Every user that lands on your website is different, but today, they all receive the same experience. It doesn't have to be that way.
Button uses machine learning to understand, and make decisions about that first experience based on a number of properties of the user, their device, where they’re coming from, where they’re going to, and ultimately… what will most likely lead to the best outcomes for them, and for you, the marketer.
Broadly, there are two buckets of decisions this system will make:
- Detecting and routing around problems (more on that later)
- Optimizing for the best outcome (particularly when a customer is unlikely to convert to purchase in that first session) based on our marketers’ goals
Optimizing for outcomes
Our goal is to increase the value of every tap — that means turning every tap into the most significant interaction possible. Hopefully, an immediate purchase, but many other outcomes are important to our marketers, including: app downloads or building SMS subscribers to convert in the future.
Beyond toggling features of the experience, Button builds a model to predict which outcome is most likely to lead to long-term value for the customer.
Dimensionality of a user
You already know this about your traffic — where a user is coming from is one of the biggest indicators of how likely they are to convert.
There are a number of obvious, and non-obvious signals that can be used to understand what intermediate experience will best optimize for a short term, and long-term outcome. These are the inputs to the intelligent routing & decisioning system.
Device & connection speed
Deep linking provides a huge benefit when the user is on a slower device, or lower connection speed, but offering app installs is often less successful and just a distraction. Different versions of the OS have support for different linking features.
Based on a user’s past behavior, are they likely to download an app? Or offer their phone number for follow-up?
Destination URL & Source
Based on source-destination pairs, Button can detect patterns of broken content, or incorrect linking experiences to make sure that they can be fixed, or routed around.
Propensity and LTV
Based on many of these signals above, as well as other data observed on Button for a customer, we calculate a propensity score for each customer — how likely are they to purchase right now?
Based on this propensity score, you can confidently reduce friction on the path to purchase for your highest-likelihood conversion customers, and add additional prompts to retain lower-propensity customers as an app user or an SMS subscriber.
This maximizes the overall value of every single tap and creates the deepest possible relationship with every users that lands on your site.
Detecting and routing around problems
Deep linking isn’t always good (or deep)
There are hundreds of environments that are actively adversarial to deep linking and blanket-applying deep linking everywhere can often do more harm than good (e.g. some prominent social media apps have counter-measures in their in-app webviews)
Button not only can work around some of these to enable deep-linking, but automatically learns when new or existing sources of traffic are not deep linking as intended, and can automatically disable functionality that may otherwise lead to reduced conversion, or route around the problem.
How does this work?
Button observes and measures an enormous amount of traffic to and from many of the biggest marketers in the world. Each of these journeys ends (or doesn’t) in a conversion to purchase.
Permutations of the dimensions that are monitored for intelligent routing are scored, and then our learning system is responsible for observing changes in overall-outcomes in those combinations, and changing our routing logic accordingly.
A small portion of traffic will be constantly exploring other routes, to make sure that we are finding the globally-optimal outcome.
A few examples …
Detecting specific issues
- A single referrer to a path on your site has dropped in conversion when deep linked
More general changes
- A publisher stops correctly deep linking users for all destinations
Broad systematic problems / outages
- All deep linking doesn’t work on a new version of your app, or a new OS version
Button would detect each of these changes and start routing the user around the detected issue. A small % of traffic would be periodically linked to the original destination to measure whether the issue has healed.
If you’re interested in learning more or exploring how you can drive 30% higher ROAS off your spend on Facebook and in other channels, reach out the team at Button today.