Saving Abandoned Carts: The Experts Weigh In
Because blog comments typically don’t see much light of day once the article is more than a day old (and most of our loyal readers subscribing by RSS or email), I wanted to dedicate a post to the topics of optimal cart recovery email timing and cannibalization of natural conversions (that would have happened without the recovery email).
Viktor from Grass Roots Store asked:
Wonder if less than 20 minutes would increase conversion further? Or is that too soon? Is 20 minutes the low end of sending out this email?
I had recalled that SeeWhy's research into cart recovery found immediate remarketing achieved highest conversion rate, so I reached out by email to Charles Nicholls to provide his insight on timing:
There is no doubt that sending a first email immediately once the abandon is detected is the most effective technique. We have run many tests on this, and have always found that the immediate send always outperforms one sent later (there's a Lucky Brand Jeans case study up on SeeWhy.com which is based on an A/B test where the real time one generates three times more revenue than one sent in batch at +24 hours).
The reason for this is that an immediate email ‘connects’ with the emotion of the originally intended purchase. This emotion dissipates rapidly. We often also see higher average order values with an immediate email for this reason (this was a factor at Lucky Brand, for example).
What about natural return rate cannibalization?
Serial abandonment is a common behavior. In fact, the more frequently a single customer abandons a cart, the more likely they are to buy. 2.8 times more likely, according to SeeWhy. Does abandoned cart remarketing eat into organic conversion rates?
James Daniel asked:
…by emailing after 20 minutes there has to be the issue of traffic cannibalisation? Would be good if somebody could clarify.
John Jones commented:
20 mins is far too soon – that would eat into my natural return rate.
Chris Sheen of SaleCycle, the company that developed the infographic chimed in:
We have done some work with our brands looking at even quicker response times, but generally advise against it. Primarily as there is a risk of cannibalising ‘natural recoveries’ where people will still complete their purchase without prompting, and we want to ensure a natural customer journey as much as possible.
As James points out, we do a lot of work with our clients to look at the impact an email might have on the customer experience – and this plays closely to the types of emails we advise our clients to send which are relevant (personalised, product details, etc) and focus on enforcing the strength of the brand (to which the customer has already bought into).
Charles Nicholls reminds us that "you can test that":
In terms of cannibalizing organic sales (i.e. those that would come back and purchase under their own steam) the ONLY way to understand this reliably is by using a control group, where 50% of the traffic gets the remarketing campaign, and 50% gets nothing. Both groups need to be run at the same time because of the effects of promotions. We have measured the organic return rate at on average 8% - so if you do nothing, 8% of visitors where an email address has been collected will come back and buy.
Charles also offered a rationale for why after 20 minutes, recovery rate drops off:
Many web servers use 20 minutes as a standard setting for a session timeout after which the user needs to start over. We’ve analyzed patterns of customer behavior and concluded that for most brands, 21 minutes can reliably be used as a cut off point after which it is statistically very unlikely that the visitor will try and reinvigorate their session and make a purchase. This works for most sites but of course there are always exceptions.
Does "it depend?"
Daniel Kohn from AbandonAid shared:
I’ve also read about the 20 minute sweet spot, but as a consultant on multiple large ecommerce websites, the honest truth is that every site is different and finding that ultimate time depends on multiple factors. I find that if the products being sold are highly price sensitive like electronics, then 20 minutes is too early because they are shopping around and closer to 1 hour works better. If it’s a very high ticket item like jewellery, again 20 minutes is not ideal because a shopper needs more time to think if they are on the fence. Some sites I’ve worked on like ticketing for holidays or concerts etc, 20 minutes might be too long…
To Daniel’s comment, Charles Nicholls replied:
@Daniel has some experience which suggests different results in some markets – we’ve not found that but I’d love to hear more about those. It could well be that certain creative approaches work better with a longer delay in some circumstances (i.e. a more marketing / promotional email might work better with a longer delay).
I can't argue with the quicker-is-better theory. The data backs it up. Both SeeWhy and SaleCycle's research has found diminishing returns as time passes.
I also understand why marketers would be hesitant to pull the trigger too soon. Brands don't want to come across as pushy or spammy. We know serial abandoners spend more money, thus first-abandon triggered emails may cut average order value if they successfully incite an immediate conversion, especially if the email is incentivized. (A holdout test that factors in the AOV of "organic" recoveries will show you whether your remarketing is rushing your customer out of a bigger basket.)
If your site's average days to purchase and visits to purchase is high, combined with a higher-than-average baseline "organic" recovery rate (>8%), you'll likely expect less of an uptick from immediate remarketing than the average business. Those who are just not done shopping and will convert at a later date will simply ignore your triggers. But that doesn't change the concept that sending a retargeting message sooner than later means a higher success rate.
Remarketing emails have to be opened, and targeting customers when the shopping experience is still fresh in their minds is better than hours or days later, when your customer's inbox is cluttered with messages from other businesses.
Testing cart recovery emails
When conducting testing, I also recommend conducting a holdout test (50% of abandoners get a recovery email, 50% don't). But keep in mind, with holdout tests, the success of your challenger depends on design, content and execution. The recovery differential of your first test may be 9%, but a follow up test with a completely different timing, design and offer may have a 26% spread. More than one round of testing, testing different variables against a holdout control, or A/B/C/D testing can provide you insight into what variables provide the most payoff.
Always measure revenue per visitor and profit alongside conversion lift.
And embrace segmentation. Here are some customer segment ideas:
Serial abandoners (revisit=1, revisit=2, revisit=3, etc)
Serial abandoners who come back and add more to cart
Serial abandoners who come back to edit cart contents
Visitors who are logged in customers
Visitors who are also signed up to email program
Visitors who have visited more than X times in the last 90 days
Visitors who have made a purchase in the last 180 days
Visitors who have purchased more than $X in the last 180 days
Cart abandoners vs checkout abandoners (the latter show stronger intent to purchase)
What step of checkout did the abandonment occur?
What device was used (desktop vs tablet vs smartphone)
What say you?
With that, I’ll turn it over to you, our readers. There are undoubtedly some exceptions to every rule when it comes to marketing optimization. Have you tested and found your industry responds better to a longer delay time? Or have you tested cart page abandonment vs. checkout abandonment and found different timing sweet spots? Shoot me an email at firstname.lastname@example.org, your story could be the next feature on Get Elastic.
Off-the-blog, Charles shared with me some tips for remarketing to visitors that haven't provided their email address by account or in the first step of checkout. Tune in next post to learn how...