Few years back, though it feels like ages ago, I was working on my first web analytics implementation. As one of the initial tasks for the implementation, we went around the organization to explore what may be the burning questions that our fancy new web analytics tool needed to answer right away. We setup a meeting at the end of each day to brainstorm the questions that we got to figure out, how the tool could be setup to answer those. And then the questions started trickling in – « How many visits do we get daily? », « How many transactions did we have each day? », « Can we find out traffic and transactions volume by marketing channel? », « What are the keywords driving traffic to our site? », « Do we need more landing pages? », « What kind of errors do we see on the site? », « How well are our promotions doing? » etc.
As we started digging into the details of these questions, a pattern started to emerge, and we divided all the questions in the following categories:
1. I need more data
2. I need to know how my work had a positive impact on our performance
3. Tell me what’s wrong – with our websites, campaigns etc.
4. How do I improve it – Marketing campaigns, promotions, landing pages etc?
The first two categories here relate to the theory that if we had more data we could certainly improve our business. It may not make sense to disagree with this statement, unless one puts it like « I can’t improve my business because I don’t have access to more data ». And this notion, more often than not, leads to the behavior that if we keep reviewing data, something will emerge leading us to higher business growth. Again there is nothing wrong with the notion, but the likelihood that we will see that pattern, insight in a short amount of time, is relatively lower.
Category #3 assumes that in order to improve, we have to know what is wrong first. And this is a rather dangerous notion, which promotes status quo. The lesson for us here was to set the complete web analytics up in a way that promotes « there is no right or wrong answer, but there may be a better answer and we must, continually, look for that ».
The last category is rather open-ended and broad. In order to tackle it, we have to dig deeper within the business model, to figure out the process and feasibility of improving any area. So that brought us back to the point that, in parallel to the implementation of the web analytics tool, we need to paint a clear picture of how our business model operates and what kind of insights will help this business model grow.
Though one can use a similar process for anyone of many online business models out there, I’ll be using the example of transactional retailing business for this article. I’m defining a transactional retailing business to be where a consumer has to buy a product from the online store, for the business to generate revenue. Typically, the revenue generated by each transaction is based on the attributes of that particular transaction. So, Amazon.com, Overstock.com, Zappos.com, Travelocity.com, wyndhamhotelgroup.com are all example of online retailers. Whether a retailer is selling books, or hotel rooms, the following four key areas define the basic tenets of the retailing model:
- Outcomes – Most of the key performance metrics of the business, such as transactions, revenue, margin etc. can be defined as Outcomes.
- Demand – This is defined by number of visitors looking for the products being sold.
- Supply – Inventory of products, such as books, hotel rooms etc. and their respective attributes (pricing, quantity etc.)
- Execution – this is the process through which products are sold, e.g. Website, mobile site/applications.
With businesses moving online, the quality of business measurement improved exponentially overnight, and hence it almost felt necessary to optimize the business with this newfound measurement visibility. Optimizing all aspects of a business at one time can be a rather daunting task to setup, as well as to ensure whether it results in an overall performance improvement. Therefore, it is best to divide that in manageable parts, and leverage analytics to optimize that part in a way that it generates Incrementality overall.
It all starts with the definition and production of “Outcomes”. How do we know we have been successful? For a transactional retailer, this is about generating incremental revenue and the success is measured against a pre-defined baseline, either in form of an annual plan or prior year. “Conversion Rate”, “Transactions”, “Sales”, “Revenue” are all examples of “Outcomes”. Of course, each transaction within online retail is a result of matching demand with supply through an online platform. So, as we think about application of analytics within the retailing environment, the first step is visibility of “outcomes”, followed by each of the three areas, Demand, Supply and Execution, where analytics can add further intelligence. Visibility into outcomes is the most common step, and most of web analytics tool come well equipped with it. In this article we skip that to focus on the other three.
All retailing business need prospective customers coming through their doors/websites to generate any revenue, and the “Demand” relates with all efforts needed to generate demand for the products that is being sold by this business. All the marketing effort come under this section and any usage of analytics relate to generating maximum revenue within given marketing spend. The role of analytics is fairly visible in the definition of Demand itself. One is on the cost side – how do we minimize the cost of acquisition, while maintaining or growing the overall volume, and the other is on the revenue side – who do we go after to consistently increase transactions out of the visitor volume. So, from the acquisition channel perspective, one could either optimize the cost of acquisition in a way that generates the same amount of visitor volume and transactions, with lower cost, or generates higher visitor volume and transactions with the same amount of spend.
This relates to the inventory of products that any retailing business may have on its shelf to sell. All of the inventory attributes, such as pricing, cost of production, inventory management etc. fall in this category. The concept of supply may vary depending upon the type of retailing business. Some businesses sell the products that they do not build themselves, but acquire from others. Most of the products that Amazon sells are built by other companies. Some companies sell services which are fulfilled by others – travel and hospitality industry is a great example here. Travelocity sells travel products which are essentially services fulfilled by airlines, hotels etc. And then there are business who build their own product and have their own retailing structure to sell just that (e.g. Apple). Depending upon the type of supply side of business, the breadth of analytics opportunities may differ, however concept of success is consistent. Of all the prospective consumers in the marketplace, how can we get more of them to pick our products to buy? For an online retailing business, that comes down to what products, along with all its attributes, do we display to consumers in order to maximize likelihood of selection and hence transaction.
Since I have spent most of my career in travel and hospitality, I have looked at inventory more from opportunity cost perspective, resulting in a focus on pricing. In this way, the question that we ask ourselves is “how much revenue can we generate with the available inventory that we have access to?” Each option in the schematic here highlights an optimization opportunity – for example, the impact of pricing on overall transaction volume depends upon the quality of pricing within its competitive set. In a similar way, communication of value within pricing sets up some great site optimization opportunity through multivariate testing.
Execution is about the goodness of the retailing platform – website, mobile sites/applications etc. it refers to the platform through which demand and supply is brought together. Within online retailing this is web and/or mobile platform – consumer experience through the search, shop and buy process, merchandising of products, technical efficiency of the platform etc. Of course, as we all know, the website or the mobile platform itself is not the business – these are merely one of the enablers. So, as we think about the analytics plan, the plan has to be balanced for all three aspects of retailing business, and not just the execution platform – website or mobile site/application.
Now that we have a set of analytical initiatives defining our roadmap, what if we were given a target of generating 10% incremental transactions out of this. Depending upon the current state of “Outcomes” and feasibility of execution, there is more than one subset of initiatives that can take us to 10% incremental transactional growth. That subset of initiatives could all be out of the demand side of business, or could be all Supply, Execution side of business or a combination, but there is a subset associated with 10% incremental growth. The subset may not be the same for every organization, since it depends on current performance, analytical maturity and available resources, tools etc. For some, the optimization of entry pages itself may send 10% increment prospective customers down to booking path, while for some it may be the question of changing the channel mix of marketing spend.
However, for each retailing organization, the analytics program boils down to picking this set of initiatives within Demand, Supply and Execution, on way to looking for a better solution than the status quo. Instead of a target emanating from top level, we can also generate a feasible performance goal by evaluating the impact, at the set of feasible initiative level. Either way, there is a predictable path to get to performance impact on business, and that is the beginning of “Predictive Retailing”.