With any marketing campaign, you will notice daily fluctuations and it is extremely important to bring some levity to the situation and not overreact. Especially with CSE use, you will notice daily fluctuations in performance. The exact amount of fluctuation will vary depending on the size of your account and the channels being used. However, this does not obscure the point which is to be patient.

Time Range

When judging performance, it is best to measure things on a weekly basis, at a bare minimum. On a recent webinar I was listening to, Tom Leung, Product Manager for Google, recommended making site tests within 2 week time intervals as a minimum. Ideally, he recommended performing site tests over a month long period.

It’s my belief that working with comparison shopping engines is extremely similar. Do not judge results over a few days time as definitive proof something should be done.

Seasonal Variations

One additional thing to take into account when judging performance is seasonal variations. Since we are looking at month long time ranges here, take a look at seasonal differences. For example, if your leprechaun and Easter bunny outfit was not performing well in January, perhaps you should take another look at it.

Seasonal variations contribute greatly to the rise and fall in popularity of products. So, be aware of these variations, even when not as obvious as holiday fluctuations. I recommend being acutely aware if your usual yearly sales decline 5% in April, so you can be aware that if you notice a drop in CSE revenue or return on ad spend, there is most likely a correlation.

CSE Seasonality and Testing : Putting It All Together

So, in conclusion, make sure you are providing enough time to test variations on comparison shopping engines. At the very least, schedule 7 days for a test, but preferably schedule 14 days or more. In testing in general, a statistically significant amount of traffic must pass through the test to make it meaningful. Even more importantly, it has to be done at the right time. Keep in mind that a Christmas ornament will not sell well in July, and if you are testing home improvement items right now, your year to year numbers will be slightly skewed. Give enough time for statistically meaningful information to build and be aware of external conditions. Then, you should be able to squeze a profit out of all that time you are investing.

When categorizing items, a comparison engine will many times simply drop the item into this “misc” category if a good match is not found by the engine’s algorithms. On this surface this seems OK, because it is extremely hard to put together a simple taxonomy capable of holding millions of products (CSE marketers and engineers, I feel your pain). The problem for retailers is that when this happens to their products, it can dramatically pull down performance.

Thinking about this in a brick and mortar sense, imagine in a store, there is a big grab basket of items in the back. These items don’t appear on a shelf where consumers will find them, instead lying destitute in a confused collage of custom sewn quilts, portable pet massagers, and exotic wall art. The chances of a consumer sifting through this barrage of merchandise to find your mis-categorized item is slim. That is the reason why it is essential to pull the item out, and find a category to place the item into.

Or maybe you are just looking to see what it would take to get on board with a shopping comparison engine. That is why I created this quick reference guide. Remember, this is just a guide, not an absolute schedule. This is based on an average depending on your experience – could take more or less time than this. It also depends on where you are heading to. If you are trying to get on Amazon Seller Central, this may take a little bit longer. So here is what you are in for on each channel:

1 hour to set up an account per engine: This includes finding the link to enter in billing information, typing it all in, entering any pre-requisites. Once you have done this 10 times, I am sure the average will go down to 10 minutes.

5 hours to format a feed: Even if you are using a third party solution, you can expect to invest some time here to really understand what is being asked for. You or anyone sending a datafeed for the company should know what is needed, and what the consequences are for not having a shipping rate, for example.

3 hours to transmit the feed and confirm proper transmission: This is the most important and sometimes over looked aspect. You want to make sure the feed is properly being updated. This 3 hour estimate is spread over a couple of days time, to ensure updates are being processed on the engines’ side.

3 hours to trouble shoot initial feed: This is an average, but I truly believe that every new feed will need to be tweaked based on the engine’s requirements. This goes back to the previous 2 steps. Chances are you mis-understood or just plain overlooked something. You definitely want to spend some time in reviewing all categories you have assigned, product names, and pricing. And if you have not spent time assigning categories, then you need to go back to the previous steps.

5 hours for observation and further optimization: This is where a good merchant is separated from the average ones. Review the analytics, product by product. What is the competition doing? Which products are selling? Which products are leading to upsells? What are the differences between this engine and other engines? Are there additional things that can be done like signing up for customer feed back tools? Ask questions, and many of them.
17 hours total, for one shopping comparison engine feed: Remember this is just an estimate, but truly anyone who is looking to simply shoot a product feed out there, I would suggest either waiting until Q1 2008, or get yourself a responsible professional to put this together. You don’t want to spend $500 to sell 3 pairs of iPod headphones.

For CSE’s in general, there may be many different categories to list an item.

If you go on PriceGrabber, and search for Leather Briefcase, you would be asked to choose which category to look in:

Laptop Carrying Cases
Briefcase & Business Case
Carry Ons & Totes
Travel Accessories
Backpacks
Power Tools Accessories

Now, if I am a merchant and selling a leather carry case, that can be used for notebook computers, but I also know that many of my customers use for general purpose use, this can create quite an issue. The most efficient thing to do in this scenario is send the category information from the home website, or merchant storefront. After sending, see where PriceGrabber places this item.

Next, after some time tracking the results of this categorization, send PriceGrabber new information using PriceGrabber’s taxonomy to insert the item into another category (for information on PriceGrabber’s taxonomy, or any most any engine’s taxonomy, you should be able to find this information in your engine’s account control panel) .

By testing and tracking different categories, the most efficient fit should be apparent. Of course, since the holiday retail season is quickly approaching, the sooner the better for any AB testing like this.

Many merchants freeze when looking over all the possible choices, but the best tactic is send what you have and then test, test, and test. After doing this, your briefcases should start flying off the shelves and into the hands of your customers.

UPC stands for Universal Product Code. The product code is meant to consist of 12 digits that give a unique identifier for  every individual product. UPC is a push to standardize product information in a way that can be applied to all products, and in theory making life easier. Some engines and marketplaces are turning toward UPC as an answer to placing products in the right place, but there is a problem for vendors…many do not have UPC in their product data.

One of the main issues for standardizing data is not only resolving a set format, but then making sure sellers have the information needed. UPC presents an issue because while it seems to be a great way to integrate products, many distributors and manufacturers do not provide this information to sellers.  The seller is then stuck trying to integrate a UPC code, possibly for all variations of products including products with size, color, or other options. This creates a huge challenge as even a mid-level business can be looking at trying to add UPC values across thousands of options. Amazon is one of the leading proponents of UPC integration based on their own datafeed requirements, and other companies are following on this same path.

This creates a big dilemma for sellers: How to get a UPC value, and then how to add this to the data. The first question is hard, the next is slightly better. Depending on the data structure, the best scenario is to have a way to integrate UPC number with the option value of a product (if needed).  The best way to integrate…well, that all depends on the seller’s products, shopping cart, database, third party marketplaces…etc. Unfortunately, there is not an easy answer here. The creative burst that led to different sites and different ways of selling has led back to next challenge: How to put all this back together?

 1. No, you can’t assign an item to multiple categories

Only a few companies have this principle in place, but most do not. The reason is because a merchant could potentially flood an engine with only a few products spread out in many places. The consumer will quickly tire of seeing the same products over and over, abandoning the engine to shop elsewhere. This same principle means options are not typically allowed to be listed as multiple items, because a single item could then take up an entire page of results. For example, some engines like GoogleBase and NexTag only allow one item per product url or link page.

2. Picking a category

Next idea: there is no perfect category! Yes, an item could conceivably fit into multiple categories so it is up to the merchant many times to pick the best category (A-B testing anyone?). Now, many engines will actually recognize a product by SKU or UPC, and have a predetermined place for that item. So, this issue could be rendered moot. However, most times a merchant is forced to pick a category assignment so it’s best to pick a few items from a given group and search for them. See where those items come up. Then, decide if you want your items directly competing with similar items from other merchants, or if you would prefer your items elsewhere in seclusion. This is a competitive analysis that could help greatly with conversions if not total click count.

3. Try sending your categories first!

Many merchants get caught running in circles and drowning in the possibilities of where to put items. Try sending your categories first. That is, send the category assignments from the home website and see how the CSEs handle those. In the CSE back office, you should then be able to get a listing of products and category assignments. Based on those results, then see what alteration is necessary. First send what you have, and then make adjustments as needed. Don’t get caught up in unnecessary work!

The article summarized the over improvement of analytic tools but the basic vacuum created by knowledge without the understanding of how to use the information being generated. For example, if a merchant now knows what keywords or landing pages are generating traffic, what is being done about it? If a merchant is using CSE’s, are the keywords that are successful on their website being used in product feeds? Are there tests to find out if the keywords are successful across different platforms and multiple channels?

The article raises excellent points on the need for training and understanding now that all these new tools are available, especially with programs like Google Analytics that are free and virtually accessible by everyone.

This article and questions I see from merchants regarding best practices in optimizing product feeds outlines a real need in the industry for more information and outlines in how to take advantage of on-line retailing now that more things are becoming standardized to the point of being able to measure with great accuracy what works and what does not. Look for an upcoming piece outlining basic steps here that will detail best practices for optimizing comparison engine feeds in 2 parts : 1) Optimizing the product data and 2) Optimizing product selection and bid strategy to generate better return on investment (aka return on ad spend).

If you have not already, click on the link to Ralph Wilson’s article here: An Analytics Consulting Model for Small Businesses