Configurations Best Practices
Configurations
Merely adding Discover to a site does not automatically create value; we must take a strategic approach to ensure that the Global and Personalized Weights configurations align with how customers actually shop within a category or vertical on the specific site.
The following cheat-sheet provides initial weight settings for each of Discover’s default parameters. While some are based on rigorous testing and optimization precedent, others are informed assumptions driven by known vertical-specific traits and shopping behaviors. Some example discovery questions that provide insight into these determinations:
- Are the products replenishable? Do customers purchase the same or similar products on a recurring basis as is often the case in Grocery and Beauty? If so, then past consumption is a strong indicator of future consumption and parameters such as Product Purchased and Category Purchased should strongly inform the Personalized Weights. If not, as is the case with Electronics and Appliances, these parameters should be buried.
- Do brands matter? Consider whether customers of a particular vertical exhibit brand loyalty or are fairly agnostic. Within Beauty, for example, the cost of switching can be high; when customers finally find a brand that they like, they usually stick to it. Within Media or Home Furnishings, that is usually not the case which means dialing back on the Brand Viewed and Brand Purchased parameters in the Personalized Weights.
- Do customers stick to the same pricing tier? Does the vertical attract classes such as high-end, midrange, or budget buyers? We frequently see this in verticals such as Electronics, Appliances, and Home Furnishings which offer a wide range of affordability options to buyers with various levels of spend. While customers may browse across tiers, they tend to purchase within the one closely aligned with their budget. As such, we downplay Price Quartile Viewed, putting more emphasis on Price Quartile Purchased. In Office Supply, both parameters are of little importance and are therefore dialed down.
Credibly hypothesizing Discover configurations requires an understanding of the particular vertical-retailer intersection and its customer tendencies. From there, you must make an informed, albeit subjective, determination as to what factors matter more than others and set the weights accordingly. Again, we are not aiming for perfection; we’re merely establishing a starting point—a springboard from which future testing and optimization can occur.
Global Weights
|
Param |
Apparel |
Luxury Apparel |
Electronics |
Appliances |
Grocery |
Health/ Drug |
Beauty |
Sports |
Media |
Home |
Office |
|---|---|---|---|---|---|---|---|---|---|---|---|
|
Click Rank |
3 |
3 |
3 |
3 |
2 |
0 |
2 |
2 |
2 |
3 |
3 |
|
CTR Rank |
4 |
4 |
5 |
5 |
5 |
3 |
4 |
4 |
4 |
4 |
5 |
|
Newness Rank |
6 |
6 |
7 |
7 |
3 |
1 |
6 |
6 |
8 |
6 |
4 |
|
Revenue Sales Rank |
8 |
8 |
6 |
8 |
8 |
5 |
8 |
7 |
7 |
8 |
8 |
|
Unit Sales Rank |
9 |
9 |
8 |
8 |
8 |
5 |
8 |
7 |
7 |
8 |
8 |
|
View Rank |
3 |
2 |
3 |
3 |
5 |
0 |
3 |
3 |
3 |
4 |
3 |
|
Brand Viewed |
2 |
2 |
2 |
2 |
2 |
3 |
4 |
2 |
1 |
2 |
2 |
|
Brand Purchased |
3 |
5 |
5 |
5 |
4 |
5 |
7 |
6 |
2 |
2 |
4 |
|
Categories Viewed |
5 |
3 |
6 |
6 |
5 |
4 |
5 |
2 |
5 |
5 |
5 |
|
Categories Purchased |
2 |
2 |
0 |
0 |
7 |
6 |
7 |
6 |
7 |
1 |
7 |
|
Price Quartile Viewed |
3 |
3 |
5 |
5 |
4 |
3 |
5 |
3 |
1 |
7 |
3 |
|
Price Quartile Purchased |
3 |
3 |
5 |
5 |
4 |
3 |
5 |
3 |
1 |
7 |
3 |
|
Product Newness Affinity Viewed |
1 |
2 |
2 |
2 |
0.5 |
1 |
2 |
2 |
5 |
2 |
0.5 |
|
Product Newness Affinity Purchased |
2 |
3 |
4 |
4 |
0.5 |
1 |
3 |
4 |
6 |
2 |
0.5 |
|
Product Viewed |
4 |
6 |
7 |
7 |
4 |
2 |
4.5 |
5 |
4 |
6 |
4 |
|
Product Purchased |
0 |
0 |
0 |
0 |
10 |
8 |
9 |
1 |
0 |
0 |
0 |