How we Saved our Funnel Drop-offs After They Happened "Product people - Product managers, product designers, UX designers, UX researchers, Business analysts, developers, makers & entrepreneurs 15 September 2017 True Design, email marketing, funnel drop-offs, jifiti, Product Design, Product Development, Product leadership, Product Management Role, Product Planning, Skills, User Experience, Mind the Product Mind the Product Ltd 2113 Product Management 8.452
· 10 minute read

How we Saved our Funnel Drop-offs After They Happened

One of the biggest problems for e-commerce, demonstrated by about 74% of users worldwide, is cart abandonment. It means three out of every four shoppers in your online store walk away AFTER they’ve put their items in the cart. If you’ve ever tried to deal with cart abandonment, you might find our story interesting — we brought back 11% of our checkout drop-offs to successfully complete their purchases.

The remarketing tactic of cart abandonment email campaigns isn’t new. Many online retailers go to the effort of implementing them, for two straightforward reasons:

  1. No matter how much you optimize every step of your funnel (as you should), you will suffer from drop offs. So much can go wrong…
  2. A user who has got as far as your checkout page has shown the highest level of intention to buy. Converting this user into a customer should be much easier than attracting new ones all the way through your funnel.

A drop-off should not be treated as a bitter end, but rather as a new branch out of the funnel.

This is the story of what we did at Jifiti to save those lost souls.

Jifiti offers online gifting. Users can purchase a gift from any retailer and send it digitally, without having to know the recipient’s shipping address or even being sure that the recipient wants this particular gift. The recipient then receives a digital gift notification and can have it shipped to them, redeem it in-store or switch it altogether for store credit.

The gifter’s journey may have various entry points and paths, but the two final steps of the funnel would typically be:

  1. Once the user chooses the product to be sent as a gift, they enter the recipient’s name and email address and write a personal message.
  2. Checkout.

Our “Before” Measurements

The conversion rates we measured for the above steps were:

  • Step #1: 28.8%
  • Step #2: 79.9%
  • Overall conversion of both steps: 28.8% * 79.9% = 23%
  • Overall drop-off in both steps: 100% – 23% = 77%

While 79.9% is quite a high conversion rate for a checkout page, we were not satisfied with the overall conversion of 23% from selecting a gift to completing the purchase.

It meant there was massive untapped potential if we were able to get even some of those other 77% drop-offs back on track.

We knew that most of them would not complete the process, otherwise they wouldn’t have dropped off.

But what if it was just a Chrome crash? What if they couldn’t find their credit card at the moment? What if their significant other wasn’t nearby to approve the purchase? (yes, we can’t do anything by ourselves…)

I was betting on an additional purchase or two. It was worth a test.

Building Some Infrastructure

We made sure that at every step of the purchase funnel, everything was logged and saved to our database (asynchronously of course, for the sake of user experience). For example, if the user enters an email address at an early stage of the funnel, we save it instantly, we don’t wait for the transaction submission, which might not even occur. This way, even if the user has dropped off mid-funnel, we still have a trace to follow.

Now we can easily track the user session and know where it ends and also what data was entered so far. This way we can “reconstruct” the session and get the user back to the exact drop-off point, because there shouldn’t be any effort on their end. With one click the user is positioned at the same stage of the funnel, with the item they selected during the original session waiting for them at the checkout, as well as with all the original information pre-filled in the form.

Our Toolset

For tracking users’ activity through the site we used our previously built proprietary infrastructure, which already had token-based hooks on our pages sending async event logs to our database. The same system was then expanded for reconstructing the sessions from the same data.

For sending the emails, we’re using Mandrill API.

For A/B testing, we’re using Optimizely.

Pay Special Attention to Mobile

Keep in mind that most of the emails you send will be opened on a mobile device — even if the user initially started the funnel on a desktop. The infrastructure described above had to be fully responsive and optimized for sending the user to the right step of the funnel on their current device — not necessarily the original one — and allow for the completion of the checkout process as smoothly as possible.

It’s important to save your session token on the server and not rely on any device specific session data storage, like cookies or HTML5 Local Storage. That would make your session reconstruction bound only to the original device.

Tips for an Effective Email

  1. First, think of the value to the user, not to your funnel. Keep that perspective while writing. A pushy sales email would only antagonize your users. Instead, provide service, answer questions and offer help. They may actually need it to complete the purchase. Write a direct, personal email signed by a customer support team member. Explain who you are and why you’re writing to them.
  2. Write a short message with a clear call to action. Mention relevant and concrete data in the message title (to increase open rates) and in the body of the email (to encourage click-through). An email titled “Your gift for [recipient name]” will generate a better CTR than a generic title.
  3. A simple text email is sometimes more effective than a designed and branded one. If you are working with a large, highly recognizable brand, you might want to leverage the brand’s impact. However, working and extensively A/B testing with some top brands, such as Ikea, we found that conversion in personal, text-only emails perform up to 80% better, compared to designed and branded ones.
    In short, this is because an email is a one to one communication channel. This is how your users perceive their inbox. An email that adheres to this perception will be better accepted & more noticed (since it won’t fall into the banner-blindness pit that the designed emails fall to) and, if written properly, even more compelling to answer, as if there’s a real person at the other end.
  4. Compose a separate message for each scenario, since the generic approach tends to be less clear and doesn’t convert as well. In our case there were two such scenarios, but you might have more.
  5. Consider a secondary call to action in the email (not sure? A/B test it). On the one hand, it might serve as a distraction, but on the other, it may help the user to come on board again, even if they don’t want to complete that same purchase. For example, we offered an option to browse a catalog of other related and relevant items. If the user isn’t interested in buying what they initially planned, they still have an easy way into the funnel again. Should they choose another item, the previously entered personal information is automatically filled out, speeding up the process.
  6. You could offer a discount to encourage users to complete the purchase. We chose not to, for now. If you do, take into account the technical complexity it might add, and the possible business implications of cannibalizing your sales. A/B test it carefully.
  7. Add an unsubscribe link at the bottom. Apart from your legal obligation to comply with spam laws, this is also a good practice and a convenient way to mitigate the risk of antagonizing the user. Users are more trusting when they know they can opt out of emails that “know stuff” about them. As a bonus, including the unsubscribe link also makes their email server more trusting, and increases deliverability rates.

Here is a sample email:

Hi [sender first name]!

My name is Stephanie, and I’m on the Jifiti Gifts team.

I noticed you didn’t end up purchasing the [item name] for [recipient name]. I simply wanted to reach out and see if there is anything I can help with.

If you would still like to complete your purchase, you can [click here] to go back to the checkout.

If you would like to select something else, you can [click here] to go back to our gift catalog.

Feel free to reply to this email if you need any help with your gifting.

(If you already received a purchase confirmation, please disregard this email)

All the best,


Setting up, Testing and Running

Ensure that everything is trackable and measurable. Just like any other MVP, at first you mainly want to measure and learn what you can. We made sure that every click in this new branch of the funnel was logged so we can learn what works and what doesn’t.

After we set up tracking, we created a list of drop-offs (from both process steps) and decided on the right timing to send. On the one hand, you don’t want to wait for too long between the drop-off and the email. You want to offer considerate and attentive customer service and save the sale. On the other hand, if you send instantly, the reason for the initial drop might still be present, for example, “I don’t have time, have to leave the office”, “I left my credit card at home” or “I want to confirm the purchase with my spouse”.

Test and find the sweet spot. We started with a one-day delay and later changed it to a two-hour delay, based on the A/B test results.

Needless to say, this is not a one-time campaign. This has become an ongoing component in our system.

Analyzing the Results

Running this remarketing campaign yielded results quite quickly. Remember the 77% funnel drop-offs we started with? Out of these drop-offs:

  • 22% of the users who had dropped off at checkout (step #2) had clicked through to go back to the funnel.
  • Out of them, a total of 25% went right back to checkout and completed their purchase. 21% went back to the catalog and still completed the purchase.
  • The final result? 11% of all our drop-offs (from both steps combined) completed their purchase due to this email remarketing campaign.

We also learned that:

  • The Step #2 drop-offs convert 7 times better than the Step #1 drop-offs. That’s no surprise. But the Step #1 volumes are much higher, considering this is the original purchase funnel. Since many more drop-offs occur on Step #1, the total numbers add up to be significant.
  • The more focused primary call to action converted better than the more generic secondary call to action, but the secondary one still adds some long tail value.
  • Open rate for this remarketing campaign emails was 44%.

One More Unexpected Result

Another advantage of the personal / text-only email approach is that people started replying to these emails. Literally thousands of people took the time to reply.

Some considered it annoying, but they were a minority. The vast majority of users responded very positively and appreciated the follow-up. Some explained why they didn’t purchase, some asked for help in completing the purchase, some thanked us for the reminder. Besides the additional purchases generated here, this created two important outcomes:

  1. It gave us a chance to engage with our users. Our support team replied back, helped with the purchase, and made them happier customers.
  2. It taught us more about our product and gave us ideas on how to improve it to avoid some of these drop-offs in the future.

Key Takeaways

  • Cart abandonment emails work. They are also one of the highest ROI actions you can take to increase conversion.
  • The further in your original funnel the user has advanced, the higher purchase intent they’ve shown, therefore the conversion is easier.
  • Be as personal and as concrete as possible with the information you include in the email. Use everything you already know about that user to increase engagement.
  • A simple, direct text email offering help is more effective than a designed sales email.
  • Track the user journey and data to the smallest detail. Then use it to take them back to where they dropped off.
  • Make your campaign device agnostic. Most remarketing traffic will be mobile even if most drop-offs were on desktop.
  • Treat this campaign as any other minimum viable product you build. Measure everything, aim for short feedback loops and A/B test any variation you can think of.
  • Your users will reply. Use it as a way to engage with them and to learn what they really think of your product and what you can improve in it.

Comments 0

Join the community

Sign up for free to share your thoughts

About the author