Category

Allgemein

How to create a truly enjoyable survey …

By | Allgemein

How to create a truly enjoyable survey… 

and why this is the key to get deeper and more actionable insights.

Do you remember those hot days in the summer of 2018? At one of those days, we were about to launch a mobile survey with 16 questions, 8 of them being open-ended. My colleague David, who were sweating next to me, said “This will never work! Who wants to respond to 8 open-ended question, on a mobile phone?”. Still, we had no other choice. We even added two more questions, “did you enjoyed this survey?” and “why did you answered this way?”

The outcome was a surprise. On average, about 50% of respondents in conventional surveys indicate they enjoyed the survey. Our mobile survey achieved a 91% enjoyment rate. Wow!

You will ask yourself “why?”. This is exactly what we asked the respondents. The most frequently mentioned reasons were “it was easy”, “these were good questions” and “it was fast”. (actually, the average time was 5 minutes, which is a rather long response period on a mobile phone).

Are you satisfied with these answers? We took them to the test. If a reason is a true reason, it must be possible to use its information to better predict the enjoyment rating (in a multivariate model). This is a ground-truth formulated by the fathers of causality research such as Nobel Prize winner Clive Granger.

This is why we build a flexible (machine learning based) prediction model and discover something quite exciting. None of the frequently mentioned reasons predicts well why respondents enjoyed the survey. Instead, two other reasons are highly predictive.

The text response “the survey was simple” was mentioned by 10% of respondents and was a strong predictor of survey enjoyment. Further, text that can be summarized as “i liked the opportunity to describe my sentiments with my own words” was mentioned by 9% and is an even stronger predictor.

Recent attempts to increase respondents’ engagement in the industry typically circle around varied and entertaining ways of surveying and to use more gamification approaches. But what we now learned is that the key is less about building more sophisticated ways of asking, it is more about to keep it really simple and give respondents the option to communicate in the most authentic way possible: to use their own words.

You may ask “ok, but is coding open-text responses economically even feasible and attractive?” Recent technologies that combine Natural Language Processing and Machine Learning can automatically code large volume of text. The system we use must be trained by a human coder. Then its output has comparable predictive power with human codings.

“Ok, but what does this mean for improving the quality of insights?” you might ask yourself…. good question!

Let’s look at what we have learned. We have learned that we receive the most authentic and high quality information from respondents (e.g. your customers) by asking very simple questions and let people respond in their own words. At the same time, we now have the technology to discover hidden drivers of success in those simple text responses.

Of course, we can not only learn what is truly enjoyable with surveys, we also can learn why customers are loyal or more likely to recommend your brand. We can also learn why they consider certain brands or choose a new product innovation.

We can learn valuable insights that are key to success by using extremely simple and cost effective research methods. What could be more exciting?

If you like to dive deeper … e.g. how to make this work for your NPS program, your Brand Tracking study or whatever you think drives value in your company, just let me know and I’ll be happy to chat.

Yours, Frank

Why customer join, is not why they stay: Artificial intelligence reveals the key loyalty drivers for mobile provider

By | Allgemein

Why customer join, is not why they stay: Artificial intelligence reveals the key loyalty drivers for mobile provider 

Mobile customers may choose a carrier because of a proper connectivity. Interestingly this reason is not the most important for their loyalty and why customers recommend their carrier to others. Instead, the key driver of customer loyalty and recommendations are attractive plans. This is a core finding from Success Drivers’ recent category CX study.

The results exemplify the power of the company’s solutions to reveal hidden drivers of customer loyalty. Two research instruments, NPS.AI and Brand.AI use two simple sources of information: a standardized scale to measure customer loyalty and an open text question asking “why?”. The study was done in collaboration with codit.co – an AI-powered platform for the analysis of open-ended survey questions.

“We asked 1,000 American customers: “Would you recommend your mobile carrier to a friend”. And we asked “why?” They most frequently answered “because of the good network connectivity/coverage” said Frank Buckler, CEO of Success Drivers and continues: “What we discovered is that this reason – although most mentioned – has a minor impact on customers loyalty and recommendations.”

The Success Drivers AI technology revealed the factors that were more likely to lead to carrier loyalty and recommendations. Although connectivity/coverage is known to be the one of the dominant factors why customers choose a carrier, it has only moderate impact on loyalty and recommendations. AI algorithms found that being satisfied with the actual plan is the largest reason why customers stay at a carrier and the dissatisfaction with their plan is the major reason for leaving.

The methodologies used in this study rely on coding text into content categories. The research tool is an AI-powered platform, which automatically codes text with great precision after training by human content experts. Its novel deep-learning-based engine enables fast, inexpensive and accurate analysis of large-scale open-end surveys or other text sources.

The figure below shows a more detailed picture of this study. The vertical axis is the frequency with which a loyalty factor has been mentioned, and the horizontal axis shows the impact of factors on customer loyalty. The results show the challenge of conventional survey research: customers find it difficulty in prioritizing what drives their behavior. Self-learning AI-based algorithms help finding the complex correlations between customer responses and their level of loyalty and likelihood to recommend.

NPS.AI: A simpler CX platform with deeper insights

Conventional CX programs are descriptive, not predictive. They do not provide insights into why NPS score has changed or reveal the hidden, non-obvious success drivers. NPS.AI’ powerful web-based dashboard shows it all: Trended NPS, our proprietary AI-powered impact of key loyalty drivers, and an explanation of the change in NPS compared with the last wave.

Established high-reputation brands like SONOS are convinced: The NPS.AI solution helps to simplify, and safe costs, and most importantly it delivers insights capable to inform top-level business decisions.

More details: https://nps-ai.success-drivers.com

Why Good Sex is not enough: Artificial Intelligence reveals the key drivers of a happy relationship

By | Allgemein

Why Good Sex is not enough: Artificial Intelligence reveals the key drivers of a happy relationship 

We just released a study that explored the hidden drivers of relationship satisfaction. The results exemplify the power of the Success Drivers solutions to reveal hidden customer loyalty and brand consideration drivers using two simple questions: a scale to measure satisfaction and an open text field asking “why.” The study was done in collaboration with codit.co – an AI-powered platform for analysis of open ended survey questions.

We asked spouses or partners why they are satisfied or dissatisfied with their relationship. They most frequently mentioned “joint activities” and “lack of cooperation”. What we discovered is that those reasons have a minor impact on their relationship satisfaction.

The Success Drivers AI technology revealed the factors that were more likely to lead to fulfilled relationships. Physical affection is important and prevents the leading reason for dissatisfaction (after cheating) in a relationship. However, our AI algorithm found that physical affection is a necessary, but insufficient condition for relationship satisfaction. Importantly, happy relationships are associated with higher reports of good communication among partners and unexpected helping behaviors that demonstrate a partner is willing to go the extra mile.

The methodologies used in this study rely on coding text into content categories. Codit.co is an AI-powered platform which automatically codes text with great precision after training by human content experts. Its novel deep-learning-based engine enables fast, inexpensive and accurate analysis of large scale open end surveys or other text sources.

The figure above shows the whole picture of this study. The vertical axis is the frequency that a relationship factor was mentioned, and the horizontal axis shows the derived impact of those factors on relationship satisfaction. The results show the challenge of conventional survey research: namely that humans have difficulty knowing and expressing what drives their behavior. Self-learning AI-based algorithms help find the complex correlations between stated responses and the level of satisfaction.

More details on the applied methodology, the codit.co text analysis platform and the new generation of Net Promoter Score programs (NPS.AI) will be provided by a series of webinars that codit.co and Success Drivers will host together this fall.

The next webinar takes place on August 28, 11am ET. Free registration over https://goo.gl/2fGoKr

Read here more on NPS.AI

Multi-client study PRICE.AI

By | Allgemein

Test all your prices at scalable costs

Manufacturers and retailers have to manage hundreds or thousands of prices. Established price measurement methods such as conjoint are too expensive on this scale. Simple methods like PSM produce rationally biased results.

The solution: In a standardized online survey of potential customers, Price.AI uses an “implicit” survey method to measure unconscious willingness to pay. This is the scientifically established way to reveal unconscious associations. We apply this to the price range to be tested (default 7 price points). A 5-minutes questionnaire later, our algorithm calculates a multiple validated price-demand function and the profit-maximizing price range.

The Multi-client-Study: 

  • Starts beginning of June 2018
  • Express your interest until May 17th.
  • Your input: In our Excel template, you write down the product description, the price points to be tested, the name of the product image to be shown and select the screener question belonging to the product.
  • You will receive: Price-sales function in Excel per product. When indicating the product cost of goods sold, we deliver a price-profit function.
  • Participation costs: 500 USD per tested product plus 1.5k setup fee (volume discount scale an request)

For further background information please contact Frank Buckler, PhD. Just send an email to: buckler@success-drivers.com

Evidenced: Which Creative Techniques Drive your In-Market Performance

By | Allgemein

Invitation: We accept inquiries for participation of your brand until Feb 25 2018

For six product categories, our syndicated study ”Creative.AI“ in 2017 has revealed which creative techniques and emotional triggers drive purchase intent – short and long-term.

In 2018 we are expanding this approach by not limiting it to purchase intention anymore. Research shows that intention and purchase are correlated, but– depending on product category, research approach and context – this correlation can be substantially increased.

In this study we show the specific impact of intentions on purchase behavior for your specific product category. We make causal relationships transparent and maximize the power to predict the in-market impact of creatives – even before anything has been produced.

With your participation in this syndicated study you will profit from the following deliverables:

  • You test your commercial with a target group that already resides within the purchase funnel instead of surveying people that will not have a need for the product or service in the near future
  • You determine the short-term and long-term sales impact of creative techniques (e.g. celebrities, spokesperson, brand song, or 100 others), emotional triggers (indulgence, family love or 130 others) and physiological emotions (happiness, surprise, sadness, etc.).
  • You receive insights that will show you which content features of your commercials have which emotional impact on your customers; this will enable you to improve creative and message design

How we proceed

  1. Copy-testing of commercials in an online survey: We simultaneously test a large range of competitive spots in order to cover a large variety of creative techniques. The tested ads will be profiled with an objectivized content analysis.
  2. Purchase intention measured with improved methods; scientifically validated and state-of-the-art
  3. Re-contacting of respondents after exposure and determining their purchases
  4. Driver analysis (self-learning, AI-based) to quantify the causal importance of creative techniques

Costs for participating companies depend on reachability of your target group, the purchase frequency of your product category and the target region. Fees start at $6k per tested commercial.

For product categories with infrequent purchases we will also provide data on information seeking behavior (in addition to purchase behavior).

Schedule: Declare your interest shortly. We finalize agreements until Feb 25, 2018. Field phase is March to April. You get descriptive copytest results in April and final study reports end of May, 2018.

Contact: Frank Buckler, PhD., buckler@success-drivers.com

The Blueprint for Successful Advertising

By | Allgemein

Creativity does not follow rules. Correct. However, if it wants to be successful, it needs clear guide rails – just as the creativity of architects is guided by the laws of statics. But what does actually work – “Celebs?”, “Voice over?”, “Which music?”, “Community or pleasure?”, “Problem or solution?” – There are hundreds of such questions to which we have now clear answers.

CONCRETE GUIDE RAILS FOR SUCCESFUL ADVERTISING? NO CHANCE!

75% of the advertising effectiveness does not depend on the budget or the media plan, but on the ‘quality’ of the creation. However, it remains largely unclear what exactly is needed for good advertising.

The result: Only 16% of the ad spots are actively recalled and only a third of all creations lead to an attractive ROI.

On the other hand, many successful campaigns have an ROI of a factor of 10 or more. Where can you invest your money better than in such advertising? Nowhere, if – and only if one knows the blueprint of successful advertising.

THE SOLUTION: PROFILING OF CREATIVE TACTICS

There are a variety of test methods for advertising. Copy test surveys, eye tracking, EEG measurements, and a lot more. All these methods are good, useful and important.

However, there are two things they fail to do: they just measure the response to advertising, but not the effect of individual design parameters. They collect data which, on their own, do not support any causal inferences. For example, any coke advertisements will tend to perform better in all parameters – regardless of how good the advertisement is by itself.

The now piloted solution combines three modern approaches into one:
1.The copy test survey, which also measures the emotional impact of  advertising.
2. The quantitative content analysis: Experts analyze each ad and encode the creative vehicles (celebrity, humor and 90 others) and emotional triggers (indulgence, family love and 130 others).
3. Determination of the causal contribution to the success of each tactic: Here, self-learning algorithms of artificial intelligence are used to prove what works and what does not.

EUREKA MOMENTS THANKS TO THREE-RATE EXPLANATORY POWER

The success factor analysis makes it possible to measure how valid the established “blueprint for successful advertising” is. Compared to the best models to date, this explains 3.1 times better how advertising works. Dozens of eye-opening findings in our study prove this. Here are three examples:

The study shows that emotions are the key to success. Even rational messages are not processed until emotions are aroused. However, only certain emotions are productive others again extremely destructive.

Furthermore, the study shows that in each product category different basic emotional messages are effective. On top of that, the messages that are usually used in their respective industry are usually not effective – just like the “belonging” message for alcoholic beverages, the “trust” message from financial service providers or the “relief” message for medication.

Also, frequent techniques like celebrities or voice-over generally are not effective. In fact very specific and old tactics are often those, which work wonders.

These and dozens of other findings will fundamentally change advertising in the future.

HOW TO IMPLEMENT THE METHOD

The quickest way to get started is the Quick-Audit, which Success Drivers offers free of charge. Furthermore, the Copytest used in the study is an inexpensive first step. With the help of the blueprint established in the study, concrete areas of improvement are illustrated by means of the Copytest.

In the medium term, it is advisable to carry out own category-specific “deep dive” studies and thus develop your own specific know-how. The use of the sales data of the advertised products makes it possible to accurately predict in the model the sales-increasing effect of an ad.

However, the best research will only bear fruit if it is translated into sound creative strategies, and if the marketing management as well as the creative agency stands fully behind it. For this reason, we have developed specific workshop concepts. Here, we use our database of more than 60,000 tried and tested exemplary spots to illustrate the different varieties of certain basic emotional messages and other creative parameters.

More details about the results can be found in the following video recording of my speech at Esomar’2017:

BrandGrowth.AI: A proven, evidence-based strategy for more growth of your brand

By | Allgemein

BrandGrowth.AI: A proven, evidence-based strategy for more growth of your brand

In recent years neuroscience has provided the scientific proof. The result of human decisions can be detected in the brain already one to seven seconds before it has entered consciousness. Well, 90% of our decisions can be explained by unconscious emotions, attitudes, and instincts. However, most advertisements still focus on the potential product benefit and other USPs. The consequences are far-reaching.

Consumers make thousands of choices every week and are exposed to thousands of advertising contacts. It is neither practical nor relevant for people to process and understand all of them, or even to enter into a “relationship” with a relevant number of brands.
The scientific models developed by Ehrenberg and Bass decades ago are based on a different understanding of consumers – with success. They predict consumer behavior and thus the market structure very precisely. Large empirical studies have confirmed this over the last decades in hundreds of product categories across the different cultures.

Ehrenberg und Bass: A milestone in marketing research

In his book “How Brands Grow”, Byron Sharp – the current director of the Ehrenberg Bass Institute – has outlined on the basis of cross-cutting empirical studies which far-reaching consequences for convenient marketing strategies the new consumer understanding has.

He proves that placing the focus on winning new customers is more effective than binding the existing ones. Coca Cola makes more than 50% of its sales with customers who buy a coke only 2 to 3 times a year. Just about every brand makes a relevant revenue share with a large number of occasional customers – the “long tail”. This opens up an enormous growth potential.

Sharp also empirically proves that positioning on the basis of emotional differentiation is overestimated and he shows that branding is about more fundamental things. It’s about

  • being saliently perceived and anchored in order to enter the choice at the moment of the decision,
  • ensuring that advertising contents are placed in a context by arousing the right emotions and it is about
  • maintaining continuity so that the brand image is strengthened over time rather than diluted.

As a third point, Sharp proves that it is important to keep the brand “available” in all respects in order to exploit the full potential. “Make it Easy to Buy” is the motto. This not only includes the dissemination of CPG products on the market, but also their visibility there or the avoidance of various, in hindsight partly trivial buying obstacles.

BrandGrowth.AI: Success Drivers’ solution for more growth of your brands

In the growth package “BrandGrowth.AI” we have compiled our patented analytics solution. It includes the key steps that we recommend to all brands in the market to strengthen their position in the long term and generate growth.

1. Be Broad –Exploit as much potential as possible

Step 1 – Define your playground

Is Head & Shoulders a shampoo or a men’s anti-dandruff shampoo? The brand is bought by men and women most of whom in addition have no relevant dandruff problems. With selective targeting, the brand would not be the world’s best-selling shampoo brand. But how do you find the right scope, the appropriate definition of the “playing field”? Exactly for this, there are methods such as the intersection analysis that can help.

If you believe your brand owns a real USP, then this would be an occasion to think about expanding or changing the category understanding. Why? Rational decision-making processes are more likely to be found in the choice of product category and less in the selection of brands.

Step 2 – Find you’re optimal set of Category Entry Points

You pass McDonalds while driving and you still did not had breakfast. If the brand is not associated with breakfast, you’ll drive by without noticing you have missed an option. Did you ever ask yourself why tomato juice is mainly consumed in planes? Flying is (for whatever reason) a central Category Entry Point (CEP) for tomato juice.

Brands grow if they are associated with more relevant CEP’s. We identify qualitatively potential CEP’s, measure the relevance using implicit methods and identify how your brand is already associated with relevant CEP’s.

Step 3 – Category.AI – Find your category drivers

For the buyers of a category, the lowest common denominator needs to be found, which allows for convincing as many buyers as possible. Our AI-based driver analysis covers the central levers of each product category by means of simple brand surveys.

Here we find that the core driver for the beer category is the refreshing experience. We discover that in the skin cleansing category, the fragrance is the central purchasing indicator. We demonstrate that a bank should primarily focus on its expertise in financing and investment.

These core drivers are the lighthouse for your marketing. At times it seems too trivial that a beer refreshes, a shampoo smells good and a bank provides good advice. It is therefore often the rule instead of the exception that brands move too far away from this lighthouse.

This case study, using the example of the SONOS brand, illustrates the procedure.

2. Build Brand – Maximize the “mental availability”

Branding aims to strengthen the mental availability by strengthening memory structures related to brands. This is achieved by the use of unique brand elements (logo, colors, figures, melodies, stories and associations) which are constantly used over years and decades. However, this anchoring can only be achieved if it is possible to involve advertising content emotionally – or when the brand is shown in emotionally involved contexts (keyword sponsoring).

Step 4 – Creative.AI – Identify the DNA of successful advertising:

Our Creative.AI solution is designed to find out which emotional archetypes and creative techniques are capable of involving customers in a positively emotional way and thus cause a deeper anchoring of the brand elements. Find out more about Creative.AI here

Many large-scale studies show that at least 70% of the advertising effect depends on how well it is done. The evidence-based optimization of creative content is therefore of crucial importance. Only when this step has been completed does it make sense to optimize the distribution of marketing budgets more precisely:

Step 5 – Media.AI – Media planning with maximized reach

According to current state of research it is a fact that the first advertising contact has the highest ROI. Repeating contacts are only reasonable if there is no alternative target group available with similar response to the advertising. Byron Sharp showed with great empirical evidence that advertising has greatest impact at light category buyers (long tail), typically not within the “core target group”.

That’s why impactful media planning is less focused on razor-sharp targeting but maximizing reach within the group of potential category buyers. Media.AI samples individual media usage profiles among category buyers and than computes (by leveraging A.I.) those combination of media channels, types and times that maximizes reach with an existing budget.

Step 6 – Mix.AI – Mix modeling in the digital age

Marketing Mix Modeling is an established method. However, it requires an update. In the digital age, channels mutually define each other to a growing extent. A TV spot generates Adword clicks – an indirect effect that must be taken into account. AI-based driver analyzes not only depict this indirect relationship, but also find hidden synergies, i.e. interaction effects between channels. More about our MMM approach.

Step 7 – SLC-Brand Tracker: The 3 central KPIs always in focus

Brand trackers nowadays primarily measure the brand image in order to check whether the brand is still perceived according to definition. However, it is much more crucial to track how salient, “likable” and continuous the brand management is:

  • Salience (brand salience): It measures how many anchor points the brand occupies to be recalled and retrieved at the required time. In order to optimally measure the brand’s salience, a separate measuring instrument must be developed that scans all brand “cues”.
  • Likability: Is the brand associated with positive emotions and with the basic benefit of the category?
  • Continuity: Is the brand communication based on existing memory patterns or  is it diluted by building new ones?

3. Reach out – Maximize the “physical availability”

Under the concept of “physical availability”, we understand that the brand is in the perception field of the consumer when required and meets certain knock-out criteria (for example, desired package size, price category, etc.). The degree of dissemination on the market here is only one of many obvious components.

Step 8 – RNB.AI: Identify the reasons NOT to buy

Brands whose apparent success is “convenience” often have eliminated certain knock-out criteria. Shining examples are PayPal, Google Search, Nintendo Wii or the iPhone. Interestingly, to marketing experts it is by no means obvious in advance what exactly these criteria are. More detailed research is required. We recommend a two-step approach

  1. Qualitative research: depth interviews, focus groups or communities
  2. Quantitative exploration: Our AI-based driver analysis reads from a quantitative survey concluded for this purpose which criteria are actual “show-stoppers”.

The approach has been designed and tested for a current product category on the market as well as for new product concepts.

Step 9 – Sales.AI: Maximize the impact force at the POS

The impact force at the POS can be increased by a variety of measures: displays, in-store furniture, demo events, sales training, price promotions, etc. However, their effectiveness is not directly reflected in the data. Thus, some measures have long-term (training) and other short-term (price promotions) effects. Sales.AI establishes a convenient record and evaluates the ROI of these sales promotion activities.

Step 10 – Price.AI: Find price points that are acceptable to the majority and maximize long-term returns

A price point that is too low will jeopardize the long-term “survival” of a product – too high a price reduces the actual availability and prevents the full profit potential from being exhausted. The price is not only the easiest to change marketing parameter it is also known to be the most effective one. Every dollar increase is one dollar profit.

Price.AI is based on an implicit measurement of the price stability and models the causes, conditions and leverage of the willingness to pay. This makes it possible to find and manage optimal long-term price points.

The BrandGrowth.AI Audit – Find out what works for you

BrandGrowth.AI is a collection of tools that are applied at the most important parameters for brand growth. It provides focus, highest validity and transparency in the jungle of methodological possibilities.
We offer new customers a free-of-charge audit workshop in which we will discuss with you what your individuals needs are to be able to move your brand to the next level. With a little bit of luck, it will be the same as with T-Mobile USA, who, 4 years after the implementation of Category.AI., have doubled their market share and today are making record profits.

How T-Mobile Doubled its Market Share through Artificial Intelligence

By | Allgemein

How T-Mobile Doubled its Market Share through Artificial Intelligence

What can you advise a brand that has a qualitatively worse product, operates in a largely commoditized market and has suffered massive losses for years? The way out of the misery reminds of the fantastic stories of Baron Munchhausen, who pulled himself out of the swamp by his own hair.

Our analytical know-how was the decisive basis for an unprecedented success story in which the market share doubled in only 4 years and today profits are at a record high. Read how similar effects will be possible for you as well.

The Brilliant Frontal Attack

The year was 2013. T-Mobile was virtually doomed to die. Years of massive losses lay behind it. But the mother company found no buyer for the unloved daughter. John Legere took over leadership of the company as CEO and was given free rein from the headquarters. Things could hardly get any worse.

His strategy was to move heaven and earth, so that customers would again opt for T-Mobile. He reduced the prices, terminated the contract obligations and threw in top smartphones simply for free. He framed the whole story with a communication story “The Un-Carrier”, in which he celebrated T-Mobile as the Robin Hood of the mobile phone companies.

The customer growth was not long in coming. Yet the question remained: Why do customers come to us? Is it the price, the contract obligation, the smartphone or the emotional attraction of the new positioning? Conventional methods gave contradictory answers. One could get any response by filtering the insights accordingly. Does this sound familiar to you?

Artificial Intelligence in Action

T-Mobile addressed this question to us because the question of why, the question about the central cause of success is our specialty. Conventional correlation and regression analyzes had not been able to provide convincing answers.

We resorted to a nationwide survey of customers and non-customers. Such “brand trackers” are carried out by almost all companies on a regular basis. In these, we measure the willingness to change to T-Mobile, the perception of the brand image, the extent to which the new positioning has already been learned, and, above all, the assessment of the customers and prospective customer regarding the central purchasing criteria such as network quality, price, contract obligation, service or devices.

Still, data is useless if it is not possible to identify the true cause-and-effect relationships from them. Simple correlations only provide spurious correlations and conventional statistical methods are neither able to take into account indirect cause-effect relationships nor to correctly represent the unknown facets of the relationships (nonlinearities and moderation effects).

We carried out a universal structural modeling – a causal machine learning procedure. It is distinguished by three captivating properties:

  1. Measures causality not correlations, and thus avoids the classic pseudo insights.
  2. Reveals indirect effects and thus, in contrast to classical driver analyzes and regression approaches, is able to estimate the true overall effect.
  3. Is self-learning and thus models previously unknown nonlinearities and interactions.

Importance of the Findings: Immense

The findings of the analysis were manifold. The most important and at the same time most surprising insight was this:

Neither price contract obligation nor top equipment has a massive direct impact on the purchase. It was the “Robin Hood” positioning that attracted the customers. The pricing, lack of contract obligation or equipment extras were merely levers that fuelled the perception of this positioning. Findings that remained hidden with conventional analyzes.

Instead of preserving the growth by fighting yet another price war, we recommended to constructively strike out new paths, which substantiate the positioning. Exactly this was the path of the company in the coming years. T-Mobile offered, for example, customers of the competitors to take over the fee in case of premature termination. Or the flat rate was extended to calls abroad.

“Success Drivers provided us the confidence to advise our business partners on the most effective and efficient means for sustaining our brand momentum.” David Feick, Ph.D. Director Consumer Insights, T-Mobile USA

Learnings

The case study shows the crucial importance of a precise understanding of what motivates the customers in their market to act as they do.

Industry knowledge of experts is a necessary but not sufficient prerequisite, because the variety of simultaneously acting cause-and-effect relationships cannot be deciphered by a human being manually. Causal artificial intelligence can help, as the example shows impressively.

The example also shows how conventional statistics in such real-world questions often are not sufficient. Success factors influence each other (indirect effects), change their meaning depending on the context (interaction) and depending on their occurrence (non-linearity). Artificial Intelligence and Machine learning algorithms can support management in extracting valuable information from the ocean of data.

Data quality cleaning using machine learning

By | Allgemein

Data quality cleaning using machine learning

New, more effective ways to avoid bad response effects, sampling bias and fluctuating KPIs

Some respondents click rapidly through the questionnaire, some erratically. The random sampling in a survey is often a far cry from a chance selection. Excessive quota systems increase the problem instead of solving it. Internal customers often ask themselves why an NPS (or other KPIs) has decreased or increased and can find no reasonable answer. Misgivings regarding the data quality can destroy the trust in an entire study. Luckily, there are new, effective ways to counter this.

The bad response effects example

Many market researchers have developed their own “practical” rules to counteract this. Overly short survey times or too many affirmations lead to the exclusion of a respondent. Occasionally, test questions are introduced to check whether a respondent is “paying attention”.

The problem is: what exactly is the right survey time that should be used to exclude respondents? Which rate of agreement can be used to differentiate a fan from a dishonest respondent? What exactly does the test question say about the survey behavior of the person?

The answer to this question is not known and can hardly be answered from data using typical means.

Self-learning analysis methods, also known by buzzword “machine learning,” offer the chance to learn of complex relationships between indicators and their effects based on data.

Bad response indicators include, for instance, the length of the interview, the response to a trick/consistency question, the number of identical responses in a row, and many more. The trick no longer tries to assume the meaning (= influence) of these indicators for the results (arbitrarily) but rather uses the data to discern the causal influence. This is possible with cause-effect analyses based on machine learning (especially NEUSREL). The analysis yields the following results. The graph shows that customer satisfaction is extremely high for interview lengths under a certain number of minutes (the scale is normalized to 100 here).

diagrammsatisfaction_2

This makes it possible for us to derive reasonable decision rules. Below a time of 20 (normalized) the satisfaction is obviously massively biased. On the right side, you will find a scatter plot containing the same variables, and you see a very different relationship there. Why? It could be the case that the fast respondents are younger people who are per se less satisfied with the provider. Machine learning delivers causal, corrected insights instead of the typical spurious correlations.

If the data is used for a driver analysis, it is recommended to not exclude the suspicious cases at all. Because the bad responses are described well by the indicator variables, a driver analysis based on machine learning (especially NEUSREL) simply filters out the bad response effects. The sorting out of cases would introduce a new but unknown selection bias to data.

The sampling bias example

Distortions in random sampling can occur, for example, as a result of a willingness to participate among the subjects that does not differ randomly. By measuring indicators of this willingness to participate, such as age, sex, income or a trait such as the “need to communicate,” machine-learning-based cause-effect analyses can determine the relation between indicator and response behavior. If there are too many seniors in your sample, for instance, you then know how this will distort each question and consequently you can correct this effect.

For the use of data in driver analyses, it is sufficient to make good indicators part of the driver model, because it automatically subtracts out all bias effects. However, this is possible only for driver analyses based on machine learning (especially NEUSREL). In this manner, an overly strong quota system, which itself creates new quota-related biases, can be avoided.

p.s. More on NEUSREL and Universal Structural Modeling here2

Lowering returns quotas: breakthrough via causal analysis.

By | Allgemein

How leading returns researcher Professor Walsh discovered effective levers to lower the returns quota via causal analysis.

The extremely high returns rates in e-commerce are an unsolved problem – a problem that represents a challenge to managers due to the tight margins in online retailing. Traditional BI analysis results very often turned out to be misleading in the past. The reason is known as “spurious correlation”.

Success Drivers  therefore identified, for example, that horizontal striped clothing was sent back less often. A classic spurious correlation: horizontal stripes are found particularly on cheap products, for which the returns rate is generally lower. Under otherwise similar circumstances, horizontal striped products are sent back MORE not less. This is a finding that can only be discovered via causal analysis.

The returns whisperer

For this very reason, Professor Walsh from the University of Jena has been involved in causal analysis with his team of researchers. His various studies provide many new insights into what drives product returns.

In a study, which appeared in the Harvard Business Manager, he and his co-invetigator found that the use of penalties and obstacles does not pay off in the long run. Their study showed a clear strategic path: return rates can be reduced by providing customers with relevant product-related information and making the buying process as hassle-free as possible. Virtual try-ons, avatars and specific product evaluations from peers are very helpful in this respect. The latter is a factor with very high ROI as the investment is low.

Self learning causal analytics covers up

With the current study the research team around Gianfranco Walsh were also able to identify a key driver of customer returns. Using consumer survey data and causal analysis, they found that the customer-related reputation of an online retailer alone has a significant effect on the return rates.

This reputation is determined by the following components: customer-orientation, being a good employer, being financially sound, perceived product quality and social responsibility. Although this measure may seem very abstract, the new causal model enables very precise measurement of the effect of bad press and to quantify the effects on the margin.

“There are also many company-specific returns levers. They may be identified most successfully via self-learning causal analysis – as enabled by the NEUSREL software”, Walsh recommends.

p.s. More on NEUSREL and Universal Structural Modeling here.