Data Science Thesis at Akanoo

Hallo!
Ich bin Tillmann Radmer (26), studiere an der Humboldt Universität Berlin Wirtschaftsinformatik und habe zuvor meinen Bachelor als Wirtschaftsingenieur in Hamburg gemacht. In meiner Masterarbeit erforsche und vergleiche ich neuartige Ansätze zur Optimierung von Uplift-Modellen zur interaktiven Besucheransprache.
Durch ausgeklügeltere Modelle kann Akanoo Kunden gezielter ansprechen und die Conversion Rate für seine Kunden verbessern. Ich möchte hier einen kleine Einführung in das Uplift Modelling zur gezielten Besucheransprache geben.

Uplift Modeling

Often times an important question in direct marketing is which customers to target. This is because some customers will not react to the campaign, while still incurring the variable costs of marketing. Worse, some customers might react negatively to receiving an ad. Generally, there are four types of customers in direct marketing:

  1. Customers who will respond without treatment,
  2. customers who will only respond after receiving a treatment,
  3. customers who will not respond because of a treatment, and
  4. customers who will not respond regardless.

The goal of uplift modeling is to differentiate customers in group 2 from those in the other groups.

Start with A/B-testing

To build an uplift model one needs the result from an A/B-testing experiment. Using two groups and a binary outcome will give four possible combinations as shown below.

Figure 1: Four possibilities in an A/B-test experiment. (Shaar2016)

Given the result one can calculate the uplift as

P(return|treatment)-P(return|control).

Unfortunately, this term is not defined for a single customer, so we cannot simply build a model that maximizes it. Instead there are serveral approaches to maximize it indirectly.

Building a model

The simplest one neglects the control group and models the probability that a customer will return given that she received treatment and given some customer specific information.

P(return|treatment,information)

Throwing away half the data isn’t optimal. For instance this model doesn’t predict which customers will only return when given some treatment, as we required above. Instead it will give all customers that will return regardless of treatment.

A simple extension is to build two models, one for each term in the uplift equation. Formally,

P(return|treatment,information)-P(return|control,information).

While this approach directly models the uplift effect, research suggests that it doesn’t always perform well in practice. One reason for that is that the treatment effect is usually much smaller than the main effect, e.g. if the main effect is 1.0% in the control group and the combined effect is 1.1% in the treatment group, then the uplift effect is only 0.1% (=1.1%-1.0%). Consequently the individual model will mostly focus on the main effect within each group. (Radcliffe, 2011)

An easy way to a single model approach is to apply a class variable transformation to the result from the A/B-test. Both the TR and the CNR groups will be assigned to the positive group (e.g. treatment) and TNR and CR will be assigned to the negative group (no treatment).

Table 1:  Example of the class variable transformation.

Customer

Group

1=Treatment

0=Control

Returned

1=Returned

0=Not Returned

Transformed
1 0 1 0
2 1 0 0
3 0 0 1
4 1 0 0
5 1 1 1
6 0 0 1

The idea is that from the four fields, we would definitely like to treat the customers in the group with the TR outcome. And because we don’t know whether the customers in the CNR would have responded if they had received a treatment, we would also like to treat them. It can be shown that under some assumptions modeling the conditional probability of the transformed class variable is equivalent to modeling the conditional probabilities of the two original variables from uplift equation above.

There are more elaborate techniques that incorporate the maximization of the uplift equation into the training algorithm. However, the straightforward implementation and applicability to any standard machine-learning algorithm make the two models approach and the class variable transformation approach very attractive first choices.

New feature: visitor targeting on product title selection

screenshot_hagebau_hochdruckreiniger

 

Our online shopper analysis has shown a strong correlation between product title selections and the conversion rate. If visitors copy the product title on a product detail page, they show a significantly lower conversion rate.

We have developed a feature to counteract this behaviour. Our technology identifies if a visitor copies the product title while on a product detail page. Akanoo then potentially interacts with a personalised campaign based on this observation and further visitor characteristics.

Currently, we are testing the feature with some of our clients and see very promising results. If you want to learn more about visitor targeting on product title selection, contact us here.

New customers: Schecker and Testrut

new_customer_schecker_testrut

We are happy to announce the collaboration with two new customers: Schecker and Testrut.

We met the Schecker employees during our pitch for the Rising Star Award and could convince them of our technology. Schecker is an specialist dealer for dog supplies. In their online shop dog owners can buy everything their heart desires. The second new client for Akanoo is Testrut, an international supplier for house and garden accessories.

We are looking forward to successful collaborations and a a lot of effective campaigns.

Akanoo User Observation: Volunteers wanted!

Obersavation Online Shopping Behaviour

You are keen on online shopping? Then you are the right candidate to participate in our user observation.

The aim of the observation is to gain a better understanding of online shoppers behaviour. Therefore, we are planning to record the participants while surfing on an online shop. The session will take around 45 minutes. The camera will be pointed at the laptop, not your face, and the video will NOT be shared outside of the team. You need to bring your own laptop. In return for your effort and time, we will provide you with some coffee, cake and interesting conversations with our charming team members. ;)

Sounds like you are right the person for this? We are happy to hear from you: data@akanoo.com. If you have any open questions, feel free to contact us, too.

New Akanoo office in France

new_akanoo_office

Akanoo has opened an office at the coworking space La Cantine in Nantes, France. This enables us to offer our French clients and prospects an extensive, personal support. We spoke with Carole, our Country Manager for France, about her first weeks in the new office:

What do you like most about your new office?

It is the perfect place to meet people from other young companies. The atmosphere is very nice, friendly and cosy. We can share ideas, learn from each other and participate in lot of workshops and events. It is great! The startup network here is very dynamic and it is fascinating to discover all the new ideas in and outside the digital world.

Furthermore, it is good to be closer to our prospects and clients. I can leave to Paris with the first train on the morning and be back in the evening. This allows me to meet our prospects and clients in person regularly.

What do you like most about your new home town?

Nantes is a fantastic city. It is big enough to have everything you want, but on the other hand you are only 45 minutes away from the sea! People are friendly and I’m happy to get back the french gastronomy (and french wines :) )

Do you miss Hamburg?

Of course, I miss all of you and I was so happy to be back in Hamburg two weeks ago! I know that I will come back often to see you and enjoy a barbecue near the Alster again. :)

Have you met any interesting prospects yet?

Yes, definitely. I think we will welcome them very soon – stay stunned! ;) I’m really pleased to announce that we start new collaborations with Bricozor, La Redoute, Excédence and Gravissimo. We are very happy that they choose our solution and we are excited to work with them.

Meet Our Data Scientist Gundula

employee_interview_gundula

Gundula works as a data scientist at Akanoo. She implements new features for our predictive models. Furthermore, she spents a lot of time on preparing and evaluating experiments and monitoring the nightly training cycles.

How long have you been working for Akanoo and what have you done before that?

I started working with Akanoo in November 2014. Before, I did a PhD in computational neuroscience: My research was about the processing of auditory information of calling songs in the brain of grasshoppers.

What do you enjoy most about your work?

I enjoy the vast amount of data that we analyze.

And what do you usually do after work?

On my way back home, I enjoy the Außenalster!

Which colleague would you take with you to a lonely island and why?

I would take Carole, because she is a great swimmer.

Which attraction/part of Hamburg do you think you should definitely visit and why?

This summer, I discovered Wilhelmsburger Inselpark, though it looks quite artificial. Altes Land is lovely, too, especially during spring when the apple trees bloom.

Which apps could you never do without and why?

Spotify.

These blogs/websites belong to my daily reading:

I read a lot about vegetarian cooking!

On the web you can find me here:

Try your favorite search engine.

Akanoo wins bevh Rising Star Competition 2016

bevh-Rising_Star_Gewinner_1

Akanoo won the bevh Rising Star Award 2016 yesterday with a big lead over the other contestants. For the fifth time the German E-Commerce and Distance Selling Trade Association (bevh) honoured innovative e-commerce distributors and service companies, that have successfully established a new product or business model. The winner was chosen via public vote from the panel of e-commerce experts in retail and technology providers. An expert jury had previously selected the three most impressive companies among all applicants and invited them to a final pitch at the .EXE Workshop Conference in Hamburg.

The e-commerce automation platform Lengow and the data analytics software Webdata Solutions presented their business in front of the conference audience and e-commerce decision makers together with Akanoo. Every visitor could vote for one of the three start-ups. Akanoo won with over 60 percent of the votes, followed by Webdata Solutions in the 2nd and Lengow in 3rd place.

Thanks to bevh for the beautiful event and congratulations to the two other finalists Lengow and Webdata Solutions.

You have seen our pitch and want to learn more about on-site targeting? Get in contact now!

Akanoo Nomination For bevh Rising Star

bevh Rising Star

Akanoo reached the final round at the startup competition Rising Star. With the Rising Star award the German E-Commerce and Distance Selling Trade Association (bevh) honours innovative distributors and service companies, that have established a new product or business model successfully on the market.

Next to Akanoo the e-commerce automation platform Lengow and the data analytics software Webdata Solutions are nominated. The final announcement of the winner will take place at the workshop-conference for interactive sales and commerce .EXE on 26th April. The three nominated companies will present their business ideas in an elevator-pitch at the conference. Afterwards the conference participants will vote on the winner.