The latest in our ‘ask the expert’ series focuses on a core skill in the digital marketer’s skillset – email optimisation.
Here are their pearls of wisdom. Note you can skip between questions using the links below.
- There are lots of things to optimise. Where should marketers look first?
- Is there a particular metric marketers should be optimising for?
- What new technologies or consumer behaviors are having most impact on email optimization?
- What types of emails can be optimized most effectively?
- How far will AI take us? How important is the creative person and their instinct?
Kath Pay, founder and senior consultant, Holistic Email Marketing:
I would focus firstly on the subject line, as if they don’t open/read the email, then it’s hard to test everything else. But over and above that I recommend testing motivations as the more you know about your customers, the better you can speak to them.
So, ask them what they like best; via testing in the channel that is one of the key drivers of traffic to your website – email marketing. Your email database contains your target market, so use email’s unique push ability and treat every email as a survey by asking your customers what they like via a scientific A/B testing program.
By using a hypothesis and testing to determine a motivation, you are not limited to just testing one factor (i.e. subject line, CTA, landing page, copy, imagery). As long as they all support the hypothesis, for example “benefit-led copy will increase conversions over loss aversion-led copy” – you will be able to test the subject line, CTA, headlines, copy, imagery and landing page as you are testing a motivation rather than a factor or element of the email.
This is what we call Holistic Testing.
By seeking these long-term valuable insights through email marketing, you not only increase results within email marketing but you can share them across other channels to drive the business objectives that are common to all channels. Apply what you’ve learned to your website copy and organisation, to your search keywords, PPC campaigns and related landing pages, and in ads and banners you run on third-party sites in remarketing or network campaigns.
Email gives you a good basic testing structure that you can build on to sharpen your insights and improve your marketing efforts bit by bit across all channels. It’s another one of email’s superpowers that marketers so often overlook or ignore.
Ultimately, it’s another reason why investing both time and money in email pays off across your entire marketing program.
Parry Malm, CEO, Phrasee:
Well, here’s the thing. I run Phrasee, a company that uses AI to create better subject lines than humans. So what am I gonna say here?
Still, jokes aside: your subject line is the crux of your email marketing programme. If it sucks, then your snazzy content won’t get seen, no matter what time you send it. So yeah, logic and statistics indicate the subject line is where to start. Sure, I’m biased. BUT – that doesn’t make me wrong.
Dale Langley, head of deliverability, Emarsys:
Searching the Internet for ways to optimise your email program can often lead marketers into a crazed frenzy of making changes with little understanding of whether it’s sensible to make those changes and what the long term consequences may be. My advice is to remember one thing; no-one knows better what your customers want than you. You just need to read the signals and have a plan in place for measuring results. Different optimisations can work better for different stages of the customer lifecycle and some are better suited to short-term gains at the expense of long-term results.
For example, using tempting offers in your email subject line such as 50% off will surprise (some of) your customers and lead to an increase in clicks, but where do you go from there? Some brands are now so perpetually stuck in the discount game that they’re unable to get out. Instead, use different offers for different segments (such as lapsing high-value customers) and use other channels, such as social media, to encourage one-time purchasers to re-engage.
You also have to recognise that a customer’s profile will change over time. This includes when they prefer to read email, the frequency that they want to receive it at and the topics that interest them, a common mistake of marketers is to build a marvellous customer journey but to forget that customers can switch personas at any point. Our advice is to build out your personas and understand the motivation for each persona to engage with your brand, model the customer journey for each persona and built-in the ability for personas to change over time. This requires effort but it’ll pay dividends in the long-run.
And if you’re interested in testing things like subject line, send time, content etc. make sure that you’re using a proper control group. This means that for a particular campaign, you will exclude a group of customers (the control group) who are similar to customers in the rest of your database and are only excluded for this one campaign. When you make your change (to subject line, send-time etc), measure the revenue generated by the control group vs. the campaign and you can determine whether the tactic you’ve employed yielded any results.
Parry Malm, Phrasee:
Many will say to focus on end conversions, and, in related news, many also don’t have a strong grasp on statistics.
Think about it like this. Say you’ve got a list of 1m subscribers, and 20% open. That’s 200,000 events you can learn from. But then let’s assume a 10% click-to-open rate. You’re down to just 10,000 events – and you run the risk of making decisions on insignificant numbers. Let’s say you get a 10% conversion rate of clicks – that’s just 1,000 events. You’re unlikely to get a statistically significant result, and thus will be making decisions based upon random variance. (Pro tip: anyone who doesn’t understand this should not be in your analytics department)
Here’s another fact: the data shows that, in the long run, open rates correlate very strongly with click rates. And guess what? Click rates correlate very strongly with conversion rates. Therefore, the dominant strategy is to use opens as a proxy metric for email marketing success.
There are, of course, caveats to this. For example, you shouldn’t resort to spammer techniques just to get a few more opens in the short-run. Never forget that email marketing is, in essence, a form of advertising. You wouldn’t put an ad on TV that was off-brand, so why would you send out emails that are?
The sweet spot is when you’re maximising open rates whilst remaining on brand. That’s when you’re winning at life. Well, winning at email, but still, you gotta take the wins you can get.
Dale Langley, Emarsys:
I’m tempted to say revenue since that’s the ultimate goal of most email programs. However, I’m going to say inbox placement rate since if your email isn’t in the inbox, it’s not generating revenue!
If you’re having difficulties due to inbox placement then you need to figure out which levers you can pull to persuade the spam filters that your email should be delivered. Spam filtering mostly occurs these days when you’re sending email to people who aren’t engaging with your brand – so improve the acquisition tactics, introduce a re-engagement program and be prepared to suppress subscribers to your email program (or target through other channels) when they’re ignoring your efforts.
Dale Langley, Emarsys:
Consumers now have a relationship with the brands they engage with, right from the first interaction you’re influencing whether a consumer will turn into a loyal advocate or a one-hit wonder
The days of your IT team triggering an email from the website upon signup and purchase, before the marketer takes over with batch and blast are gone. Every interaction a consumer has with your brand should be influenced by the marketer and through smart, AI-driven marketing automation platforms (like Emarsys), on which you have the ability to craft programs that adapt to the changing needs of the consumer.
Parry Malm, Phrasee:
To exist as a concept, “optimisation” inherently requires measurement. Therefore, the key feasibility driver is universe size. If you’ve got a list of 1000 people on your list, well, you can “optimise” to your heart’s content… but you’ll just be doing it for the lols, as any significance measures will be unattainable.
It’s pretty simple, really. Focus on your campaigns with the biggest audience, which will in turn have the biggest impact on your bottom line. You’ll have way more data to experiment on and learn from. Then, apply what you learned to your smaller campaigns, and boom goes the dynamite. You’ll get that promotion you’ve been haranguing your boss about for the last six months. Congrats, big timer!
Kath Pay, Holistic Email Marketing:
All types can be – whether they’re campaign-based or automated. For either of these, processes and planning are key to success. With automated programmes, the hypothesis is being tested over time, which reduces the chances of time-sensitive anomalies, world events etc. affecting the results. However, with campaign-based tests, ideally, the hypothesis should be tested multiple times to ensure that the results are valid. A statistical confidence calculator should be used in all cases.
Parry Malm, Phrasee:
AI can do a lot less than what you probably think it can. For every AlphaGo, there’s a Tay, the racist chatbot. Everyone seems to think we’re approaching an AI apocalypse, where the machines take over. When that happens, I, for one, will welcome our new robot overlords.
Here’s the reality. There are limited – and powerful – use cases for AI currently, and also for the foreseeable future. Here’s one powerful use case as a (totally unexpected amirite?) example: using AI to generate optimal subject lines. It is a known business requirement – to increase eyeshare on your marketing messages – that we solve by combining two forms of AI (NLG & deep learning). It’s a niche problem, for sure, and that’s by design.
The fact that we use AI is super dope. But – and it’s a Sir-Mixalot-sized BUT – that in itself doesn’t solve your problems.
Instead, here’s my advice: if you have 60 minutes to solve a problem, spend 55 minutes thinking about the problem, and 5 minutes on the solution. If the solution uses AI, awesome, and if not, that’s OK too. The important thing is that your problem is solved.
Dale Langley, Emarsys:
In simple terms, human-driven personalization can’t scale. It just can’t deliver on true 1:1 experience that consumers want and brands strive for. AI can, which provides an opportunity for brands that understand how to introduce hype-free, tangible AI solutions effectively. AI allows us to analyze vast amounts of data, understand consumer patterns and channel preference and to craft incredibly personalized consumer journeys. Furthermore, AI allows us to do this in real-time rather than spending hours creating huge (or many micro) segments. It’s taking us much closer to the 1:1 consumer-brand relationship that we’re all striving for….the promise of marketing.
However, AI isn’t human. It can’t (yet) build something from scratch without basing it on what it knows from the past and it can’t cater well for emotion and a true personal connection. This is why the combination of AI and human ingenuity is the key to successful marketing. We believe that AI can take on the burden of marketing execution, leaving the marketer more time to focus on strategy, content and what we believe will be a new creative renaissance.