Capturing and collecting customer knowledge

Customer knowledge exists throughout your organisation, but companies often have the data in a number of seperate databases and data silos without necessarily having links between the databases or being able to connect the data sources. In addition, businesses will have information from website use, loyalty cards, CRM systems and customer service and will bring in data for marketing such as list augmentation with geodata, or capturing social media links. As a result the business can struggle to maintain a single view of the customer through all these data sources.


In the last decade, companies have been moving away from transaction-based silos of information about customers through sales, towards a more holistic view of customers and customer experiences with the customer at the centre.

This customer-centric view started with CRM systems where a customer’s transaction history could be combined with contact records and direct online information about customer accounts and service history. For large accounts in B2B markets, this internal customer data could be combined with external data such as news reports, press releases and analysis to ensure sales and service staff kept familiar with the customer’s business. This broad view of a customer was what became known as Customer Knowledge.

In consumer markets, data was more typically purely at a transaction level – what products were purchased, when and then information from loyalty cards to enable repeat purchase tracking and the effect of direct marketing to be evaluated through database analysis.

As web technologies increased it became possible to track customer journeys through a website and to combine this data with end purchase activity. Modern marketing intelligence and marketing analytics is frequently used to look at how customer journeys through a website or webapplication affect conversion and return rates. With cross-site trackers it then becomes possible to link factors like advertising exposure to eventual return and via techniques like A/B testing to link this back to purchase rates.

So where customers are identifiable even at a consumer level, via tools such as Facebook ids, or log-in details it becomes possible to track customers through their web-behaviour to both on and off-line purchase behaviour, collecting details of the webpages and topics they look at or the items they search for, the advertising they have been shown. And where customers are also active on social media, this too can be used to build an individual profile of a potential purchaser.

Full ‘deep tracking’ - from advert to purchase - is probably only really available to companies like Facebook, Google and Amazon where they can follow the entire customer journey. Other companies can capture parts of the journey, but rarely all of it and so have to fill in gaps with modelling and intuition.

Customer knowledge in consumer markets has thus moved to an attempt to capture the whole picture around customers, not just in B2B, but also for consumers, and then to apply principles of statistical analysis, text analytics and forms of artificial intelligence so as improve targeting, to track marketing performance and to model and drive customer behaviour.

In practical terms then Customer Knowledge draws from a number of strands depending on the situation. In B2B markets Customer Knowledge is the collection and use of information to support account management and sales activity (account acquisition, bid management, and B2B marketing).

In consumer markets, within a telephone sales context, Customer Knowledge is still related more towards CRM systems and contact management and tracking that also ties in purchase and service history and live notes on customers. For retailers, Customer Knowledge builds on loyalty cards. In both situations, as customers move to multiple contact methods including online ordering Customer Knowledge becomes more holistic, though it is still potentially patchy compared to the pure web-based companies like Amazon.

Thus for those operating on the internet, Customer Knowledge is the collection of online information that is available on customers and the process of building this knowledge by combining and drawing on multiple sources of information – for instance combining online web-tracking data for identifiable customers with offline data such as geodatabases or customer profiles.

As with all these approaches, the aim is to try to find out what customers are looking to buy and to prompt for those purchases ahead of competitors. For instance, an insurance comparison company might use details of customer’s location to prompt with offers that have been popular for other customers in the same location.

With larger and more sophisticated databases the ability to target much more finely and specifically becomes possible, but with the challenge that there then end up with too many segments to be managed by hand. In these situations the only available approach is to start to market and monitor on a statistical or algorithmic basis with the algorithm choosing to present the offer most likely to win the customer’s business given what is known about the customer.


Collecting customer knowledge

From the description of customer knowledge and the different viewpoints, it should be clear that each potential data source and business type will have its own different approach to collection of customer knowledge.

The most basic level is pulling information together from transaction databases, specifically where data is say organised by product purchased or invoice in an accounts system, in a customer knowledge system this data has to be pivoted to reflect the transaction history per customer. Where a customer is one person this is straightforward, but where the customer has multiple accounts or, in B2B terms, a customer may be multiple people or sites, the pivot needs to include elements of data cleaning and deduping.

For sales and account managers, on top of this transaction view, they want to be able to see and track contact and conversations with the customer. With CRM systems it becomes a method of collecting specific information in the customer scripts and recording conversation topics, dates and frequency so as to tailor a sales approach to the customer. This can then be combined with external data providing a sales-based view of the customer.

In B2B markets where customer relationships are more complex, the CRM systems will contain details of the breadth of the customer relationship – multiple contacts and locations together with corporate structure, purchase unit structure and information on the background of the customer. This can then be supplemented by external information about the customer in qualitative terms like news, financial reports and press releases which can be mined on a case-by-case basis to help the sales team in their account management.

These views are mostly a view per customer, so the next stage is to pull the dataset together into a combined customer knowledge analysis database. In this way links between sales behaviour, transaction histories and contact behaviour can be mined to allow more focused sales activity that is more likely to win orders or retain customers. Qualitative factors like financial reports might need to be turned into clean quantitative data, and external data sources might include geocoding, credit score information and information about number of employees, end of year date etc to see if this drives sales.

With person-to-person sales these databases are typically reflect several thousands of customers with a depth in terms of the number of transactions and conversations plus simple demo- or firmo-graphics and draw on information from CRM (eg Salesforce) supplemented with external data sources or, in B2B markets, links to services like Hoovers or credit scoring.

For businesses with a wide reach (eg retailers), the customer databases can reach several million records with deep transaction histories from data collection via loyalty cards. These would be relatively structured databases and suitable for modelling and then data extraction for direct marketing. These type of very big databases have much less in the way of qualitative or unstructured data.

For customer knowledge in an internet world, the amount of potential information explodes. However, the difficulty is matching visitors to actual individuals and tracking a full transaction history over multiple interactions and over multiple devices. In practice much web-analysis and analytics is done at an anonymous level where the individual identity of the visitor is not known nor linked to external information.

However, as customers identify themselves online it becomes possible to link individuals to specific visitor journeys. For small companies this type of customer knowledge can be hard to obtain because it relies on someone logging in, or completing a purchase (at which point they identify themselves) in order to work back through the data to see the customer journey. Instead, visitors are tracked using cookies but without external data.

By contrast larger companies where customers are continually logged in or where trackers are linked to identifiable individuals, can have information about the customer from the first page including linking to external or identifiable information. This allows for deeper targeting and is the preserve of the larger internet companies and advertising firms. Customers, their purchase history, visit behaviour, keywords, location and even social graph can be collected and used for targeting – and this could be supplemented with external data, though in practice we don’t believe this happens. With such depth of information it becomes possible to have offers prepared for specific individuals that are then presented when those individuals come online. However, the entire system needs to be developed and delivered automatically as the data becomes too big and dispersed for human scale analysis.

The current state of customer knowledge collection therefore splits between anonymous or pseudo-anonymous website tracking data, and much smaller scale knowledge systems for thousands or tens of thousands of customers. Only the largest online companies have the depth of individual level data that can really link online and offline views of the customer together. As time moves forwards, possibilities of adding data include text-mining and social media monitoring including face-recognition fused with customer records or web-journey tracking.

This ‘deep profiling’ raises obvious issues about privacy and manipulation – for instance using customer knowledge to set higher prices for certain customers. In Europe at least, deep data brings with it a requirement for explicit consent and good data practices.

Allowing for change

It will be clear that initial steps to build customer knowledge platforms into the business can be demanding. However, it is important than any systems and approaches are also able to flex and adjust to changing product, customer or transaction requirements. Flexibility and the ability to adapt are thus core elements in a customer knowledge system with suitable API links and connections to allow the system to keep growing and changing to market and business needs. A single customer view now also needs to be a single customer view in a few years, when circumstances have moved on.


For help and advice on building customer, competitor or marketing knowledge systems contact info@dobney.com