I have been thinking about this for some time (BTW, this is a short post – hope to ignite some discussion with this), would love to get your thoughts.
Traditionally (as in most everybody I know in this world) we use financial metrics for customer segmentation, right? Either lifetime-spend, latest-spend, last-year-spend, or profitability, or what-not. We may use other aspects of segmentation for marketing (like demographics, products purchased, support requested, etc.) depending on what someone in someplace decided it would help us find the “right” people to buy our product (usually that someone was a marketer, or a focus group, or an MR firm we hired – or something altogether).
This has — well, worked for us until now since we have been focused on the shotgun approach to marketing and sales for the most (yes, I know that means anybody but you who have done a masterful job of hand-selecting your clients, unlike your competition). That is, we loaded the shotgun with pellets, shot in a general direction (segment) and some hit and some missed, if we did a decent job of segmenting we shot into a bush with tons of birds and we hit more than we miss.
This is not the bestest model in the world, but it works. Sells product, mostly targeted at the right people.
As I was doing research for a deck I did on long tail CRM, I found a case study that was too good to pass up. I have been using it for the past few weeks in presentations and talks, but wanted to get your thoughts. This is from a company that chose to remain nameless, but can tell you that they are in telecommunications. I can assure you, their work applies to either B2B or B2C or whatever letters and numbers you want to put together. OK, stage is set.
They used email marketing. As we all know, email is “virtually free” to send; but that is irrelevant. See the table below, first column is past, second column is the new model they use for segmentation (more on that after the fold).
| mass market segments | long tail segments | |
| segment size | 20,000 | 200 |
| emails sent | 18,762 | 200 |
| emails opened | 15,449 | 162 |
| emails clicked | 817 | 98 |
| leads | 42 | 52 |
| deals closed | 2 | 12 |
Now, anyway you want to look at that — it is good.
Either because they sent fewer emails, got more of them opened, more clicks,, generated a similar number of leads as percentage of people reached — and 6x more closes. These are good numbers, no matter how you look at them. As I said, they are based on a real case study from a company I talked to at length and you can see that they knew what they were doing, the number of emails opened was outstanding before the long tail segments were created.
Of course, now you want to know how they did that, what is the long tail they aimed for. That is the purpose of this post anyway…
Use Case Segmentation.
Instead of focusing on profitability, past purchases, ownership, time, dollars, or similar this company used analytics to find a very specific use case (sorry, cannot reveal details here) among their customers and found a service that applied to those people. By using the data available to them, they were able to fine-tooth-comb their customer base and found these few people who were more receptive to the message (which was also carefully crafted to reflect the use case) and make more money, for a lower cost (we won’t debate that now, but let’s assume a lot of sunk costs and just a per-email cost was used to calculate it).
What do you think? Would love to hear your comments…
ThinkJar - The Blog (Archive)! 
I was hooked when I saw the table of results. I think the issue here is that saying they used a use case but not revealing the details leaves so much up in the air that it’s tough to comment on it. I work for an analytics software company, arcplan, so I was primed and ready to hear about the specifics of this case and I’m disappointed that so much was left out.
LikeLike
Heather,
Thanks for posting and commenting. I share your frustration, but i thought the results were good enough to let people know there is a better way to segment. Unfortunately, a lot of what i do is not shareable, as with most analysts, since it is very strategic to the people we work and talk to.
I will try again, or try to find someone else who can give me more details. Now I have been “baited” to find out more….
Thanks
Esteban
LikeLike
A great example of rifle beating shot-gun. Likewise disappointed that use-case was thin.
In B2B the campaign numbers may be very small, but results at the level you show can be very rewarding. Email marketing automation tends to reduce numbers per message even further, making ROI difficult to justify.
I recently wrote a blog post on segmentation identifying 5 static profile attributes (geog, industry, company size, role, current equipment) and 5 dynamic or behaviour attributes (pain, aspirations, buying stage, info sources, risk attitude).
Mini-campaigns to a segment based on a combination of these attributes could be justified if the results you show above can be replicated.
LikeLike
Hi E,
Interesting!! My assumption is that the specific use case you are talking about is a specific “Customer job to be done”. Would you agree?
I’m also very interested to understand what kind of data-points they discovered to be unique to those Customers so that they could use it to tooth-comb others out of their entire base. In my experience that’s a big challenge, because basically most data available in many companies on a Customer level is the same or very similar for all Customers, thus far from unique. Finding unique data-points and/or combinations of data-points so specific that it only fits the specific segment of 200 Customers is not an easy task.
On the other hand.. we’re talking 6 % conversion rate here. That’s still not very impressive, though the improvement is!
Wim
LikeLike
Esteban,
OH! Such a teaser. Ever hought of a second career in marketing? No, didn’t think so.
With the objective of opening up some eyes to a new way of segmenting and to starting a dialog here, mission accomplished.
Here’s my thoughts (questions) to which you probably can’t share under that whole NDA thing. But, here goes anyway.
Those 12 deals – is the use case addressed from an existing service within the portfolio?
Or was the service created specifically for this long-tail segment?
If not, is the solution supportable within the portfolio?
What is the shelf life for the revenue stream/profit/cash flow?
Is it a one time bump? Or a long-term additive?
(This is the part i think you can talk about)
Is targeting micro segments/the long tail a sustainable model for any company? As has been proven, it works for the Amazons of the world. But does it work for Haliburton?
I guess thats it for now on this Friday afternoon. Hope to read more thoughts. I’m sure it will spur additional questions from me.
Cheers
LikeLike
I agree with Wim. Segmentation along lines convenient to the company is lazy and doesn’t address the job a segment of customers is trying to get done. For instance, the job of ABC company is not to maintain their HQ in zipcode xxxxx or to take on the persona of SIC code 1234. It amazes me when marketers draw paychecks thinking some new algorithm is going to make their approach to demographic segmentation better than the company next door.
Hey, I don’t want to make marketing’s job more difficult but aren’t we trying to make our customers jobs faster, more stable and more efficient? Shouldn’t we be making their jobs easier? That probably means we should work hard at finding the message that gets results (or clicks)
LikeLike
Hi Esteban.
Intriguing post and a definite conversation starter! Like the rest, the particular use case details would be interesting. But I also appreciate the sensitive nature of divulging too much.
Wim highlights an interesting aspect with regard to the data points used to identify the use case. I’m wondering, from an approach perspective, if the company used analytics to segment targets based on
some “pre-identfied” use cases. Or, if they used advanced analytics to surface a particular “customer job to be done” as Wim mentions, and then target accordingly. Either way, this example certainly shows the power of analytics to extend traditional segmentation models or create new ones for more effective results.
Thanks for the post!
LikeLike
Segmentation is dead.
Segmentation existed to manage the arbitrage between off line advertisers and companies. The common language between the 2 was socio demographic segments…
So the advertisers created inventory that was targetted at segments, and marketeers segmented their databases into the same tiers.
Businesses aren’t generally built around servicing socio demographic segments – but specific customer needs.
As marketing increasingly enables us to target on a per user basis – we’ll reduce the arbitrage on ad spend – and end up with more useful, targetted advertising.
Nilan Peiris
Chief Marketing Technology Officer
http://www.holidayextras.com
LikeLike
Need. More.data. Esteban.
LikeLike
Esteban,
Actually I think you gave all the details needed. The results stand to reason. A specifically tailored response to an equivalent need is bound to resonate and appeal more effectively than a more general focus which was the segment. If analytics can unearth the nuggets, then propositioning becomes a whole lot more effective
Martin
LikeLike