Sunday, June 3, 2012

My ‘Killer App’ For CRM

This is an idea I have mentioned to people from time to time. It can be applied to any extensible CRM system in the market and, in my opinion, it will eventually be an integral part of all CRM systems in the market. To my knowledge there is no CRM system doing this today.

So what is it? A configurable Bayesian filter.

What is a Bayesian Filter?

Many of us, directly or indirectly, use a Bayesian filter every day. Many spam filters rely on this technique. Essentially, a Bayesian filter uses past evidence of a number of situations to assess whether a current situation is a match. In the case of e-mail, the filter generally has a collection of ‘good’ e-mails (ham) and ‘bad’ e-mails (spam). By keeping a track of the words in those e-mails, it employs statistics to work out the probability whether a new e-mail is ‘ham’ or ‘spam’.

The unusual name comes from a mathematician called Thomas Bayes who formulated a specific case of the theorem used to run such filters. As is often the case in mathematics, the theory came first and the application came a lot later (in this case about 250 years later).

How Does It Apply to CRM?

Let us consider the bread and butter of CRM: sales force automation. Let us assume we have a large volume of sales opportunities going through CRM. Some of these opportunities are won and some are lost. Generally, while they are in the process of being won or lost, they are assigned a probability to help with predicting future revenues.

In the case of sales opportunities our filter will review all historical opportunities that were won (ham) and those that were lost (spam). It could then look at the fields on the opportunity record and associated customer record and, based on these, assign a  probability of success to all opportunities in the pipeline.

Not only do we eliminate the eternal problem of traditionally optimistic salespeople manually assigning probabilities which may or may not reflect reality but we may also gain insight into where our business is successful e.g. is there a certain industry where we do particularly well?

What Other Applications Are There?

Let’s say we have strong integration to our communication channels e.g. e-mail, twitter etc. Let us also assume we have the ability to convert this communication into ‘CRM stuff’ e.g. sales leads, opportunities, support tickets/cases etc. We, again, have an opportunity to apply our filter. In the case of Dynamics CRM, from Outlook, we can, at the click of a button, turn this into a sales opportunity or case. Our filter can look at all historical conversions and, based on the content of the e-mails, consider whether new e-mails should be automatically converted or not. Another application may be the automatic creation of cases and auto-routing them to the right queue, based on how previous e-mails had been routed and queued.

Why Will This Be More Important In The Future?

Firstly, as CRM use matures, businesses will be capturing more and more structured data and will want to analyse it for insights into their business. This tool allows us to review certain historical data and derive insights for our present situation.

Secondly, as social channels become more important, we will need a way to efficiently filter the ‘wheat from the chaff’. One of the skills one must learn when using Facebook is the skill of not caring if you do not read every single update in your feed. I personal struggle with the idea that I may have missed ‘the one really important update’. In the case of Facebook, the chance of an important update is vanishingly small but, for business trying to monitor the massive volumes of social data, the consequence of missing a key opportunity or complaint is much more costly making an ‘intelligent filter’ increasingly more vital.


So there is my one ‘killer app’ for CRM. If you plan to implement this, I am sure we can come to an arrangement on my cut of the waterfalls of cash that will flow from it. Ultimately though, I am keen to see it ‘in the wild’ because it opens up a world of possibilities for CRM beyond the simple capturing of data in a central location, which is very exciting.


Mitch Milam said...

I will have this ready for you next Thursday, at 4:00.

Leon Tribe said...

Neil Benson, fellow MVP and consumer of fine single malts, found this article from a number of years ago talking about the problem with estimating opportunity probability

Maulik said...

Amazing Article. Great work by you and Neil.

I think we can extend this to give suggested leads as well. So when a person opens a lead record just like a gmail account where on the right hand pane you can see the suggested ad's we can extend this to give suggested leads which will give the data from external companies like insideview or D&B. To suggest the leads it will use the same search theorem.