Adobe Analytics Plus ChatGPT for Ideation: Part 2 – Personalisation

Sometimes, the scope of personalisation can be overwhelming, especially if you’re in a small team or just starting out. 

How do you start doing meaningful personalisation? What are the best practices?

In this two-part edition of the ‘Adobe Plus’ series, I’m going to look at how ChatGPT can be a useful tool to give you a kickstart and a fresh perspective on ideation. In Part 2, I’m going to look at its application in personalisation.

What’s the catch?

  • ChatGPT is only as good as the context you give it.
  • You know your organisation’s data and audience best.
  • Use ChatGPT’s suggestions in conjunction with your own data about your customers to validate and make decisions on what to test.

It goes without saying, but don’t upload sensitive or personal data to ChatGPT.

Use ChatGPT to explore personalisation ideas

Start by giving ChatGPT as much context as possible – tell it at least your industry and goals, but the more context, the better!

Act like a conversion rate optimisation manager for a retail company with a website that sells shoes.

It’s highly likely you know your website audiences already, but you can use ChatGPT to provide another perspective in identifying audiences and what they might be looking for on your website.

What are 5 likely audiences on the website and the top 3 considerations of each audience?

Example response:

 Budget Shoppers

Considerations:

  • Affordability: Budget shoppers are concerned about finding shoes within their budget. They want to easily locate the sale or clearance section and know if there are ongoing promotions or discounts.
  • Shipping Costs: High shipping costs can be a deterrent for budget-conscious shoppers. Clearly communicate any free shipping thresholds or promotional shipping offers.
  • Return Policy: Knowing the return policy is crucial for budget shoppers. They want assurance that if the shoes don’t meet their expectations, they can easily return them without incurring significant costs.

Do you know how best to identify your audiences for targeting and personalisation?

How might one identify the audience ‘Budget Shoppers’ based on their website behaviour?

Example response:

Identifying the audience of “Budget Shoppers” based on their website behavior involves analyzing various metrics and patterns that indicate price sensitivity and cost-conscious shopping habits. Here are several ways to identify this audience segment:

  1. Price Range Filters: Analyze the usage of price range filters on the website. Budget shoppers are likely to use filters to narrow down their search within a specific price range, often at the lower end of the spectrum.
  2. Deal and Discount Engagement: Monitor engagement with deals, discounts, and promotions offered on the website. Budget shoppers are more likely to click on and interact with discounted products or limited-time offers.
  3. Use of Discount Codes or Coupons: Track the usage of discount codes or coupons during checkout. Budget shoppers are more likely to take advantage of discounts to lower the total cost of their purchase.

This information can be used for analysis in Adobe Analytics. For example, to identify ‘Budget Shoppers’, you might look for interactions with price filters, repeat browsing of sale sections, or usage of voucher codes. Use your analytics data (in conjunction with other sources) to validate that the audience exists on your website, and is large enough to A/B test or serve personalisation to.

ChatGPT can then be used to generate more specific ideas for that audience. Our recommendation is always to A/B test that your personalisation works before rolling it out, so we asked ChatGPT for personalisation ideas to test.

For the audience ‘Budget Shoppers’, suggest 10 website A/B test ideas that address the audience’s key considerations, with the aim to increase their likelihood of purchasing. Include which pages of the website the ideas pertain to.

Example response:

  1. Clearance Section Visibility (Homepage): Test different placements and sizes of a clearance section link or banner on the homepage to see where it gets the most attention. Budget shoppers are often interested in clearance items, so making this section more visible could increase their likelihood of purchasing.
  2. Filter by Price Range (Product Listing Pages): Test the addition of a filter option allowing users to narrow down shoe options by price range. This helps budget shoppers quickly find products within their desired price range, enhancing their shopping experience.
  3. Budget-Friendly Recommendations (Homepage or Product Pages): Test displaying personalized recommendations based on price sensitivity, highlighting budget-friendly options that align with the shopper’s preferences and budget constraints.

Using GPT-4

At the time of writing, the free version of ChatGPT uses model version GPT-3.5. GPT-4 is used in the paid version of ChatGPT, which is OpenAI’s more advanced system.

GPT-4 can browse URLs and analyse screenshots, so you can point it to specific URLs and provide screenshots as additional sources of information. For example:

Act like a conversion rate optimisation manager for a retail company that has a website that sells shoes: https://www.toms.com (please research the site to get up-to-date context). Also, use your knowledge of the provided screenshots of the website product pages. For the audience ‘Budget Shoppers’, suggest 10 website A/B test ideas for these product pages that address the audience’s key considerations, with the aim to increase their likelihood of purchasing.

Our findings were:

  • Using GPT-4, providing screenshots along with URLs had better results than just providing URLs alone.
  • There was always overlap between the results provided by ChatGPT and GPT-4.
  • ChatGPT tended to provide broader best practice & tactical ideas, vs GPT-4’s more specific & strategic ones. However, you may be able to get round this in ChatGPT by providing more context and detailed prompts!

In summary…

ChatGPT can be used as a great starting point for generating ideas for personalisation – it’s then up to you to apply knowledge of your organisation to create a truly data-driven roadmap.

To find out more about our booster programmes for data-driven personalisation, please contact us.

This website uses cookies

We use cookies to improve your experience and to provide us with insight into how people use our website.

To find out more, read our cookie policy.