Jim,
Great post – there is so much stuffed in here that I hardly know where to begin. Like deciding which present to open on a particularly good Christmas morning!
Maybe here: “Real-time decisioning probably isn’t over-rated but real-time scoring is.”
I think that’s a good re-statement and, in fact, I think it’s true to the original intent of what you wrote. Real-time scoring is at the heart of most fancy personalization solutions, but because it involves some pretty daunting technical challenges, it usually ends up being both difficult and expensive. So if you’re right and it can be dispensed with in lots of cases, then that’s a potentially big win.
When I first wrote about choosing personalization tactics, I called out two main factors: cardinality and mono-multi journey. We’ve already talked about cardinality. Here’s what I wrote about journey: “After cardinality, you should consider whether your central digital problem is getting different kinds of people to complete the same journey or providing fundamentally different journey’s depending on the type of person. Think about this as Mono-vs-Multi Journey.” The discussion so far has covered cardinality, not so much journey, but neither seem central to deciding whether or not you need real-time scoring.
Why might you need real-time scoring? The obvious answer is that you need real-time scoring when visitors are likely to change their interests either during or between visits. I think stability is a good name for this attribute. Cardinality and journey are clearly going to influence stability, but they don’t strike me as fundamental. They impact stability but they aren’t identical with it.
A nice aspect of stability is that it’s easily measured. A shift in focus, whether within or across visits, is fairly simple to identify and measure provided you’ve setup a robust classification of your digital content. If you count the percentage of times you see a visitor shift focus either within or across visit boundaries, that’s a pretty good estimate of how often a personalization system without real-time scoring will be wrong (I know this isn’t strictly true – so read estimate as just that).
If stability is high, then real-time scoring shouldn’t be an issue. If it isn’t, then it’s probably essential. That simple.
Seem right to you?
I’m going to hope so – I really think we’re pretty much on the same page here, and move on to something where I’m not so sure that’s true.
I really didn’t have “creepiness” or depth of personalization in mind when I asked about where to start finding personalization opportunities. So I’m not sure I’m on board with this:
“OK…You had asked about personalization opportunities and how you go about deciding where to identify it on a website. I am going to go all lawyerly (is that a word?) on you and say it depends. And when I say it depends I mean the severity of personalization. The fact that I know you are “Gary” and you live in San Fran and you like the Giants (Lincecum is back baby!) and love analytics and may have some cursory knowledge of horse racing (wink), using all that would be creepy in a personalized message on my website, right? However, using the fact that I know you are visiting my site from San Fran and have been here before and consumed various content and originally came from organic search tells me something about how I should message you in the future, which is probably acceptable. There is a balance."
I’ll buy lawyerly as a word (it gets by spell-check anyway), but I’m not sure Lincecum is back (he’s too darn erratic), and I’m not really convinced that creepiness is as much of a factor as most people seem to think. Truth to tell, it does come up a lot. Every time I talk to folks about personalization, I get the creepy-factor question. I’m not sure I have an entirely satisfactory answer here, but I’ll loop back to something you’ve re-stated multiple times. Most companies don’t do ANY personalization. They are so far away from creepiness that the concern feels more like an excuse for inaction than a legitimate issue. Let’s face it, in the vast majority of cases, we don’t know enough to be creepy.
Let’s take your example above. Any site I visit can guess that I live in SF. If I visit them regularly, they should even be able to infer that I travel a lot. This little bit of personalization seems to escape every news portal I use – none of whom seem able to infer that I might actually be interested in the weather where I am or in local things to do when I’m not in my primary DMA. Seems like you could sell some premium inventory that way to me. None of this seems even potentially creepy.
They might infer I like the Giants (many of us SF do), but that would be an iffy guess even if it happens to be true. On the other hand, if they had 3rd party cookie segmentation they’d probably see that I’m pretty regular visitor to several leading sports sites. That would make that Giant’s guess quite a bit stronger though still no sure thing. If you know that you’re dealing with an SF-based, middle-aged guy who visits sports sites pretty frequently, choosing a picture of Timmy ending a no-hitter might be a pretty good way to capture my attention and sell your brand.
It’s hard to see how any of this, so far, is creepy. We’ve all, I think, gotten pretty used to having display ads follow us around. I’d say that about the half the ads I see are for BI and analytics solutions or travel stuff. So even though I’ve never clicked on a display ad in my life, they’ve got me pretty reasonably bracketed.
But what if a site started talking to me using a bunch of horse-racing analogies? That might be creepy. It’s creepy because I can’t make the connection between what I think they ought to know and what they actually do seem to know. Sure, I’ve visited the Hong Kong Jockey Club and Ladbrokes in the last 12 months (for work, darn it), but I’m not a regular. I don’t subscribe to the Daily Racing Form, and I’ve never registered for a handicapping seminar with Greg Wry (I actually had to look that one up).
If you have information people don’t realize you have or if that information is inherently sensitive (I may know as a matter of public record that someone’s getting divorced, for example, but I’d hesitate to ever use that knowledge overtly in a personalization effort), then I think worrying about the creepy factor may be valid enough. But where I’m just using your past behavior or core facts around your given identity to personalize your experience, I don’t think creepy is much of an issue. Most consumers expect you do to this and actually prefer that you be responsive to them.
I tend to believe that this is the only type of personalization that’s actually available to us in 99% of all cases, so I’m not all that concerned with the creep-factor.
If you’re really in doubt, take the teller test. Personalization is really about creating the same level of service at scale that you’d expect from a well-trained teller, clerk, waiter or shopping assistant. If you’d be creeped out if a suggestion came from an attentive service-worker, you’d probably be creeped out if it came from a personalization engine (and vice versa).
Agree?
Of course, I’ve excerpted a very small part of your response, which then went on describe a more tactical five step process for identifying personalization opportunities (which was my original question). I liked this part much better. I’ll summarize your steps as:
- Mashup CRM and Web data to identify potential tests
- Create basic personalization rules inferentially and with simple statistical analysis
- Create a virtuous cycle of testing
- Gradually expand the data points you look at
- Grow organically
Make sense to me.
I want to explore #1 in more depth, though. I have some thoughts around how to do this, but since this post is getting a little bit long, I’m going to wrap-up and give you a chance to respond to anything here, take your own shot at expanding #1, or raising your own issues around personalization.
Of course you did raise one last question in your post that I haven’t addressed:
“This is a shift not just in technology to enable it but in processes and organizational behavior. Maybe Kelly and I should start a series called ‘Where Personalization Went Wrong’ and interview companies that tried to do this so we can all learn from them because I believe a lot of the problems are humans not machines. Aren’t we always to blame?”
Yes, it’s true. Take any sentence of the structure “What's responsible where _____ went wrong” and the answer will always be people and, when it comes to large enterprises, the subtext will always be process and organization. It hearkens back to the old “If men were angels no government would be necessary” adage. Large organizations can’t count on having great people who solve problems despite huge obstacles. They have to find ways to make things work with the everyday people you get….well…every day.
When it comes to building a personalization capability, I think part of that means having a sound strategy for deciding what type of personalization (and supporting technology and analytics) are right for your business. I think we’ve gone a long way to showing how people can answer that question intelligently. But what always seems to come next is “Ok – now what do I personalize?” That’s obviously going to be different for every business, but it seems to me there are some processes and techniques that can help make people productive when answering that question (see, I’m subtly back to #1 above).
Oh – and one last piece of advice before I kick it back to you - place next month’s salary on Lucky Dan…
Comments