Adventures in Mobile Marketing

Just had a great conversation with SendGrid – we’re already using their API for sending in-app messanging transactionally, at work. The API is easy to use, and you have great reporting on email deliverability through the web interface. Just chatted with two representatives at Women2.0′s Startup Weekend event- they’re sponsoring some prizes.

I’m also a fan of MailChimp- and have setup various clients on them. The Flash/UI interface is really great, and the reporting on the various email campaigns is also easy. They, like SendGrid, have a very communicative and readily available customer service – for engineering and design.

I just saw a new low volume ESP emerge on the scene, and before eye rolling and “not another one!” I wondered just how difficult it is for people who want an ESP to pick one, amongst the myriad of options. So this is what goes through my mind when I evaluate a small volume ESP:
- Do they understand CAN-SPAM
- (if they have a blog) is their advice in line with best practices recognized in the community
- What kind of clients do they have, and are those examples, examples of good email?

Now, as a person shopping for an ESP, don’t get overwhelmed. Think about these questions:

1. Do you need a lot of handholding? If so, check if they have consulting available, or a reputation for good customer service.

2. Are you interested in leveraging templates? Templates are a great way to keep costs down and ensure that your audience, no matter the browser/email flavor, can have a rich experience. Some vendors have better looking templates than others.

3. Do you want a price-per-email, or a flat rate? I personally don’t like per-thousand rates, because it penalizes a prospering business, and there are some ESPs out there that loosely according to the amount sent.

4. If you can admit you’re responding to an ad, do some research before finally deciding. Some of the smaller size ESPs will give you a lot better customer service than the ones that are large and can afford marketing. But also, the ones that are busier will have better persistent systems, failover, etc. But in the negative, they will also have other customers that are spammy which can affect your campaigns.

There are a basic set of features to good ESPs:
- Very good sender reputation handling, a relationship with ISPs and feedback loops
- Sophisticated templates that work in all major browsers, O/S, webmail and email clients.
- Recommendations on best practices in email marketing
- Metric reporting on bounces, sends, opens by campaign

There are also new and emerging technologies that some are including (sometimes free, or included):
- Preview of design in various inboxes
- Email to RSS so you can distribute content (and vice-versa)
- Sharing chicklet to leverage social media
- Certification available with GoodMail so you can use video in email
- Response filtering, reusable segmentation
- Integration with web analytics for live, transactional emailing

Questions from the overflowing mailbag…

Changes in requests from different people in our team means different database structure (and this tidbit: Salesforce doesn’t allow outer joins!).- friend at dinner tonight

That sucks. Beyond a redesign (on MySql) and an ad-hoc query software, I don’t know what to tell you but this is the 4th non-profit I’ve heard of who has taken advantage of Salesforce’s 10K free-to-non-profits deal, and subsequently been bitten somewhere in the a$$ by restrictions. Note, negotiating is not dead with some of the smaller end ESPs, and it’s never too late to relocate your mail services.

Why doesn’t anyone write about trigger/transactional emails anymore? – Ben at MailChimp
Personally, I’ve never understood those fickle bloggers. Is it not discussed because it’s not sexy? Because its old news? Because nobody asks (I really don’t think most marketing departments use them- prove me wrong). I think, for the audience, this is a battle in their workplace they are just tired of fighting (promotional vs. lifecycle). So, in arguing for more at your place of work, make the conversation about ‘lowering costs’ versus ‘revenue’ and you will win that argument.

How can I share this cool article about fundraising with email marketers…

From this twitter, started a covnersation with Tyler of Involver about how their tool works- you embed a little image of the video- but more importantly, there is a follow-up call to action after it plays, and deep links to the video for sharing, plus a little chiclet to share it out. Case study by a client at Stanford regarding its relevance to non-profit spheres, too. He saw a 23% lift in fundraising over the year, and 51% of it was online. Interesting (if long- scroll to end) post.

A continuation of a post I wrote last week, “Getting Beyond Your Data Set” I describe the additional data you can use to juice up your campaigns. Now, let’s talk about how this happens.

1) Export from source
2) Transfer to local system
3) Load to email database
4) Use data to segment

Most platforms have schedulers (Unix/Linux: chronjob, Windows: windows scheduler) that can trigger a script (written in a combination of either SQL batch exporter, or Perl, or even Windows MS/SQL packages) to export the data. Then, once the export file has been created, transfer the file to the email marketing platform- internally, or hosted at your ESP. They will then have either a triggered job to load the export, or a timed operation that loads the data. Once it is in your campaign database, you can use the data for segmenting.

Woah!

OK I went over that kind of fast.

Identify new data–> export file (flat, simple) –> timed scheduler to export –> transfer via FTP to local system –> timed job to pick up file –> import script to load data

- Use encryption for transferring files. Losing personal information on your customer is probably the worst thing that can happen in email marketing. Zip with encryption (password) is the lowest security, PGP is one of the more secure methods. In organizing the transfer of files with your vendor, don’t send the password in an email.
- For that reason above, I don’t recommend ever transferring email addresses, via excel spreadsheets, data files, or really any method. Create a customer key and use that to represent the unique user. You can do this in Excel, and of course in all database flavors.
- Having worked with a lot of ESPs to add data to their systems, they’re more than willing to help out and will mostly do all of the work once given an export file.
- Data files come in a few variations, commonly CSV (comma separated values) and fixed format. Data files also have a definition file that outlines the columns and data formats for each element.

I taught a quick session on “Don’t Spam: Lifecycle Email Marketing Strategies” at She’s Geeky, an un-conference. I went over some basic studies on email marketing behavior, common intervals, and the common life-cycle vs. promotional arguments. I asked folks what was going on in their organization. It seems that, due to the economy, email vendors are turning off free services to their non-profit and other “Cinderella Deal” clients.

For one of my audience members, the choice was to simply go to another vendor. I told her about a few (Constant Contact, MailChimp, MyEmma), and we discussed improvements to her current fundraising schedule. Her non-profit would have to pull in more funds to cover the cost, so it was an unpleasant reality.

For another audience member, the email was so tied in with the online services, that it would be a fundamental shift to change vendors. For a struggling start-up that uses email as a functional part of their application, not just as a marketing channel, I can see the frustration.

What to do? For many Web 2.0 companies and start-ups, start with keeping a fully tested suite of templates that degrade properly across email clients, and hosting your own IP for sending mail (note: with a clean bill of health). In my experience few outsourced email providers (ESP) can handle the demands of truly robust online service company. This is beyond transactional and lifecycle emails, and should be managed by web developers that understand email and can work with the functionality of email. Especially if it’s core to the business, it shouldn’t be outsourced, and not for a fee. I have a suspicion that the “free” was used to reel in the client just so at this point in time the client would have no other option but to continue for a steep fee.

So the general idea is that you have a customer database, and you feel it’s inadequate so you go to one of these companies and send them info, and they match it and send it back to you. Sometimes you get these cute scores and nicknames based on zip code, like mine, “Bohemian Mix.” I’m supposed to have an Audi, read the New Yorker, etc. None of that is true.

Managers who decide to buy data appends are crying out in desperation for insights into certain hidden segments of their database. What they really need is an analytical tool to help them dive into their own data. Acquired data is not going to tell them anything useful. Instead, they have a lot of information, they just don’t have the tools to get to it.

Three main reasons why I avoid them at all costs- is that they require you to submit data which they then add to their databases and sell the findings to other customers (anonymous-ly, but still). More importantly, there is no visibility into the logic that they use to determine models and segmentation, and neither are the sources. Also, it’s largely without consumer permission.

But companies have permission from their customers- from interactions. If the companies just take the time and effort to dive into their own data, they can find out trends and behaviors far more significant and relevant than those provided by these vendors.

Try it at home- as I’ve done on numerous occasions to prove the uselessness of these farms- randomly split a list into a control, and two test groups. In group A, further segment using your own behavioral thresholds, in group B, apply the relevant data farm’d segments (i.e. “Bohemian Mix,” “Digerati,” etc.). Give the control group a randomly pruned list of the same size. My experience, doing this on various campaigns over a year, was that data farm segments provided less or no lift than internally derived findings on purchase history, interactions, and past campaign behavior logic, to name a few.

I believe the savvy email marketer has to recognize (and avoid) their own bias, rely on previous findings from tests and analysis, and continue to test and model to create an ongoing, maturing view of the customer. Quick fixes like data appends simply muddy the waters, and worse, substantiate assumptions about the customers without data, testing, and findings behind it. Their sources are too diffuse, and the wins aren’t big enough.

Another issue for me is that to use the appends you must provide them with data- this in turn creates the very product they are selling back to you. Repackaging data and forwarding it on is unethical in my view, a disservice to the consumer, and costly.

A former colleague chatted me yesterday with this perplexing “deal”:

SalesForce is offering non-profits free licenses to their software. But we keep on running into issues with their “governors.”

That’s really great of SF to do- and explains an uptick in “do you do SalesForce?” emails that I’ve been getting. The downside is a serious throttling to list size. SalesForce “governors” control the amount of items in targeted lists, so you can only include very small mailings. Seemed like all of the downside of having a hosted solution with none of the upside. If your experience is different, please post here, my friend’s eager to have an alternative, or fix, to the solution.

The question came up of complex segmentation for small lists. What ESPs can handle it? We determined that manually building a statistical model and scoring it then uploading to ESPs and using the “score” to simple filter the lists, is probably the only real inexpensive solution for non-profits. Note: this is not for transactional, or triggered, messages, just promotional and newsletter-oriented emails.

The alternates I suggested to him: SugarCRM, StrongMail, and a few consultancies that could help him setup the database. He and I had worked on an Epiphany installation and the license for that was too expensive for his clients, so that was out. He wasn’t interested in a per-email pricing either.

Just a quick shoutout to one of my sponsors – Emma Marketing right there on the sidebar. They have a very nice 30-Second Tour. Go Emma!

(This is a series: See Believing the Numbers)

Dear Anna,

Thanks
Blinded by Ops

Dear Blinded,
Strongmail is basically built for internal mail systems like this, but you probably don’t have it. Keep it in mind, though, if you do want a more robust metrics & control system on those internal messages. Deliverability usually refers to seeding your output with specific emails that are then tracked and tested through a third party software. To analyze real customer emails going out in a custom application, that’s a different story.

Essentially, it’s a log parsing problem- as all mail systems issue logs based on the act of sending an email (and failures associated). There are log parsers out there. Getting a one-time parsing of the emails shouldn’t be costly or an issue, if you have internal resources you can point at the problem. Using Habeas as a professional services may work- or any other deliverability expert- as they probably do this as part of discovery of any project.

Getting an ongoing view of internally sent email, though, is a bit thornier. You need to setup trigger points in the application that populate a database (ideally) with various records: sent, opened, clicked, and has segment and cell names associated so you can do some live tracking of the effects of transactional systems. Essentially you want to design an API to another email service, and have the developers access that class or function instead of outputting to sendmail/qmail, or whatever internal mail system they are using.

If you have experiencing parsing mail logs with a favorite tool or also have a situation like this, please comment!