Can't We All Just Get Along? Deliverability As An Optimization Problem

by George Bilbrey, Nov 21, 2012, 9:56 AM

• Comment
• Recommend
• Tweet
• 

I’m on several spirited discussion lists where email marketers and deliverability folks hang out. Over the last few years, I’ve seen this conversation played out many times:

Deliverability Person (and Occasional “Enlightened” Marketer): “Target, target, target. It’s all about relevance. Send the right message at the right time. Take the dead addresses of your list. You will drive higher response rates and have fewer delivery problems.”

Grizzled Direct Marketer: “More, more, more. Who cares about rates of engagement (clicks, conversions, opens)? I care about numbers. If I send more mail, it may reduce my rates of engagement a bit and even cause a bit of a delivery problem, but my total numbers will go up and that’s what I’m optimizing for. Therefore I’m not cutting back on frequency, and I’m not pulling the non-responders off my list.”

Who’s right?

My take: It depends. If you have a truly opt-in list, this is an optimization problem, not a religious debate. The right tactic depends on how bad your delivery problem is, the driver of the problem, and  the economic value of non-responders.

Unfortunately, I find that many of my deliverability colleagues don’t take marketers’ full economic equations into account when making recommendations. Getting more mail into the inbox is only one lever driving ROI from an email program. It’s true that you can’t get responses from mail that’s not in the inbox, but it’s also true that you are likely to get more responses if you get “more bites at the apple” and send more frequently. Deliverability needs to be less of a one-size-fits-all set of dogmatic recommendations and more of a process of optimizing response based on experimentation and a review of what works for other mailers. Often you can trade off a little inbox placement for more response.

Here are two very sanitized sample scenarios with suggested solutions that take the full economic equation in mind:

1. An ecommerce company with a big list and a huge delivery problem driven by spam traps: Here the vast majority of the program’s value comes from users buying directly from email (click on link, visit site, buy wonderful goods). This particular client has a very active list: very few subscribers have never opened or clicked on a message. Unfortunately this client also has a huge inbox placement issue. By reviewing their reputation data, they can be pretty sure that spam traps on their list are causing the problem -- in this case, “recycled” spam traps at a handful of North American ISPs.

Recommendation: Probably the right thing to do here is cull the list a bit. Stop mailing to inactive subscribers and see what happens. The client should check on how the bounce algorithms work at their ESP -- this is probably the reason that recycled addresses got on the list.  They should review list acquisition practices as well, but this is less important because they are not hitting pristine traps.

Rationale: The economic value of frequency (someone seeing the message) is relatively low here. Culling inactive subscribers doesn’t represent a large cost. Besides, not much of this list is inactive.

2. A travel company with a medium-sized list, good-but-not-great inbox placement (mid 90% range) with low frequency, and slightly high (but not off the chart) complaint rates: The program’s value comes primarily from direct purchases of travel packages from the email, but the client thinks campaigns drive a lot of direct navigation to the website. (On days with campaigns, unattributed traffic on the website goes up quite a bit.) This list contains no spam traps but yields only moderate open, click, and read rates. The client believes the program has the potential to deliver significantly better ROI. When we compared this clients’ program with others in their industry, we saw that others’ complaint rates, delivery rates, and read rates were only a little worse than theirs, but competitors’ frequency was two times higher.

Recommendation: Test sending certain classes of offers more frequently. (More, more, more)

Rationale: It’s working for the competition -- let’s give it a chance to grow revenue.

The issue with an optimization problem is that it requires a lot of data. In these cases, true inbox placement, detailed reputation metrics, and insight into competitive practices and results were needed. However, with the right data, and by keeping the full economic equation in mind, the grizzled marketer and the deliverability person can come to the same answer.

• Comment
• Recommend
• Tweet
• 

Be the first to comment on "Can't We All Just Get Along? Deliverability As An Optimization Problem"

Leave a Comment