What does it all mean and how can you use it? Arvid Tchivzhel refers to a few buzzwords: data management platform, enterprise resource planning system, data lake (his personal favorite), business intelligence, and big data (the king of all buzzwords, according to him). From the executive to the entry-level analyst, it’s likely people have heard at least one of these terms in their quest to use the latest in technology and analytics to make operations run smarter/ faster/ better. Unfortunately, none of the above in isolation will solve your problems, he says
Database specialists and data scientists will balk at the following oversimplification, but for the business analyst or the executive, there are a few things to know about big data:
Yes, it stores massive amounts of data, but more importantly it scales well and is cost effective.
It is capable of storing ‘unstructured’ data, which means you can cast a wide net of data sources and types without concern of frustrating a DBA. This lets you keep everything without being forced to presage your schema or dataset from day one (if six months into your project you realise you need one more column for your analysis, you don’t have to redefine all your tables and data processing) – i.e. it gives you flexibility.
Setting it up from scratch is hard, and you will make mistakes along the way. Prepare for being over budget and expect to revise the timeline you put together at least five times.
It should be used in conjunction with traditional databases, data processing, analytical tools and creative analysts. Just because you have a data lake doesn’t mean you should throw out your SQL database.
Start with the end in mind. This is of utmost importance to executing a big data project.
Before walking down this path, decide what you will use it for: Targeted marketing? Churn modeling? Customer lifetime value? Content recommendations? Advertising pricing? Better e-commerce flow (better conversion through paywall)? Then, set goals for yourself, and use your new big data tool to measure your performance: improve customer retention by 2 per cent, raise advertising yield by $1, increase average online user engagement by five page views and 30 seconds, etc. Finally, build it for the person who will use it every day to actually implement and fulfill your goals.
How publishers are being successful
Starting with the end in mind, the most successful publishers map each output back to the original data source. For example, if the goal is to generate a targeted email marketing campaign, the publisher must decide what characteristics and behaviours are important in defining the email lists. Basic subscriber information comes from a customer management system, online engagement data can come from various web analytics tools (being able to capture the registration ID and match it to the CMS is a key), opt-in to a loyalty programme comes from a separate loyalty database and demographic and psychographic data comes from third party sources and is matched at the household level. These sample data sources give the publisher the power to filter and refine their targeting by subscription type, online behavior, enrollment in a loyalty programme and demographics.
While the four data sources noted above are by no means comprehensive for a publisher, they demonstrate the task of data source mapping to deliver the output desired. Once the sources are mapped, a process called ‘customer resolution’ is performed to match the various sources between each other. Reconciling customer data across multiple systems typically employs an algorithm to improve match rates (for example, Jonathan might match John if there are other records to suggest this is the same person). This creates a unique customer record with a link to each system, what generally is called a ‘golden customer’ (another buzzword).
With this ‘unsiloed’ dataset, you can begin to explore your data using standard analytical tools such as Tableau for visualisation and R for advanced statistical analysis. This is where publishers are able to drive real insights and conclusions about who their customers are and how they behave. The more you know about your customer and how they interact with all parts of your organisation, the better you can build your products/ prices, message your customer, apply a consistent experience and value them appropriately.
Taking a moment to think back on how decisions used to be made using siloed data sources and reports, quarterly (at best) responses, print-only and survey-based preferences information, the publishing industry has evolved dramatically in the last few decades. Real-time analytics, consolidated big data platforms, revealed preference from actual online engagement data, cross-device tracking and heavy competition from bloggers and content aggregators have forced publishers to take on the challenge and become smarter about how they operate. The phenomenon is certainly unique to publishers, but similar transformations have happened in other competitive industries such as airlines, hotels and car rental services.
Big data and the customer
A properly executed big data solution can be incredibly valuable for you and your customers. Once you have completed the ‘customer resolution’ described above, there are several ways to personalise the way you interact with your customer:
Targeted mail and email campaigns: This can be used for various purposes such as preempting churn or reinforcing your message for retention, upsell/ upgrade offers and driving online engagement. Create a unique list with a customised creative and message specifically written for a specific person or customer segment.
Targeted renewal pricing: Each customer has a different level of engagement with the product, varying demographics, tenure, acquisition channel and price elasticity. Therefore, following the lead of hotels and airlines over the last few decades, a unique price can be given to each subscriber to minimise stops from price-sensitive subscribers, but also maximise the yield from subscribers who are not sensitive to price.
Targeted reacquisition offers: Every publisher has a growing list of former subscribers, and knowing which ones will yield the best ROI for the acquisition expense will ensure cost savings and preserve volume.
Customer lifetime value: Use profitability per subscriber and expected retention to segment your subscribers into high/ medium/ low value subscribers. This is what airlines use to ensure the most valuable customers have a positive experience and maintain loyalty, mid-value customers continue building the relationship and move into the high-value group and low-value customers are upsold and given opportunities to engage and move into higher value groups.
Customer service: Use CLV to prioritise calls and use custom customer service scripts to ensure a customised experience for each subscriber.
Personalised messaging upon login: Customers are coming to expect a personalised online experience (Facebook, Twitter, etc.), so adding custom content, messaging, branding and advertising gives the user a 1:1 relationship rather than a one-size-fits-all approach.
Personalised messaging for anonymous users based on behaviour: If an anonymous user is observed to be continuously consuming sports content, they are unlikely to accept a bundled or print product. Thus, the next time the sports-focused user comes to your website, give them a sports-themed offer or a sports product/newsletter.
(The writer is a director with Mather Economics. He oversees the delivery and operations for all of the company’s consulting engagements, along with internal processes, analytics, staffing and new product and services development. Experienced in econometric modeling, forecasting, economic analysis, statistics, financial analysis and other rigorous quantitative methods, he has led numerous consulting engagements across various industries.)
May 2015