This holiday retailing season, as customers shopping preferences shift, will you still be marketing to them using stale data?
With the rise of mobile devices, and the “always-on” user, the amount of time spent on the
internet has nearly tripled over the last 5 years: 450 billion minutes per month in 2010 to more than 1200 billion minutes per month now. The velocity of data is going to continue unabated growing into the future, with some projections pointing to another 10X increase in data velocity by 2020. Interestingly, the increase in the overall amount of time spent on the internet has also translated in users spending more and more time with the same apps or brands.
All of this additional user time is generating behavioral clickstream data for companies at speeds faster than ever before. At large omnichannel retailers, the volume of clickstream data generated in one day now rivals one year of PoS data: or, in other words, there are 300-1000 pieces of unstructured clickstream data for each purchase.
While marketers have long understood the importance of “RFM” (recency, frequency, and monetary value), with the increase in volume of data every day, “recency” has become ever more important. Without near real-time usage of behavioral clickstream data, the value of the data decays, making it meaningless for targeting. For instance, during the holiday season, which is typically the biggest season for retailers, many users are shopping for gifts, and their purchase behavior deviates significantly from the norm. Businesses that can develop processes to understand and react to such data quickly can earn superior engagement and profits.
Despite the rise of big data technologies, most CMOs are increasingly feeling underprepared for “data explosion”. What are the top initiatives that can help CMOs get ready for this new age of high velocity data? Here are our top 3 recommendations:
- Process streaming data, and store everything: When dealing with low velocity data, you would first model the data to develop a data-warehouse schema; data that’s not modeled would be discarded. With high velocity data, however, you need to complement your data warehouse strategy with schema-less big data infrastructure that can store all data. The idea is to give analysts the ability to play with data to discover insights that can then be modeled. A good example is from Orbitz, which started collecting unstructured data around trip planning from users, and went from 30TB of data storage to 750TB, revolutionizing their hotel sort.
- Understand “identity” across platforms and channels: Users are increasingly adept at switching between devices and it’s not uncommon for an user to use 3-4 devices, often times in single day, while they shop around. To understand each user, you have to develop infrastructure that ties together seemingly disparate points of data from desktop & mobile into one unified profile. In addition to using well structured identity information like email addresses or customer ID, smart marketers are also looking at fingerprinting technology to fill the gaps in their knowledge.
- Get machines to help humans with analysis: Once you have the ability to process and store streams of real-time data, the next step is to have your analytics keep pace with the speed of data collection. Purely manual process can impose delays of weeks, and CMOs need to provide machine learning tools to assist their teams of analysts in uncovering insights. For example, creating lookalike audiences with machine learning on real-time data, can help marketers acquire more high value customers. During the holiday season, and other times when user behavior changes significantly, machine learning will always be a step ahead of human modeling. Machine learnt models can then be refined more by humans who can layer in additional business logic.
- Reduce the time to action with automation: Not only do you need to process data in real-time, you need to be able to act on your analysis faster. This requires a high degree of automation. In the old world of slow moving data, you might have tolerated a 24 hour delay for ETL processes to load the data in your data warehouse, as well as several weeks of delays imposed by manual analytics processes to leverage the data, and batch processes to act on the data.
However, in the high velocity world, actions need to be automated to respond to various behaviors, in a personalized manner. Simple automations like browse abandonment emails, or “related products”, can go a long way, as this example from Amazon shows.
The best consumer marketers of tomorrow will be the ones who embrace the challenges of high velocity data. Need for speed, and higher degrees of automation, will become critical capabilities for marketing organizations in this new world.