Before you try to refine your data feed, you first need to do the following:
-Set up a log analyzer program with conversion tracking
-Append shopping engine specific tracking URLS to the end of your product links (do not submit the same exact link to multiple engines)
At the very least, with this data in hand, you can track conversion per engine and sales/revenue per engine and hook this information up with cost data provided by the shopping engines to figure out simple things like engine-wide cost per acquisition (CPA) and ROI. From there, you can completely turn off an engine if it’s not a profitable marketing channel. Alternatively, if you’re doing great on all the shopping engines, you can increase all categories/keywords bids.

While completely acceptable, the problem with this strategy is that it looks at each engine as a basket of products when there are most likely some products which are hurting your overall ROI and other products which are helping your overall ROI.

In other words, a smarter (but more time consuming) strategy would be to figure out profitablity by product listing/SKU. If you want to succeed on the shopping engines, you must do this. If you are thinking ‘but Yahoo! Product Submit, Shopping.com, and others don’t provide this data,’ you’re right. Here’s a simple workaround: take SKU based reporting information from Shopzilla, NexTag, and Become, and generalize the profitability findings across the other shopping engines. Truthfully, I hate this strategy because traffic from Shopping.com is different than traffic from PriceGrabber is different than traffic from Smarter.com, but it’s the best way to handle the situation for most small and medium sized retailers. At the same time, everyone should be demanding (send an email a day) SKU level reporting from the engines which don’t offer it.
As for which products to cut and which to keep, I’d start by grouping your SKUs into three groups:

-Red: Products that get lots of clicks but no sales OR sales at a very negative ROI
-Green: Products that get clicks and sales at a positive ROI
-Yellow: Everything else

Before doing the obvious - cutting the ‘Red’ products from all paid shopping engines - make sure to check the categorization for your products. I’ve seen handbags, sweaters, and camcorders show up in the iPod category with proper, non-iPod pictures, but still get a ton of irrelevant clicks. While it’s a long shot that is the fix for your ‘Red’ products, it’s worth at least a quick check on the outliers. Ok, now cut your ‘Red’ products.

Now, as opposed to cutting the ‘Yellow’ products, take the time to look at the feed attributes - product name, product description, category, etc - and optimize the listing. Maybe your title is misleading. Maybe your description is pathetic. Maybe your image is so small that you’re getting curiosity clicks. Maybe you didn’t include shipping cost information. Maybe you didn’t include tax information. In other words, spend time optimizing your feed. These are the types of activities that you do with your PPC campaigns, why not dedicate the same effort to the shopping engines.

After cutting the ‘Reds’ and optimizing the ‘Yellows’, resubmit your feed to the shopping engines. Repeat this process once per week.

Disclaimer: All optimization strategies are suggestions and do not guarantee success. These are data feed optimization tactics I have used or others have suggested which I think everyone should at least think about, if not test. Use at your own risk. Or don’t use, just keep coming back.

Related Posts (which got me in trouble with some shopping engines)
-A Case For Product Level Bidding (June 5, 2006)
-Guest Commentary - JP Werlin on Product Level Bidding (June 6, 2005)