Customer success

ANWR Group uses AI to automatically tag product images

At a glance

  • AI-supported image recognition reliably identifies the properties of products
  • Manual effort and error rates are significantly reduced
  • Special procedure considerably reduces the effort needed for development.

The business situation of our client

Artificial intelligence (AI) enables process automation that cannot replace humans, but can relieve them of many routine tasks. Our AI solution for the ANWR Group, one of Europe’s leading retail cooperatives for clothing and, among other things, operator of the shopping portal “schuhe.de”, is an outstanding example for this potential. Originally, the company’s employees were required to manually record the characteristics of products on an ongoing basis in order to be able to provide necessary filter options to costumers – for example shoe type, brand or color – when searching for products. This approach is not only time-consuming, but also error-prone – especially when a large number of new products is introduced during the change of season.

We therefore developed an AI-supported image recognition for shoes, which automatically tags incoming goods based on the product photos alone. For that, we opted for a procedure that saves effort and leads to excellent analysis results at the same time.

 

Alexander Hock, Managing Director ANWR Media GmbH

“AI is a complex topic for sure. But together with ORAYLIS we were able to quickly achieve remarkable results.” Alexander Hock, Managing Director ANWR Media GmbH

The solution for our client

Our model for image recognition is based on a neural network we trained with corresponding image data. Such a training phase can take several weeks or even months if you have to start from scratch – even with specialized hardware.

That’s why we used a pre-trained image recognition model we could tailor to the requirements of the ANWR Group. First, we adapted the existing keywords to the categories, colors and other characteristics of the shopping portal. This step was followed by several test phases during which we fine-tuned the model. This procedure allowed us to save a large part of the training effort.

How data turns into new values

The project essentially shows that even complex AI applications do not necessarily require a lot of effort. Despite the extremely short training phase, the ANWR Group now has a highly specialized model for image recognition at its disposal, saving a significant amount of time when it comes to tagging products. The model is able to determine product characteristics with astonishing precision. The error rate is now lower compared to manual data entry. And: image recognition is not limited to professional product photos; it works just as well with photos taken with a smartphone.

turn your data into value.

At a glance

  • AI-supported image recognition reliably identifies the properties of products
  • Manual effort and error rates are significantly reduced
  • Special procedure considerably reduces the effort needed for development.
Jens Kröhnert
turn your data into value

Let’s get started!

Do you also want to use artificial intelligence to achieve much more with less effort?

Join #teamoraylispeople

Shape the world
of data with us

CUSTOMER SUCCESS

Predictive Policing: How criminals can be tracked using Artificial Intelligence

At a glance

  • AI solution relieves scarce police personnel resources
  • Crime hotspots are identified with foresight and task forces are controlled in a targeted manner
  • Success rates are significantly higher compared to experience-based approaches

The initial situation of our client

Artificial intelligence (AI) can make processes more efficient in practically every area and can therefore significantly reduce the needed human effort. For example, we developed an AI solution for the police department of one of the largest cities in Germany to predict crimes – also called predictive policing. Just like the rest of Germany, the city was struggling with far too few personnel resources to get a grip on the rising number of burglaries by organised gangs. Statistical methods for identifying potential hot spots hardly helped, as they practically made no use of the extensive case data that had been documented.

The solution for our client

The design of our solution shows that developing AI applications does not necessarily require high investments in new technologies. Since the police department’s IT infrastructure was already based on Microsoft technologies, we simply used the already existing SQL Server as a basis.

With the Analysis Services, we built a customised forecasting software that directly accessed the police’s internal databases, such as case details or geographic and demographic data. The Reporting Services were used to create an easy-to-use forecasting tool as well as a forecasting quality assessment. Finally, a special classification procedure identifies the quadrants with the highest probability of burglary and displays them in colour on a map in Power BI. The resulting forecasts are generated automatically on a daily basis and distributed to all police directors in the city.

How data turns into new values

With the diverse police data as a basis, the forecasting software makes an effective contribution to the prevention of residential burglaries by intensive offenders. Currently, the prediction of future burglaries is up to seven times better than previous experience-based approaches. The police forces show a targeted presence in the identified danger areas. In this way, potential perpetrators are deterred and residents are made aware of possible subsequent offences.

As the data used is constantly supplemented by new findings, the system continues to learn and provide ever more accurate insights. The new solution also proves to be an efficient repository of many years of police experience. Last but not least, the forecasting software can be maintained by the police themselves and can also be extended to other areas of crime.

turn your data into value.

At a glance

  • AI solution relieves scarce police personnel resources
  • Crime hotspots are identified with foresight and task forces are controlled in a targeted manner
  • Success rates are significantly higher compared to experience-based approaches
Dirk Ohligschläger
turn your data into value

Let’s get started!

Do you also want to use AI applications for your individual tasks in order to achieve much more with less effort?

Join #teamoraylispeople

Shape the world
of data with us

Customer success

How an energy supplier prevents customer churn

At a glance

  • Data-based churn model maps customer behaviour in the future
  • Potential churn candidates are automatically recognised
  • Model optimises itself independently through artificial intelligence and machine learning

The business situation of our client

The market for energy suppliers is characterised by a similar level of performance and quality of the suppliers as well as very low switching barriers for consumers. The willingness to change suppliers is correspondingly high. Companies react to this decreasing customer loyalty with so-called churn management, whereby “churn” is an artificial term made up of “change” and “turn”. In other words, it is about measures that prevent customer churn.

The comprehensive use of data can considerably increase success while reducing costs. This is illustrated by the example of one of our clients. At an early stage, we set up a learning churn model for the now established market entrant that not only reliably identifies potential churn candidates. At the same time, it ensures that the customers who fit the products are addressed.

The solution for our client

The model thrives on the linking of diverse sources and variables on a modern, digital platform. The goal is to represent each customer as accurately as possible. For example, existing customer feedback from the call centre and from marketing campaigns flows into the model. The respective consumption data is also included, as an unsuitable tariff can also cause dissatisfaction. In addition, information on the profitability of the contract is included. All this data is compared with price comparison portals – i.e. are there competing tariffs that could be more interesting for the customer? And: How does the company actually compare to the competition in individual tariff segments?

Further information is provided by campaigns such as “Customers recruit customers”. Because active “recruiters” are considered relatively safe. The same applies to those who have been recruited. On the other hand, if such opinion leaders leave, the danger increases that other customers will follow suit. Last but not least, external sources are taken into account, such as population density and market penetration in certain postcode areas. For example, in regions with low population density and penetration, customers migrate more quickly through the recommendations of neighbours.

How data turns into new values

On this basis, the churn model independently provides a timely warning for each profitable customer with potential switching intentions. Accordingly, the provider can take targeted – and usually successful – countermeasures with individually tailored marketing measures.

The special feature of the model is that it continuously optimises itself through Artificial Intelligence and Machine Learning functions and adapts to changing conditions. The starting point is the existing data from contract renewals and contract terminations. This “learning material” is continuously searched for regularities and patterns. Over time, relevant variables emerges, such as dissatisfaction with prices or certain services. The ideal time to approach the change candidate can also be determined in this way.


turn your data into value.

At a glance

  • Data-based churn model maps customer behaviour in the future
  • Potential churn candidates are automatically recognised
  • Model optimises itself independently through artificial intelligence and machine learning
Jens Kröhnert
turn your data into value

Let’s get started!

Would you also like to use your data to inspire your customers and retain them in the long term?

Join #teamoraylispeople

Shape the world
of data with us

Customer success

How to make your production more efficient with real-time data

At a glance

  • Manufacturer of special pumps monitors its entire production in real time
  • Companies, customers and employees benefit from more transparency and improved processes
  • Solution connects local SAP ERP with services in the Azure Cloud

The business situation of our client

Production processes must be as efficient as possible in order for companies to survive in global competition. This was also recognised by one of our industrial customers, a global leader in the manufacture of special pumps. Together, we implemented a monitoring system based on real-time data for its entire production line and achieved a large number of improvements as a result. Because: Up until this point, neither the production status nor the progress of an order could be tracked. The finished goods usually reached the shipping department without notice.

The solution for our client

Our customer’s production process involves pushing the picked individual parts of a pump from one production cell to the next on a transport trolley. In our solution, each trolley receives its own RFID tag, which is linked to the relevant order. The individual manufacturing cells are equipped with corresponding RFID gates. We process the resulting data stream via a real-time route based on services from the Microsoft Azure cloud: An Azure Event Hub receives the data and distributes it further. The data is then analysed in real time via Azure Stream Analytics. We have also connected a local SAP ERP with quality-assured reference data. Only by linking it to the SAP data do the findings from the real-time information gain concrete significance. Finally, the data is visualised via a real-time monitoring dashboard in Power BI.

How data turns into new values

The data-driven real-time monitoring generates new values for our customer on very different levels. The status quo of each individual pump can now be precisely tracked. Likewise, those responsible can obtain an overview of all current orders in real time. As a result, the various production steps can be coordinated much more precisely. For example, the efficiency of the employees in the production cells has increased significantly, as they are now informed about upcoming orders at an early stage. And of course, the company’s customers are also much more satisfied: they receive much more precise information about the status and delivery date of their goods.


turn your data into value.

At a glance

  • Manufacturer of special pumps monitors its entire production in real time
  • Companies, customers and employees benefit from more transparency and improved processes
  • Solution connects local SAP ERP with services in the Azure Cloud
Jens Kröhnert
turn your data into value

Let’s get started!

Would you like to analyse your data in real-time to make business and production processes more effective and customer-friendly?

Join #teamoraylispeople

Shape the world
of data with us

Customer success

How mobile providers can predict the value of new contracts

At a glance

  • A simple forecasting model makes the success expectations of new contracts and contract extensions transparent
  • Deviations from actual results amount to a maximum of two percent
  • The success of new products and cannibalisation effects can also be recognised at an early stage.

The business situation of our client

Assessing sales performance has always been a special challenge for mobile providers: usually, the revenues and contribution margins from new contracts and contract extensions can only be precisely quantified after a few months. Concrete sales figures for upcoming contract periods are difficult to estimate simply because customers often change their tariff or take advantage of supplementary options during this period. At the same time, sales successes are often incorrectly estimated and rewarded.

Therefore, one of our clients, a leading mobile phone provider in Germany, wanted more transparency and planning security. With relatively simple means, we developed a predictive model that reliably predicts the desired figures.

The solution for our client

The provider has an extensive data pool on the usage behaviour and thus the contribution margins of its customers. However, it is not that easy to predict these figures, as behaviour tends to be variable at the beginning of the contract and only stabilises after a few months. In addition, there are also differences between the customers, depending on the respective region and the sales channel.

We used this diverse detailed data to build our forecast model. A relatively simple technical set-up was sufficient for this. The entire solution is based on an already existing Microsoft SQL Server as well as the Analysis and Reporting Services. In essence, the model applies average calculations to similar customer data and then outputs a corresponding contribution margin.

How data turns into new values

Despite its simple structure, the forecasting process delivers very accurate estimates of future sales directly when a contract is signed. Likewise, the sales performance can be evaluated in a stable manner already on the following day. The deviations from the actual results amount to a maximum of two percent. This also allows our client to recognise the success or failure of new products at an early stage and make immediate adjustments. In addition, cannibalisation effects across different sales channels can be identified very quickly.


turn your data into value.

At a glance

  • A simple forecasting model makes the success expectations of new contracts and contract extensions transparent
  • Deviations from actual results amount to a maximum of two percent
  • The success of new products and cannibalisation effects can also be recognised at an early stage.
Jens Kröhnert
turn your data into value

Let’s get started!

Would you also like to use your data to optimise your business and unlock new value?

Join #teamoraylispeople

Shape the world
of data with us

Customer success

Retail business: How to optimise your advertising expenses

At a glance

  • Advertising success monitoring model measures the results of marketing campaigns
  • Findings enable optimal item mix for campaigns
  • Success of planned campaigns can be predicted by means of key figures

The business situation of our client

“I know half my advertising is money thrown out. I just don’t know which half.” – The much-quoted wisdom of Henry Ford has lost none of its relevance for the retail industry. On the contrary: due to saturated markets, the industry is mainly occupied with competing with each other for customers by means of massive advertising campaigns. In the meantime, a large part of the marketing expenditures of retail companies is allotted to this item – and the trend is still rising.

Unlike Henry Ford, however, retailers today can use their data to determine the share of misinvested advertising money relatively precisely and reduce it in a targeted manner. For example, we have developed an advertising success control model for a large German retail company that makes exactly this possible.

The solution for our client

Technologically, the model is based on a very powerful Microsoft platform. With the help of the solution, our client can first store and calculate the baseline sales per day and shop for each of their items. These normal values are then compared to the sales in a promotional period. The difference results in the turnover achieved with the respective promotional measure – traceable down to the receipt level and for each individual article. It is also measured whether a promotional item promotes a deadweight loss.

How data turns into values

Our customer can now evaluate the success of his advertising in detail. For each promotion, the top sellers, slow sellers and slow sellers can be precisely identified and stored in the system. A combined view of the different advertising channels is also possible. In addition, external sources such as competitive, weather or demographic data are used.

All these findings are finally incorporated into an index that measures the expected success by means of defined key figures already during the planning of campaigns – such as an offer brochure. If the evaluation is negative, our client can readjust accordingly. In this way, the optimal article mix can be defined for each advertising medium and channel, while also preventing cannibalisation effects.


turn your data into value.

At a glance

  • Advertising success monitoring model measures the results of marketing campaigns
  • Findings enable optimal item mix for campaigns
  • Success of planned campaigns can be predicted by means of key figures
Jens Kröhnert
turn your data into value

Let’s get started!

Would you also like to use your data to invest your advertising money optimally?

Join #teamoraylispeople

Shape the world
of data with us

Customer success

Mobile providers: How to inspire your customers with tailor-made offers

At a glance

  • Digital “Interaction Advisor” supports staff in sales talks
  • System automatically delivers action guidelines and individual product suggestions
  • Tailor-made offers inspire customers and optimise contribution margins

The business situation of our client

Mobile phone providers are faced with a dichotomy: on the one hand, individual contracts must generate the highest possible contribution margin. On the other hand, it is important to inspire consumers with tailor-made offers and to retain them in the long term. The example of one of our clients shows how these seemingly contradictory requirements can be effectively combined with the targeted use of data. Using Microsoft technologies, we have built a system that automatically defines offers based on comprehensive customer and tariff information, from which both the company and its customers benefit equally.

The solution for our client

Our solution is a so-called Interaction Advisor in the contact mask of the company-wide CRM. It provides the hotline and shop agents in direct customer contact with all important information about the respective purchase history and the current contract value. At the same time, the salesperson receives a guide to action as well as a selection of individual product and solution proposals with exact contribution margin forecasts. The system makes use of hundreds of predefined tariff and option combinations. Likewise, possible negotiation margins as well as the agent’s reward for successful contract conclusions or extensions are taken into account.

How data turns into new values

The information provided enables the salesperson to advise the customer much better and to respond to his or her individual needs with tailor-made offers. The customer feels personally addressed and in good hands with the provider. Thus, the new system contributes to higher customer satisfaction and customer loyalty. At the same time, the solution makes an effective contribution to churn management. And finally, the company benefits from optimised contribution margins.


turn your data into value.

At a glance

  • Digital “Interaction Advisor” supports staff in sales talks
  • System automatically delivers action guidelines and individual product suggestions
  • Tailor-made offers inspire customers and optimise contribution margins
Jens Kröhnert
turn your data into value

Let’s get started!

Would you also like to inspire your customers with tailor-made offers from which you as a company derive the greatest possible benefit?

Join #teamoraylispeople

Shape the world
of data with us

Customer success

Publisher increases circulation with data-driven forecasts

At a glance

  • Forecast model reliably predicts campaign success at every campaign level
  • Relevant key figures can be controlled in a targeted manner
  • Circulation figures and sales are sustainably increased

The business situation of our client

The digital transformation is not always a danger for publishers – it also opens up many opportunities. Because well-maintained databases can be converted into completely new values for the business with modern analysis platforms. For example, we have set up a data-driven forecasts for a large publishing house that reliably determines industry-relevant key figures – such as shelf life, conversion rates or circulation – for each stage of a sales campaign. Normally, this is only done retrospectively at defined points in time. Thus, the results of campaigns are often only available after several months. So there is hardly any possibility to specifically influence the success of ongoing campaigns.

The solution for our client

Our forecasting model combines existing order, customer and campaign data in each campaign phase in different ways and enriches them with additional information. Already in the planning stage of a campaign, initial statements can be made about response rates, conversion behaviour and resulting print runs on the basis of offers, advertising media, premium information and seasonal factors. These findings are refined with each subsequent campaign stage. For example, general information about the customers, such as geographic data or special interests, is added during address selection. Finally, with the start of the campaign, specific customer behaviour can be included, such as digital usage behaviour, payment information or interaction with customer service. The model thus continuously adapts its predictions to current developments.

How data turns into new values

With the data-driven forecasts, our client can reliably predict the success of its campaigns at any time based on significant key figures. This opens up the possibility of intervening if necessary and readjusting the measures – with corresponding positive effects on the relevant key figures or the specific conditions. Furthermore, short- and medium-term revenue forecasts, cash flow expectations and forecasts on terminations are possible. And what is decisive: for the first time, long-term goals, such as the improvement of conversion rates, shelf-life curves and payment rates, can be controlled in a targeted manner.


turn your data into value.

At a glance

  • Forecast model reliably predicts campaign success at every campaign level
  • Relevant key figures can be controlled in a targeted manner
  • Circulation figures and sales are sustainably increased
Jens Kröhnert
turn your data into value

Let’s get started!

Do you also want to use your data to control and increase the success of your sales campaigns?