Building Great Personalization: Components (Part 2)
Component number 2 — Great Delivery Driver (content placement)
Even though a great segment builder is a core element of any personalization tool, it’s definitely not going to be enough. At the end of the day, a personalized experience is about the experience. Therefore, you need something which would allow you to a) build the experience b)deliver it on-time c) deliver it to the right segment.
There are 3 categories of tools to build and deliver a personalized experience:
- Landing page based
2. Custom code injector based
3. Dynamic component-based
Category number 1
Custom landing pages are quite simple personalization delivery tools. In some cases this type of personalization could be done even without additional tools — just build 2 separate pages and drive your ad/email/partner traffic there. Or leverage the free Google Optimize version to have a split test of LP.
When to use and who will benefit.
- Split test of destination pages
- Split test of the whole landing pages
- Split test of layouts/colour schemas
- Message synchronization b/n ad and landing page
- Personalization of channel experience
- Personalization for search key-words
Suites for companies at the beginning stage of the personalization program who yet need to prove the value of personalization to the stakeholders. Who wants to improve their CRO and build some profit out of it to bring the personalization program to the next level.
Category number 2
Something interesting begins when you need to deliver not the whole page experience, but a partially modified experience on the page. Like a matching header, hero, or CTA to a segment.
This tool should be able to:
a) pull segment from the segment building tool and recognize the user “X” belongs to a segment “A”. Read more about the great segment builder in part 1 of this article.
b) allow you to set up a rule to match experience and a segment
c) execute the rule based on the segment trigger (basically, recognize that this user id “X”/device id“X” belongs to the segment “A” and that segment “A” should see that piece of code “Z”)
c) allow you to hide the “default” experience and replace it with the piece of code “Z”.
So, you select an HTML element on your page, prepare a code to replace it with, the back-end runs a JS with the rule: “if this group of user ids visits that page, then hide this HTML element and replace it with that HTML element”.
Why use it
- Test hypothesis around message/color/outline matching for a segment.
- Test hypothesis around message/color/outline matching for a segment in a particular funnel stage.
- Match experience (creatives and copy) for a particular segment to improve conversion rate (lead gen, purchase, subscription etc)
- Match experience (creatives and copy) for a particular segment in a particular funnel stage to improve conversion rate/ decrease drop-off rate (lead gen, purchase, subscription etc)
Who will benefit?
This simple “if-then” approach would suit the companies who have some resources to invest in purchasing/building more expensive tools and who are on a specific digital maturity stage:
- Whose stakeholders already believe in the power of personalization.
- Whose marketing team works with segmentation, channels, and funnels.
- Whose marketing department consider themself as KPI-driven and start their campaigns with KPIs in mind.
- Whose assign CRO as a priority and allocated time and resources (dev, UX, creative) for that.
Category number 3
This personalization delivery tool is even more exciting! This tool would allow to set up a “placeholder” for a content piece to be delivered dynamically. It would contain 3 elements:
1.A “catalog” holder (a mini-database or a Digital Assets Manager) of content, products, or page elements (like CTA, heroes, headlines) with their attributes like for
Content: title, preview of text body, image link, tags, etc
Product: sku, price, category, etc;
Page elements: CTAs, banners, colors, shapes, etc.
2. HTML / JSON renderer that would extract those content/product attributes from the database and display it in the right order and in the right location on the page.
3. “Brain” which would work behind the scenes to retrieve the right content for the right segment/right person based on the observed behavior. Most of the time this “brain” feeds on 3 types of data: customer historical CRM profile, product/content catalog, and a real-time web analytics feed which captures user engagement with the product id/content-id.
“Brain” observes engaging behavior across multiple visitors on the website, builds clusters out of customer attributes, and produces weighted affinity attribution which predicts the next-best product/content for the cluster (customers with attribute X mostly engage with content with attribute Y). When a new visitor arrives to the website, it assigns him/her to one of the clusters, and based on the most successful historical behavior recorded by the cluster, it would serve a similar content to that visitor (this visitor has attribute X, most likely s/he would like to engage with content with attribute Y).
Why use it
The beauty of this method would be in the ability to deliver not the pre-set experience, but the predicted best experience for the visitor. It’s the getaway for an AI-driven 1:1 experience.
By leveraging the 1:1 AI-driven content allocation, we eliminate the situation when we split-test content delivery, (so 50% of visitors receive a “loser” version of the experience, poor thing! ) and eliminate “guts-driven” content delivery to a person.
When to use and who will benefit.
This category 3 tool would suit those companies which are ready in invest in a massive flow of content production. They either have very cheap labor — the army of copywriters and graphic designers with a strong content plan and guidelines built for personas, or they leverage AI text and image generator tools. Those assets should be properly classified, tagged, and stored in a DAM or database and fed to the content/product catalog.
Component number 3.
Great Analytics
I guess it’s pretty straightforward why you need a great web analytics tool to bring a great personalization to life. But not all web analytics tools were created equally.
They would be different by its capability to:
- Capture users web behavior
- The more out-of-the-box web tracking features available, the better, but customization is important as well. And not all tools allow you to customize tracking easily.
- Very important — the ability to leverage a user-id synchronization script which would allow you to store the cross-device customer id (like email) into a cookie and match it with device id (like browser uuid and mobile device id) via internal user-id
- Be able to store and retrieve web behavior signals by the id.
2. Make the analysis as painless as possible: either by allowing you to download the user behavior data or have a space to run analysis and build visualization in the tool itself.
3. Have a feedback loop to evaluate the performance of the personalization. Be able to pull data to see personalization impressions and engagements per segment.
- At minimum to display results of the experiments
- Show results with a minimum inferential statistical analysis and/or regression analysis.
- Show results with a bayesian analysis (calculating confidence interval).
- Build and display a feedback loop for the AI model and affinity weights.
Thank you for reading to the end. Hope you liked my article and find it helpful! If you want to learn more about personalization and/or have a great project in mind you need a consultation on and/or just want to host a chat on the clubhouse platform, ping me @ https://www.linkedin.com/in/marinazub/
