OK, so what is an Event?

Well, it has taken a lot longer than I anticipated getting this blog rolling, but now in the relative calm of the post Christmas break, I thought I would sit down with the last mince pie in one hand and try to set out some thoughts and ideas on Events and Event Driven Marketing.

As stated in the objectives blog, we are going to cover all sorts of topics around EDM. In this one I would like to start by defining what I mean by an ‘Event’ (at least IMHO).

An ‘Event’ is defined as a change in an Individual’s circumstances today which is significant, either in fact or in their mind. The three key words here are Significance, Individual and today.

Nice words. Let’s look at this in a little more detail.

Significance

 

Significance is the act of determining if a particular action displays a noteworthy change in a customer’s behaviour or interaction pattern with the Bank. Ask the question “is this action significant for this customer?” Of course to know that, I would need to know more about the customer, their individual transaction history, etc.

A guard of honor passes out as Queen Elizabeth II rides past during the trooping the color parade, 1970

A guard of honor passes out as Queen Elizabeth II rides past during the trooping the color parade, 1970

Let’s take the example of a Large Deposit. Many systems work by setting triggers such as for a deposit of more than €10,000. The pseudo code typically looks something like this:

IF TXN_AMOUNT > 10,000 THEN {SUCCESS}

 

But this deposit may or may not be significant. Let’s look at two imaginary customers: Bill Gates and Bill Doors who have both deposited €10,000. For Bill Gates this may be a common occurrence and he is just paying his Amex bill. For Bill Doors it may be an unexpected windfall, which will allow him to put a deposit on a house or car. Thus this would be significant for one customer but not the other. In order to determine this, we need to check the deposit against the customer’s previous activity over a period. A simplified example checking significance might look something like this:

IF TXN_AMOUNT > 10,000 AND
TXN_AMOUNT > LARGEST_PREVIOUS_DEPOSIT {WITHIN LAST Y MONTHS} MULTIPLIED BY X {EG. 50% HIGHER}
THEN {SUCCESS}

 

So here we not only check that the transaction is greater than €10,000 but we also compare to see if it is X% greater then any previous large deposits in the last Y months. This is quite simplistic, but you can already see the amount of extra processing involved in doing this.

Let me share an anecdote to highlight the difference that Significance makes, eventricity was once told by a Bank that they were considering using us or a competitor to implement EDM. We were informed that the other vendor had a wonderful, sexy tool and a whole library of Events that Marketing could use. But the Bank wanted to assess the real value and so we were both asked to conduct a benchmark trial of our respective Large Deposit Events using real bank data from 1m customers.

Our competitor’s Event trigger generated 606,000 leads in a month whereas our Event produced 36,000 leads. Even more interestingly, when followed up the trigger leads generated less than a 0.5% positive response rate whereas ours returned rates of 18%-24%.

Thus Event triggers are very easy to set up and require little processing. They generate more volume and achieve a lower response rate. Events that check significance are more complex and require more processing to generate fewer leads with a higher value and response rates. That’s the difference between Spam and Quality. It is the difference between Significance and no significance.

Individual

 

Events are assessed against individuals not groups of people or segments. If we use a segment approach, both Bill Gates and Bill Doors are the same. If we analyse them individually, we might determine that a deposit of €4,000 would be significant for Bill Doors whereas this would have to be more than €245,000 to be relevant for Bill Gates.

blue_and_yellow

Following on from the example above, if we simply set the target amount for a segment at greater than €10,000 then we would catch a Bill Gates deposit of €20,000 (incorrectly) and we would miss out on customers such as Bill Doors when he deposits €9,000 yet this is actually significant to him and more than double his previous deposits.

To address this, our pseudo code should look something more like this.

CUSTOMER_SIGNIFICANT_AMOUNT = LARGEST DEPOSIT OVER THE LAST X MONTHS * Y (eg. 150%.)
IF TXN_AMOUNT > CUSTOMER_SIGNIFICANT_AMOUNT AND
TXN_AMOUNT > MINIMUM_DEPOSIT_AMOUNT
THEN {SUCCESS}

 

So now rather than simply checking if the transaction was greater than €10,000 we first determine what a significant amount is for each individual compared to their specific history. We also make sure that this is above a minimum amount. With this approach we can now find correctly the €9,000 deposit from Bill Doors and ignore the €20,000 from Bill Gates

Today

 

Once a significant Event has been detected the timeliness of customer communication is crucial. In fact, if the bank is not able to communicate with a customer within 48 hours of detecting an Event, they probably should not bother. Research has shown that the customer response rate decreases by about 66% for each 24 hours. An average response rate of 70% for contact within 24 hours is common. This drops to around 25% within 48 hours and less than 10% in 72 hours.

This means that I need to be able to do all of the above at least every day.

Interestingly, one source of the figures above was a Bank that had implemented a Large Deposit Event, but who were loading the Datawarehouse on a monthly basis. Their response rates for the Event starting on the first of the month were 70%, 25%, 10%, 5% and then 2% for the next 26 days until they loaded the next month’s data. This single fact was enough for them to set up a project to implement daily data loading.

Summary

 

In summary, a proper ‘Event’ asks the questions, is this action significant? For this individual? Today?

Compare the above to the way that many banks do analysis. Mostly, data is loaded monthly and analysis then takes a further week. Leads are therefore generated based on a change in a customer’s behaviour from the last data loading period (usually 5-9 weeks ago). The analysis is done against a background of a segment, meaning that levels are assessed against the segment norms, not those of the individual.

I think the takeaway from this is that not all ‘Events’ are equal. It may well be harder to define, develop and deliver Significant Events, but they provide significantly better results.

ferrari1

Remember, simply painting a Lada bright red does not make it a Ferrari. In the same way, simply calling something an ‘Event’ doesn’t mean you will get the same results. Many Event Driven Marketing solutions fail the previous criteria. More importantly they don’t deliver the level of results expected or achieved by people using real Events. So caveat emptor, and do your research on what is being offered.

Finally, There are many significant Events in a customer’s life. But it should be noted that Events are not just for Sales opportunities and these techniques can also be used to find many kinds of significant occurrences such as Attrition, Credit Risk, Anti-money Laundering and Service. Examples can include not only a large deposit but withdrawal of money, redundancy, changing payments, moving house, marriage, divorce, travel, a new job, first child, etc.

As background to this blog, you are welcome to check out details of results and KPIs published for different types of marketing here

No Comments

Post A Comment