Descriptive customer or marketing analytics are a way of linking the customer or the market to the business through information or data, specifically the kinds of useful data the business will require to allow it to make effective actionable decisions. The focus of descriptive analytics is always on what happened (and why it happened), and on the behaviour of the customer or market.
There are three kinds of descriptive analytics, each being used in a different kind of situation, or where the kind of decision management wishes to make varies. The three kinds of decision-making scenarios which use descriptive analytics are as follows:
1. Exploratory:
When dealing with some ambiguity or uncertainty, you know there’s a problem but you’re just not sure why, e.g. sales declined year-on-year for this particular marketing campaign or catalogue, or sales didn’t meet budget expectation; why was that? Exploratory research can be done to develop a broader understanding of the particular problem, and to hone in on the key area(s) to be tested more thoroughly. It often starts with a hunch and proceeds from there. Given there may be a number of possibilities why something has happened, this is often the only way we can start, with a view to eliminating some possible causes and narrowing things down as we go deeper.
Traditional examples of exploratory research are questionnaires, surveys and focus groups. In the digital age, these have moved online to such services as online discussion groups and forums. Other information is also available from social media or blog comment streams. The manual nature of monitoring these information-rich but highly unstructured environments has proven a problem. Formal online monitoring services take a long time and the cost-benefit of using them are hard to determine up front; you may not get anything useful at all to show for your time and spend. But new techniques in web data mining (using complex computer programming and machine learning) are beginning to allow companies to harvest this rich data more efficiently.
2. Problem-solving:
Where there is some awareness or certainty around the problem, but more research and information is needed to provide hard data to describe the extent and nature of the problem. Usually the data generated is hard numbers and statistics which describe the make-up and characteristics of a particular group of customers, e.g. who are our customers; how are they segmented; how much are they spending on our products vs. our competitors’ products; how much of their total disposable income are they spending with us, and so on.
There are two ways of collecting such hard data: actively through surveys, and passively by observing and recording customer behaviour (transactional data).
Active data collection here includes customer surveys, generating hard data about such things as: customer attitudes to the business and its products; product and customer service satisfaction scores; purchasing habits. Surveys can also be used to gather segmentation data: who are our customers, what are their interests or their main pain points (“What is your number one concern right now?”), who do they buy from, what is their likelihood to buy a proposed new product, and so on. Modern advances in customer surveys include mobile app surveys, and in-situ surveys (where customers are asked immediately about that purchase at the moment of sale). Surveys can be conducted in-house (e.g. SurveyMonkey) or conducted by third-party providers.
A key metric to survey is the Net Promoter Score (NPS), “How likely are you to recommend (company/product) to a friend or colleague?” scored on a scale of 0 (not at all likely) to 10 (extremely likely). The NPS is calculated as the % who are Promoters less the % who are Detractors, where Promoters score 9-10, Passives score 7-8 and Detractors score 0-6. A high score (up to 100) is very good but an extremely poor score is down to -100 (where most/all responders are Detractors). The NPS is a leading growth indicator, since word-of-mouth recommendations are one of the greatest drivers of organic sales growth yet come at no/minimal additional cost to the business. Tracked over time, this metric shows how the health of your brand is faring among your customers. Hard data can also be obtained from data service providers such as InfoScout which tracks and provides USA consumer shopping behaviour.
Passive data collection such as transactional data will be collected in the back-end and arise as a direct result of the customer transacting with the company, or interacting with it in some other way, such as visits to a website, etc.. Order data should be systematically collected, including number, frequency and value of order, products and product categories purchased, and so on. Passive data collection can occur through a variety of channels (web, mobile, phone, customer service helpdesk, marketing media coupons, retail point-of-sales). Third party companies also provide a data service in relation to passive data (e.g. point-of-sales retail data available from firms such as Nielsen or IRI).
Such passively-collected hard data – if systems are well-designed and systematically and consistently applied – will be complete, accurate and timely (often real-time) allowing for effective managerial decision-making across the three key areas of the business: optimising sales, improving profitability and reducing risk.
The kinds of problems businesses may use internally-collected hard data to understand and address include:
- How did customers respond to specific marketing offers or product promotions?
- What impact did display of products have on sales (e.g. in catalogues, online)?
- How do sales of products and product categories work together – are they complementing one another or cannibalising one another’s sales?
- How did first time buyers find us?
- How often did customers abandon a purchase and why?
The kinds of problems businesses may use externally-sourced hard data to understand and address include:
- How does our brand compare to our competitors’ in the eyes of the customer or target market?
- How do our customers’ buying habits change online compared to in shops or via mail order?
- Which listed products are customers or target market buying from competitors instead of from us?
- How often are customers talking about their purchase of our products, and who to?
- How often is our brand being mentioned in customer and target market conversations?
Point-of-sales data remains the most valuable hard data: what customers are buying, where/how are they buying, what triggered the purchase, what do they have in their basket, etc.. Hard data (both actively and passively collected) will have its limitations though. For example, you won’t know all of the customers behaviours and psychographics, nor the full list of competitors’ options under review at time of purchase. Nor can you use such data to perform causal research (what will happen to variable x if we change variable y by so much). The data you have about a particular promotion will show what people bought at what price, but not what you can do in your next promotion to improve response rates, order values or revenue levels.
Media engagement data focusses on data collected internally or through third parties to establish audience engagement via different media. For example, radio or TV audience levels and popularity by show or by channel, or newspaper/magazine audiences and circulations. Such engagement data is useful to know in order to maximise targeted advertising spend through those media. A modern approach will follow the same principles but focussed on online social media engagement data: what customers and prospects are saying and doing online and who they’re interacting with. Companies like HootSuite or SproutSocial provide access to such data. Social media can be monitored to see audience engagement and participation for a particular promotion, mentions of the company’s brand (tracked over time, and compared to mentions of competitors’ brand mentions), and for sentiment analysis (qualitative views rather than quantitative – positive vs. negative mentions).
A further category of online data is web data, including website visits, online searches, etc. – both on the company’s own website, in the broader marketplace and on the internet as a whole. An example would be to track unique visitors, page-views and website engagement metrics for the company’s website as compared to competitors. Insights can also be gained by reviewing what products/trends are being searched for on other websites (in search engines, or on site such as Amazon, Etsy, Facebook, etc.). Web data should be collected for (1) The company’s own digital and web properties, (2) Online sharing and engagement through social media, and (3) Online advertising activities and their effectiveness. Activity in one area should impact on the other two areas, and each area can be expected to cross-fertilise and improve results in the other areas.
Again, advanced modern methods such as powerful computing algorithms and machine learning can be leveraged to track and analyse social media and other web-based data to provide useful additional consumer insights and competitive advantage.
3. Causal:
Where a problem is well-known and clearly defined, consisting of one specific factor or variable which, when changed, will have a direct impact on the outcome of another variable (e.g. if we increase the number of customers we mail a marketing campaign or catalogue to, how will that affect the results received – will it increase response rates and therefore sales or not?)
It should be obvious that there are a vast array of analytics which can be monitored and evaluated, especially those falling within the causal category. For example: if we change the headline on our pay-per-click ad copy, will it increase the click-through rate; if we make a change or improvement to our landing page, will that improve conversion rates; if we change the marketing proposition or offer (e.g. discount rate) will that lead to increased numbers of units sold and hence increase sales revenue, or will the discount erode our revenue even despite an increased conversion rate.
Other kinds of questions businesses might seek to answer using causal research are:
- What impact is pricing and discounting having on sales?
- What products/offers will optimise new customer sales?
- Which customer segment is most likely to lapse and why did they stopped buying?
- Why did first time buyers purchase from us?
- Why are customers abandoning their shopping cart and what can we do convert these back to a sale?
Given the large number of highly-correlated variables (which may or may not also be causal variables) seen in a digital or direct-response marketing-led business, the data collection and analysis methods the business employs should be carefully and systematically designed and implemented. In particular, they should be strictly controlled to allow collection of compelling empirical proof of causal relationships between two isolated variables. An example will be a split test (A/B test) conducted on a sales letter or the ad headline on a landing page to see which one produces the best conversion rate. The tests should be designed such that only the two variables of interest can change, with all other variables being kept constant. As one of the two monitored variables is manipulated, it’s effect (or otherwise) on the second variable can then be measured, evaluated and optimised.
A word on causal relationships. In order for there to be a causal relationship between two variables, all three of the following conditions must be met. (1) There is evidence of some correlation or association between the two variables, that is the two variables move together in some unspecified correlated manner (e.g. price and sales revenue, where price falls sales revenue also falls). (2) The change in variable x occurs before the change variable y, also known as temporal antecedence (e.g. the price changed first and only after that did the sales change). (3) There could be no third variable which was driving the change in both the two variables under review, all other possible variables have been held/made to remain constant.
The best way to establish the difference between causal and correlated relationships in any business (especially those which are digital/DR-driven businesses) is to systematically and consistently test, retest and test again, all the different variables in that business, through a regular program of split testing. The kinds of variables which should be systematically manipulated as part of this ongoing program of causal research include: prices, discount offers, product-led promotions, landing pages, ad wording variations, sales copy variations, icon choices, image placements, font size, checkout processes and pages, and a whole host of other optimisations. The same principles also apply to allow optimisation of mobile sites and mobile apps, and a wide variety of firms are geared to helping companies automate their split testing research (e.g. Optimizely, Google Analytics).
Such optimisation can be extended to cover not just one website but also different websites optimised for different customer segments. As tracking becomes more advanced, customer segments can be targeted more and more specifically, leading to greater and greater personalisation of the website shown to an individual customer (think of Amazon recommendations driven by the individual customer’s past purchases and known interests/browsing behaviours).
The digital data revolution:
The benefit for digital or direct-response marketing businesses is clear. While all the above retail and sales data may be available to traditional retailers via third-party data/analytics companies (at a price), a modern digital/DR business is able to systematically collect and analyse this data themselves on a real-time basis – provided they have the systems set-up to allow this and the managerial culture to promote it.
The digital/DR company which embraces the data culture – investing in the systems and skills needed to enable that culture to thrive – and makes it a strategic goal to prioritise the consistent measuring, monitoring and evaluation of their customers’ and prospects’ behaviour, will be well-placed to overtake their more backward-looking and technophobic competitors.
The digital data revolution is upon us.