32
AUTODESK
How Big Data Is Transforming The Software Industry
Background
Autodesk are a Californian software publisher with the core business of developing commercial computer-aided design (CAD) software. Starting with AutoCAD, they have gone on to develop specialized applications targeted at individual fields in design and architecture, such as Revit (construction), Moldflow (production) and Maya (graphics and effects for entertainment media). Their products have become industry standards in many of these fields.
Recently, in line with many other big software producers, Autodesk have made the leap to offering their products via a software-as-a-service (SAAS) model. This means they now have access to a far larger volume and richer variety of data on how their customers use their products.
What Problem Is Big Data Helping To Solve?
Before the arrival of SAAS, companies which developed software had access to relatively little information about how it was used. Usually the only channels for gathering feedback were customer surveys and feedback forms included in physical product packaging. However, software developers were always aware that only a relatively small section of the user base would ever use these. This would be particularly true if everything were going well: realistically, how often do you bother to contact a business whose services you are happy with simply to congratulate them on a job well done?
Of course, just because the software was working fine and the customer was able to do the job they purchased it for didn’t mean improvements couldn’t be made. Software applications must constantly evolve to keep up with the competition as well as the growing demands of their user base – and customer feedback provides the most obvious and valuable source of direction.
Charlie Crocker, Business Analytics Program Lead at Autodesk, tells me: “In the old world, being able to understand our customer was relatively difficult. We would understand them on a broad wavelength – perhaps every six months or so we would understand them, when they filled in a survey or we invited them into a focus group or went to visit them in the field.
“We had a couple of tools in the products that could collect error reports and that sort of thing, and for a long time that’s what we did – we got very successful at it.
“But in this new world, we need to be able to understand our customers on a daily or hourly basis. We need to understand what the bottlenecks are in their customer experience.”
How Is Big Data Used In Practice?
With the product being hosted in the cloud, Autodesk can closely monitor and track just about every aspect of how their customers interact with their products. It also means updates and fixes can be applied at any time. Autodesk’s software developers can accurately gather deep insights into how, when and why their products are being used, meaning focus can be shifted to providing support and enhancements for features that a significant proportion of the user base relies on. Meanwhile, lesser-used functionality can be scaled back or removed, if Autodesk’s analytics tell them their users are not getting value from it.
Of course, ultimately the aim is to make sure customers re-subscribe to the services they’re paying for, when the time comes for them to make a decision. Data relating to how a subscriber has used the service in the last 90 days before their current subscription expires is seen as being the most relevant and is subjected to the deepest scrutiny.
As Crocker told me: “Understanding the user’s behaviour and what drives people to renew their subscription is a big deal. Keeping customers is much easier than constantly having to go out and find new ones.”
In order to gain additional user feedback, the company also make early, pre-release builds of many of their popular products available through their Autodesk Labs service. This gives them valuable insights into the sort of features and functionality that their users are interested in seeing included in upcoming services, and new plugins and extensions for existing packages.
What Were The Results?
The most dramatic result has been the speed at which insights into user behaviour can be gained, and consequently the reduction in time before action can be taken on them. This has led to a distinct closing of the gap between issues being highlighted as problematic by user actions and a solution being deployed. Crocker said: “It used to take six weeks for a signal that started in one of our products to end up in our data lakes. Now with these SDKs it can show up in a couple of hours.”
What Data Was Used?
Autodesk monitor around 600 data points on how each of their users interact with their cloud-based SAAS platforms. The length of time a user is engaging with the service, as well as precise details on what functionality is accessed and what is ignored, are recorded and stored for analysis. They also include internal business data such as transactional records into their equations so the overall expected lifetime value of a customer can be taken into account, and features which attract the big spenders can receive priority attention.
They also monitor how frequently the inbuilt support channels, such as live customer service, online chat and support forums, are accessed, and data from these sources shows where people are having problems and gives clues as to what actions could be taken to pre-emptively solve them.
One indicator that remedial work needs to be concentrated in a particular area is when one issue is generating a higher-than-average number of one-to-one contacts with customer services, via telephone, live chat or email. Data showed that these problems were disproportionately expensive in terms of time spent by Autodesk responding to customer service contact. At one point, one such problem revolved around failed attempts to activate products by entering registration keys. When it became apparent how expensive it was becoming for Autodesk to respond to technical service contacts, further resources were allocated to the development teams working to fix the bugs which were causing the problems.
What Are The Technical Details?
Autodesk have gathered around 800 terabytes of data on how their customers are interacting with their cloud-based services, and are currently accumulating it at a rate of 100 gigabytes a day. The company use a distributed storage network running Hadoop on Amazon S3 servers, which they analyse with Amazon Elastic Map Reduce. Other Big Data technologies and platforms used include Datameer, Splunk, Amazon Redshift, Hive and Google’s BigQuery engine.
Any Challenges That Had To Be Overcome?
In the old days, the initial cost of providing a new customer with a service would have been limited to sending them a DVD and instruction manual in the post. While this has been eliminated by the SAAS model, each customer now brings ongoing costs, in terms of computer resources that must be made available for their needs. This means each customer now has an ongoing “running cost”. Crocker says: “Every customer has a cost, so we have to understand that or it will be very difficult to provide our services and still make money and keep the shareholders happy.
“As the number of users goes up, we will continue to drive cost consciousness. We’re able to now understand which parts of the system are being used most, and which are cost inefficient. You cannot operate without deep visibility into that information, and that information is structured and unstructured – it’s messy, but there is amazing context within it.”
What Are The Key Learning Points And Takeaways?
The focus away from a “ship and move on to the next product” mentality towards providing a constantly updated and evolving cloud-based service has brought technical problems and data headaches but has also provided businesses with opportunities to far more deeply understand and connect with their customers.
Capitalizing on this – in order to mitigate the increased cost of providing cloud customers with ongoing processing bandwidth and storage – is the key to success in the era of Big Data-driven SAAS.
By integrating data analytics into customer services and product development, these businesses have the opportunity to more closely match their offerings to what their customers want. Eliminating any middlemen involved in the process, such as retailers, inevitably brings the customer and the service provider more closely together. In theory, and as seems to be occurring in practice, this should mean less disconnect between what a customer wants and what a provider provides.
REFERENCES AND FURTHER READING
For more insights on the use of big data at Autodesk, see: