Optimize

Channel and Data Feed Optimization

To make this clear, first I want to break this article apart into 2 components. The first component revolves around product data. Product data refers to the actual information being sent to the shopping comparison engine or marketplace. This data needs to conform to the destination site’s rules, and this means the shopping comparison engine data may look very different than the data taken from the origin website or merchant platform.

The second portion of optimization revolves around product selection and return. I am not aware of any merchant who sees equal profitability on all products. Many products will loose money no matter what is done to the data, and essentially should be removed. Product selection should ultimately be determined by ROI (return on investment) or ROAS (return on ad spend).

1. Product Data

Product data is defined for this article as any type of data relating to the listing and display of an item for sale. Product Data is necessarily different for each item and includes everything including product name, product sku, product description, etc. The biggest misconception about product data is that if a successful format is found on a merchant’s website, this formula or structure should be the same across all selling channels.

Myth! Busted!

Actually, product data should be changed and altered based on the destination site. Every good public speaking or writing class will emphasize either speaking based on the audience being spoken to. CSE’s and marketplaces are the same. Each is a different audience and the information needs to be altered for that audience.

I am going to skip past basic manipulations such as indicating information such as ‘In Stock’ or what type of shipping is available. This should be an automatic first step (see CSE Basics), but finer adjustments are best illustrated by the description field and product name.

Attributes

A big deal is being made about product attributes as the future of data structure. Attributes are an important piece, and engines like Shopzilla and Google Product Search are turning to attributes as a way to classify products accurately across millions of skus. The best way to show how these fields come together is using Shopzilla. Shopzilla/Bizrate uses the product name and product description fields to place products using product attributes.

Shopzilla has a pre-defined list of product attributes that look like categories, but are in fact, attribute groups. There is no category id or indication of category that can be made to place products in these groupings. Each product may be sent with a Shopzilla Category ID, but this will not place a product in the attribute level of categorization. The attribute terminology must be present in the name and description fields regardless of assigned category.

In the example shown below, if a merchant sells golf clubs, and has a set irons for sale, the merchant should absolutely make sure the term ‘iron set’ is included in the product name, product description, or both. Using this term allows Shopzilla to correctly identify this set of golf clubs as an iron set, and will then place the clubs in the proper group. If a merchant uses a term like ‘set of irons’, this will not place as highly and may not even place a product into the ‘Iron Sets’ group.

Using the correct Shopzilla category id when sending these items, cat id#12070300, is advised, but this will only take a merchant so far on this engine. By changing the product names and descriptions to match what Shopzilla is looking for, a merchant will see much more targeted traffic and then more sales.

Creating New Data

Another possible manipulation would be adding a short description specifically geared for Shopzilla. Since in this example, Shopzilla is only displaying the first 50 characters of the description, most of the website description is never even seen by the consumer! As we have already pointed out, the engine in this case is using the description field mostly for placement, so what if for the first product, a description was instead set like this, where bold characters are within the first 50 characters to be displayed, and the rest is for placement:

Free Shipping on Orders over $50. Thousands of clubs available. Callaway Golf 06 Big Bertha Iron Set. S2H2, Tru-Bore and VFT — are part of the DNA of Big Bertha Irons. Add extreme Notch Weighting, a deeper 360-Degree Undercut Channel and a constant-width sole.

These are just a couple examples of data optimization. Many more are possible and advisable. For example, the actual ranking of these items on Shopzilla are determined by many factors including relevance to the group or search term, participation in the customer ratings program, price, and bid amount. Of course, data is just one factor in optimization, the next are is in regards to product selection and return on ad spend.

*In the below illustration, a search was made on the Shopzilla.com home page for ‘golf clubs’. Then when the selection came up, the prospective customer clicked on ‘Iron Sets’. The arrows indicate the ‘Iron Sets’ attribute grouping and then where the term ‘iron set’ is used the product names and descriptions to have products come up in these sections. Actual merchant names have been hidden.

2. Finanical Perspective

Optimizing Shopping Comparison Engines from a financial perspective brings in many more variables. For our purposes here, we will assume commission based selling is not being accounted for. Commission based selling on Amazon, PriceGrabber Stores, or using Shopping.com’s Wallet System is a different animal.

To make this readable, lets take 2 perspectives.

A) The simple perspective: ” I need to make more money. How do I do that?”

Using this outlook, it is best to consider a straight forward profit and loss equation, assuming the best product data in already being used (see Optimization Pt. 1)

B) The more complex, and in some cases, best perspective: “Understanding that bidding may play a significant role in achieving the best results, how do I maximize profits?”

This outlook understands the all important variable of bidding. When items are are listed at a set cost, a proper formula is easier to construct. However, when factoring in bidding as any user of SEM or Google Adwords – Overture/Yahoo based advertising knows, can significantly impact profitability.

So lets take a look at both scenarios:

A. The simple strategy

The most important thing to understand here is the need for product level tracking and reporting. Seeing whether an entire feed or even category is profitable does not provide accurate enough information to truly get good results. The cummulative effect of all products is important, but without being able to take apart the engine that drives profits, there is no good way to fine tune the engine for performance.

Using a simple formula like this should suffice:

X= listing price of the item (cpc)
Y= profit from a sale (selling price – cost of the item to the merchant)
N= number of sales
C= number of clicks

N*Y – (C*X) =

So, for a sample product, lets imagine a futon bed:

X= $.35 per click for listing the item online
Y= $90.00 ($199.00 selling price – 109.00 cost of goods)
N= 3 sales through using an engine
C= 125 clicks generated through the engine over a given time range

$90.00*3 sales= $270.00 in profit
275 clicks * $.35 per click= $96.75

Net profit or return on investment (ROI) in advertising the futon= 270.00-96.75= $173.25

Using this scenario, this is where profit and loss is essential. The conversion rate in this scenario

is a low 1.09%, however the profit is significant.

Using this equation, you want to sum all products to see how the engine is doing as a whole, but of course any product which has a negative or unsatisfactory profit level, would be removed from the feed.

Product	Clicks	CPC	Sales	Profit/Prod	Total Cost	ROI	Feed?
Futon	275	.35	3	$90.00	$96.75	$173.25	Yes
Pillow	55	.25	1	$3.00	-$13.75	-$10.75	No
Pen	11	.25	2	$1.00	-$2.75	-$1.75	No
Table	21	.45	1	$55.00	$9.45	$45.55	Yes

B. Look Deeper

Looking a little closer, many engines allow product level bidding. This increases the variables one step further allowing for increasing and decreasing bidding based on a products performance. This requires an analysis of bidding and results over time and with different variations.

Lets assume over a given time range, we have changed the bid on a product and measured the ROI (all other variables kept constant). This change in bid structure has resulted in high visibility in some cases, and low visibility in other cases. So the results are as follows:

Bid	Clicks	Sales	ROI
.25	91	1	$67.25
.35	275	3	$96.75
.45	388	5	$275.40
.55	752	6	$126.40

In this scenario, more clicks through more traffic but not necessarily more sales. For this product, $.45 would be the optimum scenario. However,continual testing should be done because in a real world environment, other merchants would also adjust their bids resulting in changes in exposure.

ROI is one method for determining profitability. Many retailers use different qualifiers for success. We have already covered ROI, but there is also:

Cost Per Acquisition (CPA): Total Advertising Cost/Number of Sales
Return on Ad Spend (ROAS): Revenue Made in Sales/Amount Spent on Advertising
Conversion Rate: Total sales/Divided by Total Number of Clicks

The following depicts how these products from the first example would show using these indicators:

Product	Conversion	ROI	CPA	ROAS
Futon	1.09%	$173.25	$32.25	279%
Pillow	1.81%	-$10.75	$13.75	22%
Pen	18.18%	-$1.75	$1.38	36%
Table	4.76%	$45.55	$9.45	482%

Across hundreds or thousands of products, an ideal solution would be to continually adjust bid and data values. If a product continually under-performs across a determined time period, then that product should be discontinued or bidding should be set to zero. Using this type of process, return can be gradually increased even if overall traffic and clicks may actually decrease. An increase in profit does not always mean an increase in sales or traffic. Sometimes efficiency is quiet but will keep you in business and lead to success in long term where it matters most.