There are many ways that CSS and merchants could agree to collaborate. A given CSS may offer several service models, or tailor services to meet your needs. CSSs can offer a variety of services and expertise. Others specialize in specific categories, such as apparel. Others manage campaigns and feeds for you. While others offer tools to help you manage your product data, bidding strategies and other details yourself.

There are also hybrid models, and CSSs might offer additional services. While some CSSs charge per click, others may pay a commission if an ad clicks on a sale. The combined strengths of multiple CSSs can improve the performance of your Shopping ads. To place Shopping ads, you must use at least one CSS. However, you can combine the strengths of several CSSs if you wish.

Merchants may wish to compare campaign performance among different CSSs. We understand that merchants might want to test their shopping campaigns due to the many differences between service models. This document outlines the factors that you should consider in order to set up a fair test for both CSSs. What should a CSS do if he or she wants to test two CSSs for a merchant?

Before considering a test between CSSs we strongly recommend you to review the objective of the test, together with the Google CSS Partner. The Shopping ads auction is CSS agnostic. As a result the exact same setup between 2 CSSs (new and identical Merchant Center feeds, new and identical Google ads campaigns,
same budgets, same bids placed in the auction) will provide the same performance results.

We recommend starting with an identical setup (including bid level in the auction) to check performance and ensure setups are comparable at the start.

How can you best identify the level of CSS Service provided during these tests?

Sample goals could be:
a. Measuring who can deliver the best results for the same ROAS;
b. Measuring which CSS can get the most volume;

In evaluating the different services provided by CSSs during these tests, for example optimisation techniques, make sure that the campaigns have the identical setup when the test starts (such as geography selection, set of items to be advertised). From there on, CSSs can start with their own optimisation techniques. This is important to make sure the merchant can ultimately make their decision based on how one CSS has optimised compared to the other.

We recommend testing at least 6 weeks up to 3 months because it gives time to evaluate the difference between optimisation strategies.

Having done all the above, we recommend both CSSs to agree on metrics they’ll use to evaluate the test and to create a shared hypothesis with the merchant to ensure all parties have the same agreed expectations for the test results review.

The reason you should do this is to make it clearer for all parties if all parties agree on what will be tested and (dis)proved. This reduces confusion / different interpretations of the results after the test by different parties. In case of setting a certain ROAS for the test, it is best to consider calculating the ROAS net of any service fees the CSSs may be applying.

In order to have the most trustworthy results from the test it is very important to start with identical conditions. Here are some recommendations you can apply:

1. The merchant creates a new Merchant Center with CSS 1 and CSS 2 (please check “opted in the property and beyond search” option to ensure identical setup)
2. Upload the exact same product data in the two MCs. We recommend a small portion of the inventory to limit the impact of the test on the merchant’s business. All product data must be the same.
3. The merchant should create two brand new campaigns, one per Merchant Center, and build the exact same campaign structure.
4. Set campaigns corresponding to both Merchant Center accounts with comparable settings including geo targeting, language and budgets.
5. Once you can confirm that all aspects of both set-ups are comparable (in order to ensure that the testing scenario is fair), CSSs can start optimizing.

Note: at this stage, please confirm the KPIs (goal and metrics that will be tracked) once again, as different goals may need different strategies.