Here’s the Metadeal:

Twitter Facebook Linked In Paper Plan

FMTC Announces Updated Product Offering

Written by Brook Schaaf

Allow me, with great delight, to announce FMTC’s improved deal data structure alpha.

FMTC’s core product offering is our Deal Feed. This feed encompasses every available deal, sale (category, sitewide, or other), BOGO, free shipping, product, special customer (new, student, etc.) or discount. Sourced from over 19,000 merchants, each deal is hand-checked, tagged, categorized, SEO-optimized, normalized, and distributed by our API all day, every day to our enterprise subscribers, who comprise, as best as we can tell, a majority of revenue in the coupon and reward space. Without FMTC, our subscribers would face the daunting task of managing their own data (which can be difficult and costly) or post unprocessed data (resulting in a poor end user experience with reduced click-through and conversion rates).

This offering has worked well since FMTC’s inception in 2007. Seventeen years later, we are honored to have some of the original feed subscribers and to have added hundreds more along the way. The value proposition is pretty straightforward: Customers love and will never stop loving good deals, and correspondingly, they hate bad deals (invalid, expired, etc.), so they often seek out the best ones on affiliate sites.

FMTC has begun parsing natural language (that is human-comprehendible but not structured) which now allows sites the improved opportunity to filter, rewrite, compare, and negotiate. Using AI, we added fields for currency, a regular discount value, a Boolean true/false for percentage vs. flat amount, a maximum value, and a price or quantity trigger. We will add more fields for stipulations and other information after we receive feedback from our subscribers.

In parallel, we will also be using this data to cross-check existing fields such as free shipping, the old price, the new price, and the coupon code. This prompts the question: Can we simply have AI process everything? To this, the answer is no, at least not yet. As Chamath Palihapitiya, co-host of the All-In Podcast, recently noted, “You’re referring to this idea that you’re stringing together elements of a model or multiple models that each have some non-trivial error rate. And by multiplying all these error rates together, you get your final product.” Fellow co-host David Friedberg praised AI as a work tool but soberly noted that about 25% of its answers are wrong. For example, when I searched for a transcript of the podcast, AI sidebar tool Merlin incorrectly returned, “All in Podcast is a popular podcast hosted by comedians Joe Rogan and Alex Jones.” We see incorrect results even with the micro dataset of a deal label. This can be fixed with human processing — which, as it happens, FMTC is quite good at.

To the best of my awareness, there is no commercial offering with similar data, nor has any other company developed this internally. Therefore, product-market fit is unknown. I, at least, am wildly optimistic that we can meaningfully boost the economic value for everyone involved: merchants, affiliates, networks, and end customers.

This enhanced data is available immediately to our enterprise subscribers. We look forward to collaborating with you to take the customer experience to the next level. 

Leave a Reply

Your email address will not be published. Required fields are marked *

Related Posts