Machine Learning at MatchCraft
MatchCraft employs a combination of off-the-shelf and proprietary software to deliver valuable solutions for scalability and efficiency
Sitelink Suggestions
Sitelinks are a valuable asset utilized in both Google and Bing paid search ads that take users directly to specific pages on an advertiser’s site. Using sitelinks increases the size of the ad itself giving the ad increased visibility. Sitelinks also provide helpful insight into how users are interacting with your ad. For advertisers, they are a no-brainer when it comes to PPC ads, but for campaign managers it can be challenging and time consuming to create relevant sitelinks for hundreds (or even thousands) of campaigns.
MatchCraft designed its sitelink suggestion tool with efficiency of scale in mind. Utilizing the Bing Web Search API, our engineering team created a solution that generates a list of URL names and links that are relevant to the destination URL specified for the category. After some minor clean up, the list is returned to the user who can then select from the suggestions. The end result is a significant time savings since the campaign manager no longer needs to navigate through the advertiser’s website to find pages that meet the needs of the campaign.
MatchCraft designed its sitelink suggestion tool with efficiency of scale in mind. Utilizing the Bing Web Search API, our engineering team created a solution that generates a list of URL names and links that are relevant to the destination URL specified for the category. After some minor clean up, the list is returned to the user who can then select from the suggestions. The end result is a significant time savings since the campaign manager no longer needs to navigate through the advertiser’s website to find pages that meet the needs of the campaign.
Category Suggestions
MatchCraft’s keyword and ad copy library, AdVisor, offers campaign managers more than 11,000 predefined categories to select from (and that’s just our US English taxonomy). When creating large volumes of campaigns, sometimes for businesses in niche industries, it can be very time consuming to select the categories most relevant to the advertiser’s business. Campaign managers may miss categories that could be beneficial or, not uncommon, utilize the first category in the library and enter entirely custom keywords and ad copy. Neither scenario results in the most efficient campaign setup and delivers a less than ideal user experience.
In order to improve the quality of category assignment, MatchCraft created the Category Suggestion Tool utilizing a combination of Google Ads Keyword Planner and a version of multilingual Universal Sentence Encoder from Sentence Transformers (SBERT) python library. After entering the final URL for the segment, the user is provided with a list of relevant categories that can then be applied to the campaign.
Like so many things related to machine learning, the campaign manager retains their role of human oversight and can edit and customize the categories, keywords and ad copy as necessary. The difference is that now the campaign manager can do so more efficiently and armed with relevant information.
Ideas from URL
Ad copy creation can be one of the most time consuming functions of setting up a new paid search campaign. MatchCraft’s AdVisor taxonomy significantly reduces the amount of time it takes to create ad copy by providing a multi-layered library of a copy for campaign managers to select from. But sometimes custom ad copy is necessary to refine the campaign and to meet the specific needs of the advertiser. In an effort to make the campaign manager’s job easier, MatchCraft developed its Ideas from URL tool.
Utilizing a combination of text scraping and text classification, the Ideas from URL tool can generate a set of potential ad copy derived from the content on the final URL of the segment. With the BERT model as its foundation, the tool combines off the shelf technology with proprietary software to deliver a solution that delivers relevant ad copy consistent with the language found on the advertiser’s website. The end result is both significant time savings and highly customized, relevant content.
Initial Bid Estimator
MatchCraft’s proprietary bidding and budget management technology has a proven and reliable history of consistently pacing budget across the duration of a budget cycle. The starting bid value for a campaign plays a significant role in getting a campaign up to speed efficiently. For a campaign targeting a competitive vertical, lawyers for example, CPCs can be very high when compared to a less competitive vertical or small market. The bidding process needs to begin at a value that is appropriate for the vertical and market being targeted. Taking into consideration the categories, geo-targets, total budget, and historic data, MatchCraft developed a model used to predict the initial bid value needed for the campaign. To train the model, the team utilized campaign level data layered with SBERT distiluse model and XGBoost regression modeling. The end result is a starting bid that has reduced the ramp time needed for campaign spending and has also been effective at reducing overspend at the beginning of a campaign.
To learn more about how MatchCraft’s AI enabled solutions can help transform your business, request a demo today.
