Ron Shevlin, who writes one of my favourite financial services blogs, quotes me today in his post on financial services innovation. He notes in my previous comments on the process of innovation that I don’t identify how, exactly, we prioritise the things we’re going to focus on.
That’s a very good question, and one that I’m delighted to answer, since I think that Ron’s post hits the nail right on the head.
We have two gating criteria.
Firstly, we ask, “can we get this up?”. In other words, if we spend some investment dollars to prove the point, will anyone care? We have some good ways to get the answer to this question. At Lloyds TSB, each product line and channel has an IT manager who has responsibility for every aspect of the IT delivery of that particular segment of the business. This individual is able to give us a quick read on whatever it is, and if they can’t tell us, they know who can. If the innovation, whatever it is, spans multiple channels, products or business lines, we just take the read on several of these managers and get a consensus. A side effect of this, by the way, is that each is tacitly signing up to support us if we decide to push things forwards.
The second thing we ask is “will this make us money?”. Now, clearly, we are prioritising innovations that implicitly have a dollar value using this approach, and that’s not going to be true of every innovation. But we’ve decided to measure our innovation work through the number of points of potential difference we drive in the group cost to income ratio a quarter. Note that point: potential. We want the pipeline to be full of things that are material, but we don’t expect that everything we find will be something we can get up.
We measure the potential return using a four year cash flow. For innovations for which we control the adoption decision, that is a simple analysis which can be done relatively quickly. Where we don’t control the adoption decision, things swiftly become much more interesting.
Our approach is to use a modified form of the well known Bass Model (described here). Our modification takes the form of a left or right shift of the s-curve depending on whether the innovation attributes (see diffusion theory) have the effect of speeding up or slowing down adoption of the innovation. We calibrate the parameters of the shift, as well as the co-efficients of innovation and imitation, using historical data from which we forecast by analogy (for example, if we are thinking about a variation in a telephony offer, we use our historical phone banking adoption data). Add cost and income data, and you have a forecast we can use for decision making. The result are not perfect, but they are good enough.
Actually, the side effect of measuring innovation in terms of hard dollars in this way is that we’re able to show hard dollar results. We report our pipeline status monthly in terms of the number of points of difference we’ve found. We only count innovations that haven’t been discounted (through the tests I’ve just described). In other words, we’ve got a set of quality things happening, any of which we think have got a chance.
Then, finally, it is easy to prioritise: choose those things that will be easiest to get up and which have the most value.