Daily Daze

Source: Groupon Doesn’t Have The Cash to Build Its New Data Center

groupon Source: Groupon Doesnt Have The Cash to Build Its New Data Center

Apparently short a few G's.

In the run up to its recent IPO, a number of outlets reported that the daily deal giant Groupon was running short on cash. Over at the Motley Fool, Evan Niu noted that the company had $243.9 million in cash and equivalents at the end of September, compared with $465.6 million in accrued merchant payables, that is, the money they owed people who ran Groupon deals.

The company ended up raising $700 million, but according to a source familiar with its business, there still isn’t enough cash on hand to make critical structural improvements the company needs to grow. Groupon is shelling out millions every month on hosting costs, and paying a premium to third parties. The company is very eager to construct it own data center, but simply can’t afford it.

As ZDNet notes, Groupon’s massive scale is a challenge and an opportunity. “Groupon’s growth—revenue, subscribers and merchants—is off the charts and under the hood rests on whale of an information technology story. In many respects, Groupon is the ultimate IT petri dish. In two years, it has grown from a company that could be run on a simple spreadsheet to one that needs systems spread across the globe.  Meanwhile, Groupon is a greenfield opportunity—there aren’t legacy systems dating back decades.”

The company actually made a note of this in its IPO prospectus: “We have spent and expect to continue to spend substantial amounts on data centers and equipment and related network infrastructure to handle the traffic on our websites and applications. The operation of these systems is expensive and complex and could result in operational failures. In the event that our subscriber base or the amount of traffic on our websites and applications grows more quickly than anticipated, we may be required to incur significant additional costs.”

You can see some of what Groupon had in mind on this job application the company put up for a Performance Engineer.

Follow Ben Popper via RSS.

Comments

  1. Anonymous says:

    Groupon doesn’t really need to “build its own data center.”  There are dozens of modular DC companies around the country that could easily give it the infrastructure it needs.   And there’s no reason why traditional colocation wouldn’t work for them.  Groupon is not like Facebook even remotely.  They don’t have millions upon millions of users logged on to their systems simultaneously chatting, uploading, refreshing, etc.  Groupon’s main needs are probably basic redunancy for web, application, storage, and financial systems.  Even with the sales volume they have I would think 5-10K square feet of colo space would be more than sufficient and there are several dozen top-tier data center vendors that could give them that (and hundreds of 2nd- and 3rd-tier).  The main cost is the hardware they’d have to purchase for the technical infrastructure – i.e., servers, network switchgear, routing gear, etc.  That probably would be tens of millions in cost upfront if you include primary and failover/load-balanced sites then still millions a month for their power/space/bandwidth costs (and upkeep – i.e., SAs, network engineers, storage engineers, etc.).  My biggest concern for them would be heavy over-investment in technical infrastructure “just because they can.”  Oh, we’re a big deal now so we need a hundred-million dollar data center.  No, you don’t.  Not even close.  Groupon should keep its infrastructure costs to a minimum and let someone else run it – i.e., Amazon – and wait until it has survivable profitability before it begins exploring those options.  

    1. David Candle says:

      Sounds like someone who (a) has no idea what they’re talking about, and (b) wants us all to know it.

      1. Anonymous says:

        Groupon is NOT a technology company on the same level as Facebook. ‘boece’ is almost spot on, aside from having Amazon run their infrastructure (AWS is overrated) there is no reason to be building datacenters. The hard part is the software complexity in getting Groupon to work on multiple datacenters. Groupon is quite clearly spending all of their money on people selling to businesses to get more groupons.

    2. David Candle says:

      Sounds like someone who (a) has no idea what they’re talking about, and (b) wants us all to know it.

    3. Anonymous says:

      I’m 

    4. Ben Popper says:

      I’m not an expert on this stuff, so I appreciate the input. 

      My reporting is based on source who works closely with Groupon infrastructure. Groupon has 150 million daily subscribers and growing. People are transacting deals both on their website and on mobile devices. My understanding is that you end up paying a 2-3X premium when you rely on Amazon at this scale, versus building your own. So a $100 million outlay would pay for itself within 1-2 years. 

      1. EH says:

        How experienced is your source? There are a myriad of hosting possibilities between Amazon and “building your own datacenter.” Having 150MM daily subscribers is something, but there are a lot of companies who have been sending out well over 150MM notifications (even emails) every day, constantly, for years and years. The deal flow in response to that information is a real challenge, but there’s no reason to frame this story in terms of  2-3x premiums when wouldn’t reducing that even to a solid 1.5x premium be a huge win at their scale? 1.25x? 

        I’m not sure if Boece was touching on this WRT infrastructure, but from reading the ZDNet story I think it’s likely that what Groupon is having trouble budgeting for are their Salesforce and NetSuite ensembles. The story is written in terms of Groupon’s customers, who are the businesses offering coupons, so Salesforce and NetSuite are going to be implicated highly there. It doesn’t take much infrastructure to let people sign up and claim coupons, it’s the other stuff that’s used to manage the customer side. Not to mention that Salesforce-type consulting has a long, storied and extremely-profitable history.

  2. Crash says:

    I’m perplexed why this company hasn’t moved their stuff to the cloud yet. Plenty of excellent companies out there like salesforce.com, which could make running their stuff a breeze.

    I guess when you’re always a week from going bankrupt, these decisions aren’t at the top of your list.

    1. EH says:

      If you read the article, you’ll see that they’re already using Salesforce.

  3. Crash says:

    I’m perplexed why this company hasn’t moved their stuff to the cloud yet. Plenty of excellent companies out there like salesforce.com, which could make running their stuff a breeze.

    I guess when you’re always a week from going bankrupt, these decisions aren’t at the top of your list.

  4. Ryan Lackey says:

    The reporter doesn’t seem to be aware that there’s a spectrum of hosting.

    At one side is platform as a service or managed platforms on top of EC2 — Heroku, EngineYard (which Groupon appears to use), etc.  These are easy to set up, and have lower overhead for small or medium organizations.  They have the highest per-unit cost, but for rapidly growing or small organizations, can be worth it.

    At the other side is “build your own datacenters”.  This has the lowest per-unit marginal cost, but high fixed costs, in time, money, and focus.  Only companies for whom datacenter construction and operation is a core value (e.g. Google) really do this on good advice, and only after a certain point (Google used to lease space in existing datacenters).  Facebook, Amazon, etc. are on the path to this — Amazon resells the EC2 platform from largely leased datacenter suites (with their own equipment buildout internally).

    In between are EC2 (an infrastructure as a service provided by Amazon, priced less than the high-value-add platforms, and with a utility model, but yes, more expensive per unit), managed dedicated hosting (e.g. Rackspace, Softlayer), colocation (various, ranging from single servers to cages to suites to floors to buildings, e.g. Internap, Equinix, …).

    Saying “doesn’t have enough to built its own datacenter, so they’ll stay on EngineYard” is pretty vague — that’s like saying “doesn’t have enough money to buy a Bentley, so will continue to walk.”  There are a whole lot of cars less expensive than the Bentley but faster than walking.

    1. Ben Popper says:

      Thanks for the input. 

      So you’re saying the setup they have now works best for small and medium sized organizations but has the highest per unit cost. Groupon is clearly a large business now. 

      To me that fits with the idea that they are looking to make a transition but have not accomplished it yet, whether that change is moving to EC2 or building their own data center. 

    2. Jana Boruta says:

      They are hosting on a PaaS provider but they grew so fast they didn’t have enough time to hire system admins in house or gain the stills needed to keep such a massive site like that running. 

  5. Ryan Lackey says:

    The reporter doesn’t seem to be aware that there’s a spectrum of hosting.

    At one side is platform as a service or managed platforms on top of EC2 — Heroku, EngineYard (which Groupon appears to use), etc.  These are easy to set up, and have lower overhead for small or medium organizations.  They have the highest per-unit cost, but for rapidly growing or small organizations, can be worth it.

    At the other side is “build your own datacenters”.  This has the lowest per-unit marginal cost, but high fixed costs, in time, money, and focus.  Only companies for whom datacenter construction and operation is a core value (e.g. Google) really do this on good advice, and only after a certain point (Google used to lease space in existing datacenters).  Facebook, Amazon, etc. are on the path to this — Amazon resells the EC2 platform from largely leased datacenter suites (with their own equipment buildout internally).

    In between are EC2 (an infrastructure as a service provided by Amazon, priced less than the high-value-add platforms, and with a utility model, but yes, more expensive per unit), managed dedicated hosting (e.g. Rackspace, Softlayer), colocation (various, ranging from single servers to cages to suites to floors to buildings, e.g. Internap, Equinix, …).

    Saying “doesn’t have enough to built its own datacenter, so they’ll stay on EngineYard” is pretty vague — that’s like saying “doesn’t have enough money to buy a Bentley, so will continue to walk.”  There are a whole lot of cars less expensive than the Bentley but faster than walking.

  6. Robert says:

    These are the same guys that turned down billions from Google!  Idiots