Getting started with data analysis
While large scale data analysis increasingly relies on machine learning and automated approaches to extracting and applying insight from large datasets, businesses with smaller information sets can still get a great deal from a systematic approach to looking at internal data.
Companies have data about customers, sales and suppliers but often don't make the most of that information in terms of monitoring business performance or improving operational performance. Strong businesses keep good records, but also use those records to understand and improve.
A first barrier can be data extraction. Information sits in multiple databases and data silos - contact information, newsletter sign up, sales team appointments, orders and delivery notes. For one off projects the data can be pooled by working through each data source and tying the information together around customer name or contact name - often needing secondary information like phone numbers or email addresses to ensure data is matched correctly.
For continuous monitoring, businesses will need to standardise and centralise records. Tools like Salesforce and other CRM systems are designed to help with this process. However, this is not yet analysis.
Analysis starts with simple information. How many (unique) customers do we have? How many have bought in the last year/quarter? What's the average sales value? What are the most purchased products? How frequently do they purchase - seasonality? These themes of who?, what?, how much?, what value?, what frequency? recur through analysis as it gets deeper - recency, frequency, volume. In drilling down, the aim is to identify changes and differences from the norm, and to understand why and how these differences might occur and so to spot opportunities to increase sales, purchase frequency, or purchase value.
The first comparison analysis is to split the customers into segments. Who are the biggest customers, and who are standard customers - using a decile analysis (splitting the customer base into equal sized 10ths), or pareto analysis - comparing the 20% who are the largest customers to the 80% remaining.
This can be extended with geographic analysis - where are the biggest customers? Or the most frequent purchasers. Is this related to sales teams, or distribution?
Secondly is to look at time. How does customer activity vary by month? Are large customers different from smaller customers? Are there regular buying cycles? What customers are increasing purchasing, and which are reducing buying? Do the sales team know why?
With this base data, how does marketing and sales activity sit on top of this? What uplift does an advertising or marketing campaign give against normal trends. If possibly, businesses will run test and control based marketing tests to see what uplift marketing will have before rolling it out more widely?
Are there connections between say, outbound calls, and sales conversion? What about customers calling in with service queries? Monitoring website visits or social media can be difficult to connect to direct customer activity, but adds a potential additional layer to tracking how marketing activity moves customer through the sales funnel.
Within the data, are there links between buying of one product, with another. Sales of bikes might run with sales of bike helmets - can this be used to create bundles, or be encouraged to increase sales value? Sifting through this level of cross-information combined with segmentation and time-based data is why much data analysis is now highly automated looking for segments and creating likelihood models.
The data can also be augmented with external information. For businesses, this can include company financial information, or business news. For consumers, information about geography or supplemented with additional information about habits can be available, depending on GDPR and data privacy rules.
While all this takes into consideration data the company has to hand, businesses rarely have a complete picture from their internal data. Prospects and non-customers are missing from the dataset, and the internal data itself is backwards looking - it reflects customers in the past. If the business creates new products, or wants to win new customers, it needs to think ahead of and outside its customer base. For this reason, database analysis requires the perspective of fuller market studies.
The business can also become so obsessed with the data, that it forgets the individuals. When a customer calls in to a support line, they aren't just a data point, but someone with a personal and unique perspective. They could be the first person to spot an issue with a new product, or someone who has an insight that would change the business. While the aggregate helps with overall management, the details also matter for case-by-case delivery of service.
For help and advice on starting with turning data into insight contact info@dobney.com
