cost of the acquisition). It then sits on your balance sheet as a provision - a kitty of expenses already taken against a rainy day. However if you find the provision you've taken is too big you can then write these costs back through the P&L. This write-back doesn't have to be disclosed on the face of the P&L, and reduces your costs in the period (inflating your profits).
- Capitalising R&D: This is a notorious one in the software world. When you develop a big product which is going to be sold over the next few years you typically don't expense all the R&D costs up front - instead you capitalise them on the balance sheet and expense them through the P&L (in the form of amortistion of capitalised R&D) over the period you sell the product. That's all when and good, as in the long-run amortisation of historic R&D should match up with captialisation of this year's R&D. But an obvious trick is to capitalise more this year then you did 2-3 years ago. This creates a mismatch between the (positive) capitalisation and the (negative) amortisation, boosting profits.
3) Changing the magnitude of revenue
- Round-trip transactions: Another one which has been hinted at in the Autonomy case. I sell software to a telco, they pay me not in cash but by giving me free phone services for a year. I record the software as revenues and the phone services as an equally-sized cost. That makes sense if they would have otherwise paid you cash and you would have otherwise taken that cash out and spend it on phones, but if that isn't the case the transaction does not reflect the underlying economic reality, and could potentially be inflating revenues.
- Side-agreements to modify a sales contact: This was the Informix example - they booked what looked to be perfectly respectable contract representing the sale of their gear with no recourse and standard payment terms (which is what they showed the auditors). But they also wrote side letters effectively saying "actually you don't have to keep the product or pay for it if you don't want it" (which they didn't show the auditors). Hmmm.
- Holding back revenue at acquisitions: It's April, you've just acquired a company but the deal doesn't close until June. You have a quiet word with your future minions at the target and ask them to hold back on sales until the deal closes. Then once the deal's closed in June they they go out, sign all the contracts and give you a bumper month which also includes sales held back in April and May. Effectively you are stealing revenues and profits from the previous owners of the company (who don't care as they've agreed to sell it to you anyway).
- Ignoring the bits with aren't growing: All software companies like to show off great growth figures. You can do this in two ways - produce great growth figures or change the basis of comparison. So you can do stuff like say "we are sunsetting 15% of out products which are declining at 50% y/y, so we don't include them in our growth calculation". Now there is nothing illegal as this - its simply a way of adjusting the numbers that any idiot with a calculator can see through. But you'd be surprised at how many analysts then use the "growth rate excluding the bad stuff" as the basis for future forecasts (which implicitly assumes "there will never be any bad stuff in the future"). Hmmm.
- Falsifying orders: Okay if you're going to fiddle your numbers you may as well fiddle them big. Invent a customer, photocopy their signature onto a draft contract, the possibilities are endless. In theory the auditors should catch this quite easily when they a) ask to talk to the customer and b) ask why the cash hasn't arrived, but you'd be surprised at how much they miss. This is especially true in software where you don't actually have to ship a physical product and produce a delivery receipt.
4) Changing the magnitude of costs
- Restructuring costs: This is linked to the point about merger provisions above. Sometimes there are costs which are so large ("Exceptional Costs" in accounting jargon) that companies tell you to strip them out of earnings. Restructuring costs when you're firing lots of people in a recession is a good example. These will be included in the statutory accounts, but often excluded from the "adjusted" EPS used for valuation purposes. People then go and blithely whack a P/E on that EPS to capitalise the value of the company, effectively saying I don't believe this company will ever take big "one-off" costs ever again". Completely above-board, and completely stupid.
- Using stock options: Another cost contained in statutory accounts (at least under IFRS) but often stripped out of "adjusted" numbers are stock option costs. Again completely nuts. I'll let Warren Buffet make the point: If stock options aren’t a form of compensation, what are they? If compensation isn’t an expense, what is it? And, if expenses shouldn’t go into the calculation of earnings, where in the world do they go?
A couple of caveats here:
- Not all of these are illegal or against the rules. Many of these techniques are, in the right hands, perfectly responsible ways of reporting. However in the wrong hands all of them have the potential to mislead.
- While I kicked off the post with our friends at Autonomy, this is a broader industry discussion. I am not alleging these things went on at any one particular company. I'll leave that to HP!
- Finally this isn't a comprehensive list - I am not a trained accountant, merely a (modestly) experienced practitioner. I'm sure there's stuff I've missed or just mis-described. There are certainly many tricks I have omitted, in particular ones less relevant to software companies like mucking around with investment income, plant depreciation charges and FX rates.
That's all for this week. It's getting deep into Friday afternoon and I really want to have a go on my new copy of Assassins Creed 3 which finally arrived yesterday (I still stick to the PC version, which is on delayed release). Ubisoft FTW! *
* Actually sat in on a meeting with the Ubisoft CEO last week. I think he's got a great thing going on the console execution front - Far Cry 3 also looks like another surprise hit given the Metacritic scores. I do think they have a yawning gap on the mobile side though, particularly if tablets/post-PC devices steal the market from nextgen consoles. But that's another story...