The FindWAtt Blog

Google Shopping, Product Feed Management & Product Data Optimization


Share it!

Pros and Cons of Automatic Product Title Fixing

The benefits, challenges, and implications of automatically fixing title problems

Tim Gilbert 2017-10-03

Why do my titles need to be fixed?

Having title problems costs you money. Bad titles get less impressions, clicks, worse click-through rates, and cost more per click. In a case study, we found that optimizing product titles improved CTR by 47%, increased clicks by 151%, and reduced CPC by 28%. And that's not even including potential extra benefits of optimizing titles to match search terms and buyer-intent.

Enhancing product titles resulted in 151% more clicks, a 47% increase in CTR & a 28% reduction in CPC.

 

Why should I consider automatic fixing of my e-commerce product titles?

Cost

 • Optimization — computers can calculate optimizations that aren't otherwise possible

 

This applies to creating multiple title versions that are optimized for different delivery channels with different requirements such as lengths or forbidden text, or tweaking titles for maximum SEO/PPC performance using statistics extracted from a search query report. Missed optimization opportunities and missed title errors both cost you money.

 • Labor costs — human time costs more than computer time

 

Your attention should be saved for the problems that require more judgement and product awareness, not the application of standard rules.

 
Quality

 • Consistency — having a cohesive product feed 

 

One person fixing one batch of titles won't produce the same results as another person. They will use different grammar and capitalization, select different ways to express sizes, add different attributes, etc.

When all titles are the same quality, you can more easily compare relative sales volumes, optimize on particular title patterns, and maintain successful techniques.

 • Thoroughness — making sure every correction is made

 

People are prone to miss title errors or forget steps in a process. Computers will make sure every step is done and won't overlook problems.

You want to know that mistakes weren't made because someone was busy, or in a rush, or distracted with other issues.

 
Time

 • Product Turn-over — keep up with new products being added every day

 

This is especially important for merchants with a constantly changing stock, marketplaces working with a changing set of merchants, or any source dealing with used/resold products. If you're are dealing with new product data constantly,

 • Outside data — Titles coming from another source

 

If you don't have direct control over the source data (if titles are coming from manufacturer, retailer, or other department), you either have to accept whatever errors they put in the titles, or you need a correction process that can apply the same fixes each time the title is re-sent with the same problem in a way that won't waste your time.

 • Speed — allow rapid release of new or updated products

 

Computers can fix thousands of titles in a few minutes, while humans can take many hours to accomplish the same work.

 
Reporting

 • Auditing — what can be measured can be improved

 

Computers can keep a record of every change made to each title. This allows any issues in how processing is done to be traced and fixed to improve future results. If you are working with titles from outside sources, it also allows to monitor whether they are improving or worsening.

 

Product Title Performance Grader

 

What are the challenges in creating a system to automatically fix titles?

 • Evaluating title problems in context

 
  • Identifying brand names and trademarks before correcting spelling. Find trademarks phrases are particularly challenging since they are often not marked with a ™ character in titles because some marketplaces don't want special Unicode characters in the title, and because it is more challenging to figure out how many words are in the trademark phrase.
  • Adding the attributes that are most appropriate for the particular product category. This is particularly important to optimize your SEO/PPC performance.

 • Making sure title still looks like it was written by a human

 
  • Not splitting phrases by inserting a word in the middle or removing part of the phrase.
  • Keeping products attributes in a reasonable order when adding new ones or re-ordering them. This can be using standard english order of adjectives, or following standard practice for order for that category of products.
  • Not breaking grammar. This is less an issue for titles which don't follow exact English grammar rules, but there are still problems that show up with quickly cobbled together fixer.

 • Calculating confidence levels for each problem

 

 • Fixing problems in the correct order

 
  • Deal with problems revealed by fixing other issues (when you remove gimmicky punctuation such as "*F*R*E*E*S*H*I*P*P*I*N*G*", you can then detect the remove the title text forbidden by distribution channels like Google Shopping such as "Free Shipping").
  • If removing a word from the title creates punctuation problems such as repeated commas, or leading/trailing dashes and spaces, make sure they are resolved afterwards.

 • Having extensive knowledge bases of exceptions

 
  • Creating a system that gets more accurate over time and spending the time to maintain and train it takes a lot of effort. Especially when a particular pattern might be allowed in one product category, but wrong in another.

 

What automatic fixing can't do (yet)

 • Researching outside the data feed to fill in missing information

 

Sometimes the product feed just doesn't have enough information to infer missing data, so you'll need to examine images, search for similar products on the web, or even contact the manufacturer.

 • Resolving disagreements between structured and unstructured data

 

If the title has one brand and the brand field has another, which is correct? Identifying the conflict between the values doesn't reveal which one to use.

 

My recommendation:

To do title fixing correctly and efficiently takes some in depth knowledge of product data in combination with some expertise in programming. But if you're dealing with large catalogs, high product turnover, or external data, it's an important part of your e-commerce product data process to either create your tool, or find someone who can do it for you. FindWAtt is now offering its free product title performance grader to help you find 20 kinds of title problems in your product feed. We are currently doing internal testing of the first version of the automatic title fixing tool I created to start addressing the most common product title problems we see, and we'll be rolling out fixes for more complicated problems as I develop them.