Listening to your users: Inferring Affinities and Interests based on actual time spent vs clicks or pageloads

Listening to your users: Inferring Affinities and Interests based on actual time spent vs clicks or pageloads

Personalized recommendations rely on the idea the you know the interests of your audience. In absence of explicit feedback, interests are generally derived from clickstream data: session and event (e.g. click) data. But given that sessions can be short lived (bounce) and clicks can be unintentional, they are unlikely to reflect true interests of your audience if you simply count them.

At Blueshift, we choose to actively follow along the individual’s storyline and extract intelligence from each event to gather insights of the user’s intent and interests, so we can provide better recommendations.

Let’s look at a real user example

In the table below, we see an actual clickstream of events from a user on blueshiftreads.com.

Timestamp Session_id Event Category Book title
12:30:24 session_id1 view Biography & Autobiography > Personal Memoirs Eat Pray Love
12:31:29 session_id1 view Drama > American > General Death of a Salesman
13:48:49 session_id2 view Science > Physics > General Physics of the Impossible
13:49:02 session_id2 view Biography & Autobiography > Personal Memoirs Eat Pray Love
13:49:09 session_id2 view Health & Fitness > Diet & Nutrition > Nutrition The Omnivore’s Dilemma
13:49:19 session_id2 view Health & Fitness > Diet & Nutrition > Nutrition The Omnivore’s Dilemma
13:49:35 session_id2 view Poetry > American > General Leaves of Grass
14:09:47 session_id2 view Poetry > American > General Leaves of Grass
14:10:02 session_id2 add_to_cart Poetry > American > General Leaves of Grass

This specific user interacted during two different sessions, browsing books from different categories. If we try to come up with the top categories for this user, based on total number of sessions, we get:

Rank Category Session count
1 Biography & Autobiography > Personal Memoirs 2
2 Health & Fitness > Diet & Nutrition > Nutrition 1
3 Poetry > American > General 1
4 Science > Physics > General 1

As you can see in the table above, Personal Memoirs is the top category while the three other categories tie to second-place (they have been alphabetically ordered in that case), but other tie-breaking rules can be applied.

Time spent ranking

At Blueshift, we developed algorithms to re-rank these categories according to the time the user actually spent on your products and categories:

Rank Category Time spent
1 Poetry > American > General 1212
2 Biography & Autobiography > Personal Memoirs 72
3 Health & Fitness > Diet & Nutrition > Nutrition 26
4 Science > Physics > General 13

Here, we rank ‘Poetry > American > General’ above the other categories. Note that at the end of the original event stream above, the user actually did add the book from that category to the cart. Even if we would have ignored that event, our time based ranking would have indeed capture a category of interest to this user.

There’s more: decayed time spent

You should be careful not to rely on detailed information from a single user on a single day: if the user indeed bought the book he added to the cart, that might just be an indicator of no longer being interested in that specific category of products. Furthermore, you would want to adapt to changing user interest over time.

That’s why we implemented what we call a decayed time spent algorithm, that combines the time spent by users over a certain period of time (say last week) and that weighs recent time spent as more important to the ranking than time the user spent before (say 14 days ago).

Decayed weighting of recency this way allows recommendations to adapt quickly to shifting user interests when they are shopping during holidays and might be looking for gifts for others as well as themselves.

From user-level signal to site-wide signal

Many product recommendations are related to some site-wide top categories of products, like ‘top viewed’. Using our time based algorithms, we can better rank these top categories. Let’s look at another example from blueshiftreads.com where we show you a part (20-25 to be exact) of the top 25 most popular categories.

Using classical session counting, we obtain the following ranking:

category session count
Juvenile Fiction > People & Places > United States > African American 5358
Juvenile Fiction > Girls & Women 5291
Juvenile Fiction > Family > General 5265
Fiction > Contemporary Women 5215
Fiction > Thrillers > Suspense 4971
Fiction > Mystery & Detective > Women Sleuths 4804

However, when we rerank these categories based on actual time spent by the users, we see that ‘Juvenile Fiction > Girls & Woman’ drops from position 21 (above) to position 23 (below), even though it had 76 user sessions more in the 7 days over which this was calculated. User sessions are no guarantee for actual interest (i.e spending time).

category time spent
Juvenile Fiction > People & Places > United States > African American 102164972
Juvenile Fiction > Family > General 100447985
Fiction > Contemporary Women 98897169
Juvenile Fiction > Girls & Women 98340874
Fiction > Thrillers > Suspense 91140081
Fiction > Mystery & Detective > Women Sleuths 87372604

Furthermore, if we rank the categories using our decayed time spent, we see that ‘Fiction > Contemporary Women’ is actually ranked the highest (21) while it was the lowest (23) in the original list. This indicates that this category received the highest time spend by users in the most recent past.

category time score
Juvenile Fiction > People & Places > United States > African American 28461106.29
Fiction > Contemporary Women 28179308.93
Juvenile Fiction > Girls & Women 28068989.26
Juvenile Fiction > Family > General 27608048.02
Fiction > Thrillers > Suspense 26102829.31
Fiction > Mystery & Detective > Women Sleuths 24597921.38
Ok, why bother?

So why bother re-ranking? Well, most catalogs will exhibit a Long Tail in the distribution of popularity of their content: very few items will be very popular while lots of items will be very unpopular. No matter how you rank the popularity of the top-10 categories (sessions, clicks, time, …) out of a 1000 category catalog, these extremely popular categories will always on top. Just have a look at the top 20 categories from blueshiftreads.com:

blog_post_time_spent_top20

As you can see, the top 5 categories do a lot better than the rest. For most businesses there is a lot of value in promoting content from categories other than these few favorites. Therefore, if you can avoid down-ranking interesting categories for users and do this consistently over your whole catalog, you will be able to recommend products from the appropriate category to the users who care for it. In other words, you will avoid the pitfall of recommending an overly popular yet generic product to your users.

But time spent relates to sessions/clicks anyway?

Yes and no. It is true that more sessions correlate to more time users will spend on categories, but not to the same extent: a session length can range from a second to tens of minutes. Have a look at the next graph below.

What we see is the ranking of the 1000+ categories (on the X-axis) for blueshiftreads.com by popularity (on the Y-axis, logarithmic scale) over 7 days, in terms of 3 different metrics:

  • The blue line represents ranking by session count. It is very smooth because it really ranks all categories just in descending order of session count. This is the standard ranking.
  • The red line represents ranking by time spent by the users. It is equally smooth in the beginning (left) because it ‘agrees’ with the session ranking: as mentioned above, the top popular categories will always be on top. But quite soon, the line becomes spiky: the ranking disagrees with session count, and the spikes indicate that this ranking would reorder the categories in a different way (promoting different categories to the top).
  • The green line is the decayed time spent ranking: the same holds as the time spent ranking. This algorithm also disagrees with session count and would reorder lots of categories in the long tail to promote categories of interest to the user.

blog_post_time_spent_ranking_plot

This re-ranking is exactly what you should do to stop recommending the same popular categories to users that might have indicated (time) interest in other categories.

Beyond Basics : Advanced Personalization strategies for Re-marketing Triggers

Re-marketing based on behavioral triggers like abandoned cart, abandoned search and abandoned view are must do’s for any digital marketer to engage customers and bring them back to your site or app. More often than not it’s very tempting to do just the basics, may be your e-commerce platform vendor gives few out of box “widgets” to replay the content or products and you can tick a box and call it done. But that would be waste of a great opportunity to engage with your users fully at the moment they are most interested in your offerings and showcase the full depth and breadth of your catalog.

Imagine your self in the shoes of your customer – They were on your site or app for a reason. Why did they abandon their visit? Are they looking for better price? Are they looking for affordable alternatives? Are they looking for quality recommendations? Are they looking for things that go along with their previous purchases? Are they looking for hot new products they heard about? Or are they just window shopping? May be it’s mix or all of them and you can’t tell. But you can tell a better story than replaying what they have seen. Multi-touch campaigns over their preferred channel with content personalized to each user are a great way to showcase and up sell your offerings.

Here are few advanced personalization strategies to try, going beyond replaying the products your customer has viewed or added to cart.

  1. Related Products : “What Other Items Do Customers Buy After Viewing This Item?” Based on the product that was abandoned you should consider including these in your messages. Very often this is helpful as a first touch in moving your customer down the conversion funnel helping them make informed choices.
  2. Up-sell Products : “What are the top sellers in the category of the item viewed”? Based on the category of the product that was abandoned you can showcase your best sellers from those categories and optionally you can restrict it to specific brands.
  3. Trending Products : May be some of your products are not best sellers yet but they have been recently added to your catalog and are already selling out. Consider including them in your messaging with appropriate call outs.
  4. Affordable Alternatives : Your catalog likely has a breadth of products that are in the same category but are more affordable. Add them as alternative recommendations.
  5. Most Discounted Products : Your customers will likely want “quality” products but want them to be affordable getting most value for their money. And may be you are doing seasonal promotions or roll backs. Consider tying them together to the product and category of the abandoned product.

Depending upon your catalog and the number of touch points you have with your customers you can do all of the above or a mix of them. At Blueshift we have built a DIY Personalization Studio to do all of the above and more without needing to go through development or IT cycles. Re-marketing based on behavior triggers is the most effective messaging you can do as a digital marketer. Do not waste a great opportunity to win over the customer by doing just the basics. Learn more about behavioral triggers with our comprehensive e-book. Learn how our clients like UrbanLadder are using advanced personalization strategies to improve their conversion rates by 400%. If you are ready to take your triggered marketing to the next level say hello.

5 Tips to Reignite Referral Marketing

Referral marketing, or also known as word-of-mouth marketing, happens organically by enthusiastic or satisfied customers, but it can also be influenced by companies with the right strategy. You would naturally trust a friend’s opinion more than a basic ad you see on the web. And statistics also say that “53% of users who clicked through a link shared by a friend of Facebook went on to make a purchase” (SociableLabs). So if you have been on this bandwagon for a while or just starting to realize its power, here are a few tips to boost your referral program.

  1. Find users who have experienced a high-five moment: From the time a customer makes a purchase to the time they are excited to use it and eventually forget about it, there is a small window of opportunity when you can high-five a customer and influence them to refer your product to their friends. Its important to ask and ask at the peak of their enthusiasm, because it could be the difference between sharing your product with 20 friends and ignoring your request. Blueshift’s segmentation engine and API makes it easy to find customers who share out to friends from your website and reach out to them with a personalized message.
  2. Think about app referrals: Mobile browsing has now overtaken desktop users and mobile marketing can no longer be ignored. Rethink your app referral strategy to make it flow better for the end user. Companies like https://branch.io/ help you track referrals from apps to check your progress and the impact it is making on your sales number. Testing your efforts and measuring the outcome continuously will give you a good idea of your success rate and what steps can be taken to improve your numbers.
  3. Visibility is everything: Make sure your referral option is very visible when you do ask for it. Not only that, it can be included on a product page, or in the signature of a transaction email. It can even be as simple as a check box on the checkout page. The more visible you make it, the more likely a customer is to use it.
  4. Easy peasy: Make sure your referral option is easy to share across the main social networks and email so your customer doesn’t have to jump through hoops to tell their friends about your product. Also create some attractive looking messaging around your product or brand so it makes you look your best to first time customers.
  5. Make an offer: Offering incentives for referrals is a sure way of getting a customers attention and keeping their business. A time sensitive offer creates a sense of urgency for the customer to act now rather than later. Store credit or a discount on a purchase attracts them to share your product and keeps them coming back to your store to redeem their rewards. Either way its a win win situation for your company.

    Naturebox referral

    Example of a time sensitive offer by NatureBox that doesn’t cost your customer anything, it is time sensitive, and makes them look good in front of friends.

There is no one recipe that works for every kind of business. It is an ongoing interaction with your customers that has to be monitored and adjusted with time. Changing up your content and offers from time to time keeps things interesting in the relationship, and keeps your customers engaged.

 

The Case For Predictive Segmentation – Part 1 of 2

Philip Kotler Segmentation Quote

Retention & Growth marketers are often interested in taking action on a segmented base of users. Classic segmentation methods include

  • Lifecycle based segments: new, active, lapsed etc.
  • Behavioral segments based on user behavior on the website/app
  • Demographic: Age, gender, location, household income, education based
  • Traffic source based
  • First purchase product/category etc.

Given all these ways of segmenting, how should any marketer approach segmented marketing, for the purpose of improving their core retention metrics? Some of the metrics CRM or growth marketers might be interested in improving through a segmented marketing strategy may be around activation rates, repeat purchase rates, or churn/retention rates.

Here are 2 steps for how you can use segmented marketing to drive higher response rate on any metric, say, repeat purchase rate:

  1. Use criteria that lead to a big spread in response rates in the steady state: Researchers on segmentation, have pointed out the need for segments to be  identifiable, substantial, accessible, stable, differentiable and actionable.  In the digital world, the tests on identifiable, accessible, and actionable are often easy to meet with segments. However, what really separates useful segments from the rest is the differentiability of the segment especially in response rates, their stability (i.e. whether the same criteria continue to map to differentiated responses over time), and whether the segments are substantial in size. In other words, you need to identify large segments of users whose response rates are substantially (3-10X) different from the average
  2. Test what happens to the steady state when you introduce a message or an offer: Once you have identified your segments in a steady state, you want to be able to test how you could increase the response rates further by introducing new variables like product features, offers, or content.

We will focus this first post in a 2-part series on the 1st of these challenges:large segments of users whose response rates are substantially (3-10X) different from the average. While this sounds simple enough, in practice, quickly figuring out the criteria that lead to a big spread in response rates could take some work. For example, in a typical setting, your gender based response rates might not look very dissimilar for repeat purchase: Let’s say your average repeat rate is 50%, and that men have a repeat purchase rate of 45% and women 55%. Now, there is clearly a difference in repeat purchase rates, but not enough for you to build compelling campaigns around it. At the other end of the spectrum, you might be able to identify outlier users, whose response rates are very high – but these outlier users may not meet either the stability test (e.g. will they continue buying the same way), or may not be a substantial enough segment.

This is where predictive segmentation & machine learning have a role to play. Machine learning can quickly figure out variables that are important, and combine them to come up with models that give you a 3-10X separation on large buckets of users. At Blueshift, we make this easy by making predictive segmentation scores available “on tap” for you to use.

Blueshift Predictive Repeat Purchase Score

Visualization of predictive score percentile with corresponding repeat purchase rates

For instance, in this graph, you can see how multiple variables have been combined to generate a repeat purchase score that gives you a nice spread in predicted response rates. In this example, users in the top 10 percentile of scores have a 90% repeat rate, whereas users in the bottom 10 percentile of scores have a repeat purchase rate of only 9%.

Now that you have identified a way to find segments of users who have a much higher response rates than the average in the steady state, how do you test the actions that will help you improve these metrics? Our second post in this series will look at that.

5 Essentials For E-commerce Push Notifications

Mobile commerce is growing much faster than ecommerce. Mobile apps are not only extending commerce beyond the desktop, but also enabling new e-commerce use cases like  on-demand services.

However, e-commerce apps tend to have much lower retention than other categories like messaging apps. According to a study, e-commerce apps only have a 13% retention after 1 month. Push notifications are often used for increasing user retention and re-engagement within apps. While push notifications can be extremely effective when done well, they can also be annoying when done wrong. Compared to emails, the bar for relevancy that a push notification has to meet is unusually high: firstly, there’s far less that you can say in a push notification; and secondly, unlike emails that can be perused later, push notifications interrupt the user with the expectation of an immediate response.

Relevant rich push notification from Groupon

Relevant rich push notification from Groupon

Here are our 5 rules for driving superior e-commerce retention through relevant push notifications:

  1. Personalize based on the user’s history: There are many ways of personalizing your push notification based on the user’s past behavior in your app and on your website. You could tailor the content to the categories they have shown affinity with in their browsing and buying history. You could time the push notifications to go out at the times when the user has interacted with the notification in the past. Reacting to the user’s implicit preferences in these ways can make the push notification highly engaging.
  2. Target the user in-context: Another dimension for mobile personalization is the user’s context right now, or their recent behavior with your app. Has the user been highly engaged with the app recently? Have they abandoned certain activities ? Has their search behavior shown them to be in market for certain items? Targeting based on such context, especially around location or recent activity, is highly effective.
  3. Surface real-time and time-sensitive content: Since push notifications encourage immediate action, time-sensitive content like new product launches or time-bound promotions do extremely well. Flash-sale sites do this really well by notifying users of new events.
  4. Engage users with rich content: Rich push notifications enable you to send multimedia or app content to your users. By providing more information about the notification upfront, rich push helps increase the relevance of messaging.
  5. Deep link to make it a smoother experience: Deep linking into relevant content within your app enables users to act on the push notification instantly, and drive increased conversions. Instead of opening the home-screen, users perceive the push notification to be more relevant when it lands them directly onto the offer or product promised by the push notification.