The FindWAtt Blog
Google Shopping, Product Feed Management & Product Data Optimization
How Much Are Your Product Titles Worth?
Maximize the Value of e-Commerce Product Titles Using 2 Types of Words
Each product title in your catalog has a value to you because of how many sales it generates, but that's not the kind of value I want to talk about today. (If you want to know how much revenue bad titles cost you, you can read our case study where improving titles resulted in a 151% more clicks while reducing cost per click by 28%).
What I want to focus on today is the value a product title conveys to a potential buyer, to someone who wants to make a purchase and needs to know if your product will meet their needs and be worth their money, or if they should keep looking.
Every word in your product title is either conveying your product's value to a customer or it isn't. If you don't convey enough value by the time they finish reading the title, why should they click on it? That's challenging enough by itself, but in a google shopping ad, the price is also displayed next to your product title, which means you also have to give enough information to show why your product is worth the listed price.
To make it even more difficult, you may be competing against ads featuring low-quality, bargain-bin products. You can't beat them on price, so you have to show customers why your products have a higher value to price ratio.
That's the problem. So how do we figure out which of words to put in our product title that can describe the product convey the most value to a potential customer? To start with, there are two potential sources of data on value.
- Direct from cost, indirect value to customer from your Google Shopping feed
- Indirect with cost, direct value to customer from your Search Query Report
I'm going to describe how to use the first source, but both will need to be taken into account to improve your final product titles for optimal SEO/PPC performance.
Every product in your shopping feed has a listed price and text describing it from structured (e.g. color, brand, material) and unstructured fields (e.g. title, description, bullet points). Inside that text are a group of words about what the product is, what it does for the user, what it looks like, what material it's made of, etc. Each word conveys some value to the customer, but how do we calculate how much value in order to optimize the title?
If a particular word only appears within two products' text and the products are both listed at $50, then we would say that word correlates to a $50 perceived value. If the word appears in 20 products with prices from $45 to $55, that is a very strong relationship. On the other hand, if the word appeared in products scattered from $10 to $1000, then that word has no particular relationship to a price range. So our value calculation looks at the average price, and the variation of prices of the products.
(Note: since we are calculating averages for each word, this works best with larger product feeds. Or we can include training data from across a larger dataset containing products that from other sources that overlap the product categories we want to examine.)
Once we have calculated the perceived value for all words, we can calculate how valuable a product title appears to be to a consumer. We take the words that correlate to specific prices, add their average prices and divide by the number of words. Then we compare the perceived value to the product's listed price. If the title value closely matches or exceeds the listed price, your product ad appears attractive to customers as a good deal. If the calculated price is much less than the listed price, a customer looking at your product ad is going to think you are asking too much or your product is offering too little, and probably move on to the next ad.
Let's imagine our product feed contained many TV sets ranging from discontinued CRT televisions to the latest OLED displays. If we ran the price correlation analysis, we should find that "TV" doesn't match a very specific price range, "CRT" matches a very low price range (generally $50-$150 unless it has some special feature) , and "4K" matches a high price range (generally $500-$4,000). If you have a product title for a 4K TV that doesn't doesn't "4K" in the title, it is going to look a lot less valuable than both its price and real features imply. The same thing also shows up with size. A "19-inch TV" and a "55-inch TV" are typically very different in price range. If you see both advertised at $400, the 19-inch will seem vastly over-priced, while the 55-inch will seem like a sweet deal.
That's the simple version.
To get more in depth, there are actually two types of words that we are interested in. The first type that I used in the example above is from the group of words with very specific meanings that relate to what a product is, and to its unique features. The second type of word relates to a more generic set of product attributes that increase a product's value but don't match a particular range.
For example, you can many things made of leather (jackets, hats, gloves, couches, etc), and they will generally be pricier than each of those same kinds of products made out of other materials. So even though a "leather hat" and a "leather couch" have very different prices, the word leather increases the perceived value compared to just "hat" or "couch" by themselves. If "leather" is left out of the product title, your ad will seem overpriced.
To find this type of word, we need to measure the impact each word has on the values of the products it appears in to see if it generally increases, decreases, or doesn't change the perceived value. The word "refurbished" should always correlate to a lower price than an equivalent title without it. The word "gold" should generally increase the price of titles it appears in. The word "men's" or "women's" should have no specific impact on price.
If you have a small product catalog, know all the types of features intimately, and have analyzed customer feedback thoroughly, you can probably do this product title optimization by hand. Otherwise you need a way to automatically find the relationships of words to perceived value, measure the effectiveness of your current product titles, and suggest which words to add to each title that would improve your product title's SEO and PPC performance.