Customer knowledge analysis
Having gathered a large amount of customer knowledge, it is important that you are able to perform analysis across the customer base to help you make better decisions, and to identify gaps in your knowledge. There are two types of information in customer knowledge databases, unstructured news style reports and more formal structured data such as questionnaires, tracking data or other forms of sales or accounts-type database data.
Approaches to structured information
Companies have structured customer knowledge in the form of databases monitoring transactions, contact records, questionnaires and surveys (which although anonymous, provide an overview of the customer base), websites tracking and sales management systems.
From the structured data, a straightforward first analysis is simply to investigate whether your customer's revenue is increasing, static or declining. If you combine this with an overview of the strategic value of each customer, you can start to identify key customers and key segments along similar lines to the well known Boston Consulting Group (BCG) portfolio matrix.
Classic forms of analysis of structured data like recency, frequency and volume, and customer journey/experience modelling as might be used for web-site analysis can also be used with customer knowledge.
Ideally, what you are looking for from structured data are opportunities to group and harmonise customers together who have similar needs allowing the business to develop more streamlined operations which meet the needs of a large enough group of customers, while avoiding cutting corners and service for those customers who are most important to you. The streamlined operations may include more flexible products and production, increased and better controlled communication looking for cross-sale and up-sale opportunities.
However, you may also want to look at individual cases, looking for accounts worthy of extra care and attention. A critical question is how much you are willing to invest in each account based on the return that you hope you might get.
Analysing ad hoc data
Ad hoc data, by its nature has the potential to be extremely rich in content, but all too often it is dealt with on a case-by-case basis, often only by the account manager. The value in ad hoc data comes from uncovering common truths and wider opportunities across the customer base which signal the potential for new products and services.
Ad hoc, or unstructured data can be analysed only by going through the data, pulling out key trends or valuable information. Tools such as a search engine, text analytics or keyword filters help the analyst extract the relevant information but it can be useful to try and build pictures too, using ideas such as knowledge maps and case study approaches that allow a range of information to be seen together.
For an individual customer, the most common method of analysing ad hoc data is a simple historic approach, putting the information in a chronological order and looking for the underlying story. This is made far easier by having a centralised means of collecting and sharing the information as patterns occur that could not be seen with single or partial pieces of information.
However, there are other approaches to analysis. For instance, you can also take a diary approach - looking for key dates and moments that occur regularly. The simplest is just having knowledge of a customer's accounting year, but they may also launch products at a particular time or with a particular interval for example.
Another approach is to look at the customer's network of relationships - who are their strategic partners, customers and competitors. Can you plot and spot how these relationships work from the ad hoc data that you have to hand. Simple examples of how this might be used include helping your customer keep track on their competition, or partnering with one of your customer's other suppliers so you both increase your value to the customer.
You might also consider looking at the stream of communications by volume and type to identify recurring themes within the relationship and so identify latent needs or areas that have not been covered in other ways.
For analysis of unstructured data in bulk, more algorithmic tools are needed. For structured data like web-site journeys or click-through data that can be tied to sales or other structured data, tools like regression, model building or bayesian learning techniques can be used to find relationships within the data.
In bulk, unstructured data needs to be tamed before full analysis - for instance using text analytics to extract keywords or apparent sentiments from the data before trying to relate this to other data. This might also include 'coding' the content of the unstructured data to try to capture meaning in a more effective manner. The volume of data may make it difficult or impractical to code up all the unstructured information available, but it is also possible to do this type of coding on samples of the data to test and builds model. As more data becomes available the number of potential data combinations rises exponentially, as does the chance of false flags so automated learning may be more effective than individual level analysis.
One problem with the potential volume of data is that it is all historic data (backwards looking). A more effective way to use a Big Data stream, may actually be to take a more experimental approach. Rather than looking for relationships, different ideas are tested to see how the data stream responds to the ideas. These tests, particularly online, can be small scale without affecting the wide customer base and then rolled out when key successes are found.
It's also extremely important to realise that customer-facing staff are often identifying patterns and needs among the customers they see, without needing to go to the database. Though these views may be anecdotal, people talking to people may actually provide the best initial insight or guess as to what is going on. So an effective way of leveraging and exploring the ad hoc information your business has is in the form of workshops where customer facing staff are invited to talk about what they think customers want.
Where the ad hoc information does start to reveal useful information, then this is when more formal and structured approaches can be used to assess the value of this data.
For help and advice on building customer, competitor or marketing knowledge systems contact email@example.com