Insight
10 questions you should be looking to answer with your data
February 6, 2018
A good data analyst should always be asking questions, digging into the stats and asking ‘So what?’ of any findings to try and come to a genuine insight. But it can be hard to get to the nub of the problem if your data setup isn’t telling you the full story.
By going back to basics and exploring how and why you’re really using your data, you can start to make some strategic choices about the campaigns you run, the products you promote and the customer segments you target.
Knowing your bounce rate isn’t enough. Asking deeper questions that focus on the customer journey will enable you to structure your data in a meaningful way.
So what are the big business questions you need to be looking to answer to improve performance and profitability?
– Are products with a low customer rating more likely to be returned?
– What impact does poor or average ratings have on sales?
– Does seeing a ‘low stock’ indicator impact customer behaviour?
– Which customer segments have the highest cost to serve?
You might have noticed that these questions all have one thing in common: They’re linking up data sources that tend to sit within different tools or parts of the business.
Returns data, ratings data, media spend and stock data are usually all housed in silos, and this means it can be hard to make connections and put the information in context.
Although joining up data sources isn’t necessarily a new concept, linking this data with your web analytics is rarely done and can really drive business performance if the data’s used and analysed correctly.
For instance, we might have identified a VIP customer segment with a high repeat purchase rate and average order value in Google Analytics. But it isn’t until we link the returns data that we see that average return rate for some members of this segment is at 90%. This helps us identify a new segment who have a very high cost to serve (and are obviously not quite the VIP’s we initially thought).
Linking these sources together is something that can be done within Google Analytics or through BigQuery, a cloud-based storage and mining tool that can be used to store and query hit level data.
– Do users purchase more when it’s raining outside?
– Are your users more likely to buy if they’ve read three reviews on other sites?
– What is the socio-demographic make-up of our most loyal customers?
First-party data is the data you have collected yourself about your audience, second-party data is the data a partner or other organisation has collected, whereas third-party data is large data about huge sections of audience – this is usually great for demographic, behavioural and contextual marketing. Like your other data sources, third-party data allows you to apply further filters and segments to your total data set, enabling you to undertake deeper analysis of your customer base and personalise your marketing accordingly.
You can buy integrations to public data sets such as ComScore, which has pixels on most sites. This allows you to further segment by, say, audiences interest, e.g. those users who have also visited three review sites are more likely to purchase.
With one of our clients, Hillarys, we used Experian mosaic codes, a geo-demographic segmentation system. We have mapped each of the 61 main mosaic types to each of Hillarys customer segments. This allowed us to personalise the content offer and create tailored homepage content per segment.
Weather API’s are another simple way to use third-party data to inform tailored homepage content. If you are a travel company, you may choose to display warm weather locations on your homepage to users that are in cold locations, for instance. Or if you are a paint manufacturer, you may choose to lead with an interior paint message on your homepage if it is raining and an exterior paint message if it’s a beautiful day.
– Are people more likely to sign up to your newsletter after reviewing 6 pages?
– Do users who land on a product page look at more products and have a higher AOV
– Do 45- to 54-year old users who visit via email transact with more products in their baskets?
These questions can be answered through predictive analysis, which takes a list of users and their behaviours and then predicts, based on the past and present behaviours exhibited, how likely they are to purchase.
We can define what those behaviours are (for example, ‘product view’, ‘sign up to newsletter’, ‘add an item to their wishlist’), and then we can see how much weighting each of those behaviours has in terms of how likely someone is to purchase.
Predictive analysis can be extremely powerful especially in the use of re-marketing to users who are predicted to have a higher propensity to purchase; by using these tools, you can be much more efficient with your advertising budget.
Google are already doing this through their Session Quality data to evaluate user engagement for each session as well as the likelihood to convert.
A more advanced use of predictive analaysis is the opportunity for real time personalization. For instance, if someone views more than three products do we personlise the site in real time to cater for them? We can use mosaic codes in this instance to, so for instance, with Hillarys Blinds, if we know a user lives in an area with lots of terraced houses, do we show them creative for that house type?
Hopefully we’ve inspired you to go back and look at your data with a fresh set of eyes. These are only a small set of examples, but if you’ve got other questions around your data that you’d like help answering, then get in touch: hello@codecomputerlove.com.
