Personalization – Digital - A Practical Implementation Perspective – Part-3

Continued from Part-2…….

Segmentation Engine

Segmenting the audience is primarily used when it comes to email and print media campaigns. It can also be useful for targeting specific promotions to a specific segment of customers. Even though it sounds complex, a basic segmentation engine can be nothing more than a simple UI so that you can dissect the target audience in various groups like women, millennials, baby boomers, sport enthusiasts etc. Let us take a concrete example to explain the segmentation process. if you merchandizing team wants to have a special promotion for a particular brand of clothing line for women living in warmer climates, then you want to take the entire dataset and filter it down to something like this

For entire customer dataset where demographics = southern states and gender = female and customer propensity toward this line of clothing > 70 (on a 1-100 scale)

The above will give you a customer segment which can be used to target for this particular campaign. This way creating a segmentation UI can be useful to your business, product and merchandizing teams to rapidly create and deploy segments with minimal intervention from your engineering teams. 

UI/UX Design challenges

There are multiple ways to display personalized products, campaigns, promotions to the end customer. A few common ways are

·      Home page

·      Search Results Page

·      PDP (Product Display Page)

·      Shopping Cart Page

·      Promotional Emails

·      Promotional Print media campaigns

My recommendation again is to start small in all the above areas. For e.g., in the home page, you could allocate a slot for products recommendations based on past purchases. In the email, I would recommend not to allocate more than a 2-4 slots for personalized promotions/products.

The aim is not to clutter the pages or have too many options for personalization which can lead to erroneous products being displayed or also lead to too much algorithm processing which in turn slows down the display process. Especially if you want to send burst emails (1 million/hour), it takes a lot of precision and processing time to get the right promotions in the right slots. Hence, it is always recommended to start small and then gradually expand. 

ESP – Email Service Provider

For your email campaigns you need to pick an ESP that has not only capabilities of sending large volumes of emails but also burst emails (> 1 million/hour). Picking a right ESP with these capabilities is the key. For e.g., if it is supposed to snow for the next 1-2 days, you want to send out a burst email to all your winter segmented customers about show gear promotions. The capability to receive and process data and sending them out in the stipulated SLA is very important. Hence, before you pick your ESP, ensure you have all your requirements clearly laid out and then match it to their capability list. 

Operational Considerations 

Reporting

Reporting in personalization is about getting data about data. The success and failure of your personalization program depends on the capability to measure and report on each and every single aspect of your personalization flow. Clicking on the recommended products, promotions and it leading to conversion which means the customers actually were influenced by the personalization and bought the items recommended. Each of these flows have to have reports in real-time generated so that the end of day metrics can tell you what is and what is not working which in turn will help you in tweaking your algorithms and refine your customer 360 records.

Alert and Monitoring

Typically, in software engineering projects, alerting and monitoring always comes as an afterthought. As the saying goes, the strength of the chain is determined by the weakest link. Having a robust monitoring and alerting framework will give the SRE (Site Reliability Engineering) team robust tools to ensure that this whole machinery flows smoothly as designed. This framework should be designed and co-developed by both the SRE and the core engineering teams so that the minute details for every single flow can be monitored from both a software engineers and an operational engineers perspective. 

Conclusion

Once you have a plan for all the above systems, then you can determine what can be bought and what be built in your buy vs build strategy. I would also caution you against all the fancy Big consulting firms who will come and give you high-level strategy’s which have excellent PowerPoints but lack the depth. My recommendation would be engaging with firms who have actually done it end-to-end and ask them to recite what worked and what did not. The latter will give you more idea as to the qualifications of the firm. I believe it’s the mistakes you make are what make you a better person as long as you don’t repeat them.

As the saying goes, the strength of the chain is determined by the weakest link, hence, please note that every single item I pointed above has equal importance. Please ensure you pay attention to the details and detail flows and not just a high-level plan. Taking an agile approach is good as it gives the chance to pivot quickly as and when needed. 

Good luck in your personalization journey, if you have any questions / clarifications, please feel free to ping me at Sudhir.ganti@gmail.com