About MContextual

How well does it work?

Advertisers using MContextual for targeting custom contextual audiences report large increases in click through rate and decreases in cost-per-click (CPC). One advertiser targeting a niche audience saw a 12% increase in CTR, and 28% decrease in CPC compared to legacy keyword contextual providers (Peer39).

Is MContextual AI?

Yes, MContextual's LLM uses the same technology that powers services like ChatGPT and OpenAI and all the other public large language models ("transformers", in case you were wondering).

Should I worry about my info being used to train models for my competition or leak to places I don't expect?

No. Your data won't be used to train any models, and all MContextual services are self-hosted and self-contained.

What about privacy and GDPR concerns?

Not a worry! MContextual ONLY processes text you give us and content published publicly on the internet. Being totally free of the complexity of all the privacy rules these days is a big benefit of contextual data over other methods of data-powered advertising.

What about cookies?

Not used! We would not suggest targeting with both audience data and contextual data at the same time.

How can I get started with MContextual as an ad buyer?

There's a lot of applications of MContextual data, so there's not just one way. If you are a buyer, looking to get a URL targeting list, to get started just give us some examples of content for which you want to find lookalikes. This can be a few articles (not necessarily ad supported!) or perhaps a chat session with ChatGPT. After this, you send the examples to us, and we'll get your URL list within 24 hours typically. We can customize it to be as aggressive or conservative as you like, and work with you to get as many URLs as you need.

How do I actually use the MContextual data in a DSP?

Once you have your URL list, the next step is to create a campaign in your DSP of choice that targets the URLs. In Xandr, this is possible through the Inventory List Item service. On The Trade Desk, this is possible using the Direct URL Targeting feature (note: this URL is currently broken as of this writing). On DoubleClick Display and Video 360, this can be done using the Apps & URL targeting feature. Or, you can use our recommended DSP (Pontiac, below).

Do you have any recommended DSPs?

Pontiac Intelligence is already integrated with MContextual and you can get a campaign running quickly.

How can I get started with MContextual as a publisher?

If you'd like to have MContextual identify URL lists within your existing content, for example to create an inventory package (perhaps for a private deal) we can do that. It will involve scraping your site and running some of our content analytics tools on the data. Send us an email and we can figure this out.

How can I get started with MContextual for analysis?

As a trader, you may want to use MContextual data to analyze trends and patterns in your campaign's reports to find hidden patterns. To do this, you'll need a report that shows URLs where your ad served, which can be obtained from a verification vendor (e.g. DoubleVerify, IAS, Human) or from the ad exchange directly (e.g. Xandr has this data in their backend and reporting interfaces). Take a look at our content analytics page if you want to learn more about this.

If you are planning to run a campaign with MContextual and would like to add a pixel, send us a note. We can give you a URL-detection pixel to place in your creatives to get URL data for analysis (either to assist in optimization of the campaign, or to analyze post-hoc).

How is MContextual priced?

If you require URL targeting list, then it's based on the number of URLs you need. For other data services (MFA detection, content analytics, or Ads.txt data) we price based on how often you want the data (i.e. one-off, regularly). Send us an email and we can figure out what works.

I'd rather get a list of domains instead of URLs, is that possible?

Yes, that is doable. This will result in less specific targeting, but definitely can improve scale. For URL based lookalike model, MContextual's algorithm contains a few neat tricks to give you more reach, but using a graph-traversal algorithm and prefix mining algorithm to find optimizations.

About MFA sites and how MFA works

What is MFA (made for advertising)?

Made for advertising (MFA) sites are websites made about generic inoffensive subjects to garner advertising dollars. The content is low quality and diverts dollars from higher quality publishers.

How does MFA work?

MFA sites work by arbitraging traffic generation and content generation. Essentially, buy users, earn ad revenue at a higher rate than the cost of generating the (low quality) content. There is a high risk of fraudulent users in this model.

How does MContextual find MFA sites?

MContextual is able to find MFA sites based on their content signature, rather than properties of the users themselves. Basically, low quality sites form a unique pattern in the text that we can spot. Compared to some anti-fraud vendors that rely on user data from cookies and IP addresses, this can be a lot better.