When it refers to customer info, privacy and compliance are critical. Consent channels are just as crucial as sale funnels in this new privacy-first era. It’s no longer only about gathering useful information. Consumer permission must be incorporated into all procedures that use customer data when organisations create targeted and personalised experiences. 

Customers and business analysts will be able to honour their customers’ consent within existing workflows in the Dynamics 365 Customer Insights customer data platform, thanks to new integrated consent functionalities in Dynamics 365 Customer Insights (CDP). Without writing code, administrators may link to consent data and establish the rules for its use. These features will be available in 2021 release wave 2 as a public preview. 

Segmentation and consent regulations

To learn how to connect consent data and apply default consent rules to the segment data flow in Dynamics 365 Customer Insights, check out this video.

In order to access the embed, you must accept cookies from the embed’s site. Accept cookies by activating the link and viewing the embedded content.

https://cloudblogs.microsoft.com/dynamics365/it/2021/11/02/streamline-consent-management-with-dynamics-365-customer-insights/

Customer Insights now offers complete data management, including customer-identity data collection, unification, and enrichment, thanks to its consent enablement features. We use dataflows to provide a low-friction experience for integrating and harmonising consent data from several sources in this release. Everything begins with Consent Centre’s self-service data preparation service (see below).

The integrated consent data can be mapped to actions and inclusion or exclusion rules, and these rules can subsequently be used to steer business activities. Consent rules operate as global “guardrails” in segmentation, allowing profiles to be automatically filtered based on client consent and preferences, as seen in the example below.

This incorporation of consent data into the CDP can also aid data analysts. They can use consent data to train and improve machine learning models like customer churn and customer lifetime value forecasts and accuracy (CLV).

Steps to take next

These features will be gradually rolled out across regions as part of the public preview. Visit the help hub for more information on how to make the experience work in your setting.

Visit the  dynamic 365 insight portal to learn more about the new consent enablement capabilities.

Check out the free trail  of Dynamics 365 Customer Insights if you haven’t already.

To share your ideas, go to the Dynamics 365 Customer Insights Community forum.

The post Streamline consent management with Dynamics 365 Customer Insights appeared first on Microsoft Dynamics 365 Customer Insights.

Gaining and maintaining clients in these unpredictable times necessitates a fuller insight of their brand and interest affinities in order to provide relevant, personalised experiences. Dynamics 365 Customer Insights uses proprietary Microsoft data from Microsoft Graph to help us better understand your clients and then segment them into groups for relevant and effective experiences. If you own a home remodelling business, for example, knowing which tool brands your consumers prefer will help you prioritise the products you sell.

What is Customer Insights and how does it work?

Connect to your customer data sources first, then merge them to get a complete perspective of your customers with Dynamics 365 Customer Insights.

After that, you can add your clients’ brand and interests affinity to your data. There are two steps to setup: To enhance your data, choose brands and interests. Go to Data > Enrichment in the Dynamics 365 Customer Insights app. Select the Enrich my data button from the Discover tab’s Brands or Interests tile. Then, either choose your industry and let the app choose appropriate brands or interests for you, or make your own five-brand selection. Make a diagram of the demographic segment fields. From your customer profile data, map the demographic segment fields for age, gender, and location. The enrichment is essentially a look-alike model depending on the demographic section of each consumer.

The customer profiling is enhanced by identifying the brand or interest affinity for that group based on the demographic segment information. After you’ve finished setting up, you can either execute the enrichment process right away or schedule it to run with your next scheduled refresh. You can review the results to see how many customer profiles were enriched and by which brands or hobbies once the enrichment process is finished, which normally takes a few minutes. Under Data > Entities and on individual customer profiles, you can see precise information about brand and interest affinities. Now that you’ve completed the setup, you can use the affinities data to provide targeted and tailored experiences to your clients. Do you want to know more? Details can be found in the documentation.

Modern people have more material, purchasing channels, and brand selections than they did in the past, and their expectations for excellent customer service keep rising, especially in downturns. To find a way to keep loyal consumers, businesses must first comprehend, value, and believe them, then serve them effectively through smooth, personalised experiences across all channels.

Companies require a consolidated view with a single perspective of each customer to deliver these types of experiences, and this source of info must provide not only a present view of customers but also a potential future perspective.

Why is it so difficult for businesses to obtain such information? Data is pouring from everywhere, thanks to the creation of so many customer contact points, the evolution of a company’s infrastructure through time, and the exponential data expansion of recent years.

Data is frequently fragmented across various systems and departments, making it difficult to collate, organise, and mould into a holistic image of a client, let alone extract useful insights that empower staff with predictive, next best actions.

Microsoft Dynamics 365 Customer Insights can help with that. Dynamics 365 Customer Insights is a user-friendly and adaptable customer data platform (CDP) that lets you consolidate all of your customer data from a variety of sources to create a single, authentic view of your customers.

You can use Customer Insights to:

• With pre-built connectors, you can link data from any source, like point-of-sale, transactional, behavioural, and consumer preference data, or connect to a data lake to intake data and construct the view.
• With out-of-the-box and custom models, anybody from data analysts to marketing and sales professionals can use the data to uncover insights and take action.
• Use Microsoft Power Platform, Microsoft Dynamics 365 applications, and other connections to fourth solutions to consume, adapt, and expand these capabilities, as well as Azure Machine Learning and Azure Synapse for advanced analytics.
• So that all client data is safely managed and adheres to GDPR legislation and high privacy standards, rely on the trusted, hyper-scale Microsoft Azure platform. Dynamics 365 Customer Insights is also accessible on the Microsoft Government Cloud, which is designed to meet the government’s compliance requirements.

Connect to a wide range of data sources using pre-built connectors

Pre-built connectors for a variety of first- and third-party data sources, such as Dynamics 365, Azure SQL Database, Azure Blob storage, Salesforce, and SAP, make importing data easier. Check out the documentation to learn more about data sources.

To develop unified consumer profiles, map, match, and integrate data.

The mapping, matching, and merging of data into a single, unified image of your consumers is a three-step procedure. Go to the documentation to learn more about the map, match, and merge processes: Unification of data

To gain a better understanding of your customers, expand your customer profiles.

You can add data from other sources to your customer profile, such as Microsoft Graph or firm data from Leadspace or CivicScience. This information can subsequently be used to develop actionable categories, such as for high-value clients. Alternatively, broaden the categories to include lookalike modelling for targeted marketing and sales initiatives.

Make predictive insights a reality.

On top of your single customer profile, Customer Insights allows you to take advantage of out-of-the-box AI capabilities. The subscription churn insight, for example, can help you estimate whether a client is likely to stop using your company’s subscription products or services.

You may train the model using data from your unified customer profile, and then have it retrain itself on a regular basis to keep improving outcomes.

Azure Machine Learning and Azure Synapse Analytics connect consumer data with enterprise data to design, train, and fine-tune machine learning models for unique predictive insights.

At every consumer touchpoint, expose data and insights.

Customers’ unified data and insights can be surfaced and consumed at every touchpoint where employees interact with them, such as the following:

• Create a Power App that displays extensive customer data such as transaction history, customer-specific segments, and suggested next best actions.
• To access unified profiles directly within Dynamics 365 business applications, download the customer card add-in from AppSource.
• To export data for usage in other third-party programmes, use third-party connections like LiveRamp, Facebook Ads Manager, or other pre-built connectors.
• For your own in-house apps, leverage APIs to augment the customer view.
• Use Power BI to connect consumer data and generate bespoke dashboards and reports that help you make better decisions.

Recent enhancements

Here’s a quick rundown of newly announced features:

• Common Data Service is now available for building data sources in Customer Insights, allowing you to import analytical data from Dynamics 365 apps. Learn more about how to connect data in a lake maintained by the Common Data Service.
• Export to LiveRamp as a destination – Connect with over 500 platforms across the digital, social, and TV ecosystems by activating your data in LiveRamp. In Customer Insights, you can learn more about how to use LiveRamp.
• Become a member of Microsoft Teams – Connect to Microsoft Teams and use an add-in from AppSource to look up customer profiles with the help of a bot. Learn more about how to use the Teams plug-in.
• With Microsoft Dynamics 365 Online Government, you can meet stringent security and compliance standards. Dynamics 365 Customer Insights is now available on the Microsoft Government Cloud, a platform designed to help US government organisations achieve their stricter compliance requirements. Learn more about Microsoft Dynamics 365 Online Government and Dynamics 365 Customer Insights.

Keep up to date with what’s coming up in the next six months and what’s just arrived by following these links.

Become a member of our community and get started right away.

The Dynamics 365 Customer Insights community page has the most up-to-date information, including forthcoming events, featured blogs, and other relevant pieces. Also, take a look at the following links:

• ms/CIdoc How-tos and more detailed information about capabilities can be found in the documentation.
• ms/TryCI To develop unified customer profiles, start a free trial, see a demo, and import your own data.
• View a video overview of Dynamics 365 Customer Insights.
• The post Dynamics 365 Customer Insights: Unlock insights and fuel personalised experiences appeared first on Dynamics 365 Blog.

Dynamics 365 Customer Insights is a platform that allows you to combine customer data from several sources into a single view. This unified data is a great place to start when creating bespoke machine learning (ML) models to create important business indicators. We saw how to use Azure Machine Learning (AML) Studio to create custom ML models that operate with the unified data from Customer Insights in a previous article. We provide AML Studio–based model pipelines for three often encountered use cases to speed up the initial model development:

Dynamics 365 Customer Insights is a platform that allows you to combine customer data from several sources into a single view. This unified data is a great place to start when creating bespoke machine learning (ML) models to create important business indicators. We saw how to use Azure Machine Learning (AML) Studio to create custom ML models that operate with the unified data from Customer Insights in a previous article. We provide AML Studio–based model pipelines for three often encountered use cases to speed up the initial model development:

• Customer lifetime value (CLTV)
• churn analysis
• Productrecommendations

In this blog, we’ll go through these models in further depth.

Incident at a Hotel

The models in this post will be based on the Contoso Hotel scenario. CRM data is collected at the Contoso Hotel and consists of hotel stay activities. The information regarding the dates of stay for each registered consumer is stored in this data. The data also includes information on the booking, room kinds, and spending details, among other things. From January 2014 to January 2018, the data covers roughly four years.

Customer profiles provide the second set of data. Each customer’s personal information has been stored in these profiles since they were originally registered with the hotel or since their first visit. Name, birth date, postal address, gender, phone number, and other details are included. The hotel’s additional services are included in the third set of data. For example, spa services, laundry services, Wi-Fi, courier services, and so on. Each registered customer’s information is likewise logged. The typical usage of services is associated with a stay, although it is not essential; in some circumstances, consumers can utilize services without being in the hotel.

Analyze the Churn

Churn analysis may be used in a variety of scenarios (e.g., retail churn, subscription churn, service churn, etc.). We’ll look at service churn in this case, especially in the context of hotel services, as indicated above. The model will give insights into all the phases in constructing a custom model using Azure ML and Customer Insights, even if it does not explicitly cover all of the cases. It will also provide a functioning example of an end–to–end model pipeline that can be used to create any other sort of churn model.

Churn is a word that has several meanings.

Depending on the situation, churn can be defined in a variety of ways. In the instance of hotel activities, our definition is that a client should be categorized as churned if he or she has not visited the hotel in the previous year.

The AML Studio experiment may be imported from the gallery, as mentioned in our previous blog. The import blocks that read these tables from the Azure blob storage location are shown below.

Featurization

We start by identifying the raw characteristics that will have some causal influence or impact on the label, based on the concept of churn. Then we convert this raw information into numerical features that may be used in machine learning models. As a result of the data integration in Customer Insights, linking these tables is as simple as utilizing the “Customer ID” produced by Customer Insights.

The featurization for churn analysis model construction can be a little hard. The topic is commonly framed as a static task, such as picture classification, yet the underlying data is not static, unlike a set of photographs. The information is time-based, with new hotel activities being logged on a daily basis. As a result, featurization should account for this and build static features from dynamic data. In this scenario, we use a one-year sliding frame to derive several characteristics from hotel activities. We also use one-hot encoding to break out category variables like room type and booking type into independent features.

The following is the complete list of features:

Number	Original_Column	Derived Features
1	Room Type	RoomTypeLargeCount, RoomTypeSmallCount
2	Booking Type	BookingTypeOnlineCount, BookingTypePhoneCallCount
3	Travel Category	TravelCategoryBusinessCount, TravelCategoryLeisureCount
4	Dollars Spent	TotalDollarSpent
5	Check-in and Checkout dates	StayDayCount, StayDayCount2016, StayDayCount2015, StayDayCount2014, StayCount, StayCount2016. StayCount2015, StayCount2014
6	Service Usage	UsageTenure, ConciergeUsage, CourierUsage, DryCleaningUsage, GymUsage, PhoneUsage, RestaurantUsage, SpaUsage, TelevisionUsage, WifiUsage

choosing a model

The next step is to determine the best algorithm to utilize after the feature set is complete. The majority of the characteristics in this scenario is categorical traits. Typically, decision tree–based models do well in such scenarios. Neural networks may be a superior alternative for solely numerical characteristics. In such cases, the support vector machine (SVM) is also a strong contender; however, it requires a lot of tuning to get the best results. The first model we chose is “gradient boosted decision tree,” followed by SVM as the second model. AML Studio allows you to compare and contrast two models via A/B testing. It’s usually advisable to start with two models rather than one to get the most out of this.

AML Studio’s model training and assessment process is depicted in the diagram below.

In addition, we use a method known as “Permutation Feature Importance.” This is an important feature of model optimization. All of the built-in models are black box models that provide little to no insight into the influence of any one element on the final forecast. The feature relevance calculator employs a unique method to calculate the impact of various characteristics on a model’s ultimate conclusion. The relevance of each attribute is scaled from +1 to -1. The negative impacts imply that the related features have an illogical effect on the outcome and should be eliminated from the model. A higher positive effect means the characteristic is making a significant contribution to the forecast. These numbers should not be confused with correlation coefficients because they are two separate measures.

completing the Customer Insights integration

We may develop a predictive service as mentioned in the previous blog after the training experiment is completed and the generated metrics are satisfactory.

Ascertain that the projections are accompanied with the Client Insights customer IDs. The predictions may then be exported to the same blob storage and re-imported into Customer Insights.

Prediction of customer lifetime value (CLTV)

The computation of customer lifetime value (CLTV) is one of the most important measures that a company may use to evaluate and classify its customers. Knowing your consumers is crucial in the hotel industry. Understanding the difference between guests who add value and those who don’t, for example, may give critical information to hotel management. This form of segmentation can aid hotel management in determining which elements they should priorities and develop in order to please their high-paying clients against less critical characteristics. These choices can have a significant influence on sales and profits. The CLTV will be defined in this case as the total money brought in by the client throughout the specified time period.

CLTV’s definition

The CLTV of a client as of today will be defined as the total dollar amount the customer is likely to spend in the next 365 days or 1 calendar year. To forecast this figure, we’ll utilise data from all consumers over the last three years.

Featurization

The featurization will be quite similar to the churn scenario in this case; however, the label and projected values will be different than those stated previously.

Selection and training of models

Because the projected value is a positive valued continuous variable, predicting the CLTV is a regression issue. We use Boosted Decision Tree Regression as one approach and Neural Network Regression as another to train the model based on the feature attributes.

completing the Customer Insights integration

The output of the CLTV model is attached to Customer IDs and fed back into Customer Insights for additional analysis, as outlined in the churn model.

Next Best Action or Product Recommendation

In the context of a hotel, product recommendation is recommending to consumers the hotel’s services. The goal is to match clients with the right services so that their consumption is maximized. The issue is comparable to that of video streaming service consumers receiving movie suggestions.

Definition of Next Best Action or Product Recommendation

The aim is to maximize the dollar amount of service usages by providing hotel clients with the finest matched services based on their interests.

Training and Featurization

We’re combining the hotel ServiceCustomerID with Customer ID in a similar way to the churn model in order to produce consistent recommendations per Customer ID.

The data is obtained from three separate sources, as stated earlier, and characteristics are extracted from them. When compared to churn or CLTV situations, the featurization for the problem of recommendation is different. Three sets of features are required as input data for the recommendation model. The first set of features reflects the customers’ prior usage of services, the second set contains the specifics of each service, and the third set represents the customers’ information.

To train the recommendation model, we utilize the Matchbox Recommender algorithm.

The Train Matchbox Recommender model accepts three input ports: training service usage data, customer description (optional), and service description, as shown in the diagram above. The model may be scored in three distinct ways. One is for model evaluation, which involves computing an NDCG score to rank the assessed elements. The NDCG score in this trial is 0.97. Depending on the business requirement, the model can be scored on the full recommendable service catalogue or solely on products that consumers have never used before.

Looking at the distributions of the recommendations over the full-service catalogue, we see that the top services to be suggested are phone, Wi-Fi, and courier. This is in line with what we discovered from the service usage data distributions: