Most B2B marketers would say they know their customers. They understand their needs and motivations and use this to target their offering to them. But it’s not always about what you know. Examining what you don’t know and comparing this to what you do know can bring the view of your customers into sharper focus and create the basis for a more profitable relationship.
Confused? It’s the premise of one of the most powerful modelling and targeting techniques propensity modelling.
By using propensity modelling, marketers are able to take what they know about their customers and analyse it against prospects they don’t know. This enables them to refine their understanding of existing customers’ needs and wants and find other, similar prospects to target. It can be used to identify which characteristics of a prospect or customer are predictive of high take-up rates, low service costs, high rates of customer retention and low levels of debt in order to target resources more effectively.
In short, it allows marketers to pinpoint what a business need is for a product or service and the value it might represent. It’s a ‘rifle shooting’ rather than ‘spray and pray’ approach and enables the focusing of commercial resources on high profitability customers and prospects.
Target practice
So how are high value customers and prospects identified? The starting point is to take a customer base of known users of a product or service, and analyse this compared to a universe of non-users, scoring against as many pertinent predictors as there are available. This gives a score on the needs a business with certain characteristics has for the product.
For example, as a first step, a company car fleet provider would analyse users of company cars against those who don’t have company cars, using the basic predictors of industry sector and size of business. These indicators are predictive as it might find it penetrates one industry sector for example, clothing retail at 10 per cent and another building and construction at 30 per cent. This penetration rate could be used as an indicator of their propensity to buy fleet cars. Clothing retailers tend to stay closer to home and have less use for fleet cars compared to a building products company with a direct sales force. The next step would be to look at the value of spend amongst building products companies to pinpoint who might be more interested in a Mercedes than a cheaper car.
Financial variables and measures of competitors’ performance will distinguish a company that can afford new cars from one that can’t. Measures such as growth are also predictors of immediate appetite. A third area of predictability comes from behavioural variables age of the business, international activity and group structure.
Customised modelling
Of course, all this is hypothesis, but examination of the customer database versus the universe will confirm, or otherwise, its validity.
There are generic propensity models on the market such as pre-calculated scores for propensity to buy a fleet. These have a place, but they won’t distinguish Mercedes campaigns from mass market ones. So generic models only take you so far in terms of how targeted they are. It’s half way between ‘spray and pray’ and a ‘rifle shot’ approach.
Some providers have developed automated propensity modelling and it may sound automatable, even based on specific customer data, but again, there is a gap compared to customised modelling.
B2B data sets, unlike B2C, are more disparate and difficult to run automated processes from. So some automated systems stay at the level of base camp. A machine can’t sort out the apples from the oranges, it’s just not intuitive. It may fail to identify who uses company credit cards because it did not distinguish between who uses them as an additional line of credit, as opposed to using them for travel. These businesses would have opposite profiles to each other and the automated system would average them out. Misreading the data could result in missing golden opportunities.
Finally, automated systems don’t have a history of data. But to find similar clients, it’s important to know what a business looked like before it became a customer.
Customised propensity modelling doesn’t just make the most of what a business knows, it’s about exploiting the unknown to help it find prospects and target existing customers more effectively. The cost involved in generating a customised model will be far outweighed by the cost of deploying the sales and marketing assets of the business in a less efficient way. In the current economic environment, this is highly relevant.
