You don’t eat artificial anymore, why should your marketing budget?

Artificial Intelligence is like artificial food - yack! Biological Intelligence is like organic food - yum!

In marketing technology, Machine Learning and Artificial Intelligence are to the layman seemingly interchangeable, overused concepts and jargon designed to say a few things:

  1. Our technology is proprietary
  2. Our process is complicated
  3. Our results are superior

And most of all, Artificial Intelligence says a marketing technology company might just have a high-tech solution to marketing’s all-time challenge – how to reach the right person at the right time, with the right message.

Artificial Intelligence has been intriguing us, scaring us, or both on the big screen for decades but it hasn’t been until the last few years that it’s made its way into our businesses as marketers.  For most of us, AI has become synonymous with Ad Tech, accepted as ‘table-stakes’ for being more than just the next platform – especially on the buy-side of advertising.

However, in many recent Ad Tech missteps, Artificial Intelligence has been front and center, delivering anything but the results marketers had hoped for.  While the promise seems great, the reality is not.  How is this so?  It seems simple enough.  Computers love crunching numbers even without artificial intelligence.  Give them AI, lots of ‘Big Data’, many iterations and success should be a given.  That’s the sales pitch at least.

The challenge with Artificial Intelligence in marketing applications is that real business is anything but artificial.  The engine that drives commerce is not a machine, it’s the human heart, forming behavior patterns only nature can explain – not 1’s and 0’s.  In fact it’s those very binary digits that holds AI back from truly connecting with human behavior the way it should.  Those values, along with ‘Truth Values’ (True/False) are major limitations in Artificial Intelligence.  No matter how hard it tries, AI is only as strong as 1 and 0 – with no room in between.  With no intermediary values, traditional AI is archaic when it comes to replicating, or in the pursuit of Ad Tech, ‘predicting’ human behavior.

Enter Biological Intelligence.  ‘BI’ is flexible.  It can adapt to change in the universe in a way ‘Artificial’ never could.  BI sees the world not in 1’s and 0’s but in a fundamentally flexible and adaptive way found in nature – the right way, if you prefer natural over artificial.  With Biological Intelligence, for example, algorithms can solve for the behaviors of people in time and space, relative to who they are, where they are and why they are there.  BI is superior at deciphering such behavior, because like the human brain, it has a continuous range of feelings about things, not just binary values.  Marketers must connect with consumers from all walks of life with varying sentiment about their products; there is no room for a one size fits all approach like Artificial Intelligence.

So, given its flaws, and dubious results, why has AI proliferated throughout Marketing Technology when what we need to better understand humans is BI? Because AI is simpler.  It’s more widely understood in computer science.  It’s easier to program.  And in an ecosystem requiring compatibility, portability and open standards, it’s simply the cheaper way to market.   The scientists programming Artificial Intelligence models best understand computers.  And that’s useful in matters of bits and bytes.  But in matters of consumer preferences, like marketing, marketers who want results, must seek product design built based on more advanced science – human behavior.  It is indeed rare to find a computer scientist who is also a people scientist, but that’s what it’s going to take to understand people as data.

For a really low-tech and ‘common sense’ understanding of the issue at hand, consider turning to another domain – something we all consider on a daily basis – our diet.  While we all enjoy occasional sweets, and most of us occasionally indulge in the fattiest (and tastiest) of hamburgers and the like, we are quickly losing our tolerance for artificial ingredients in our food.  Brands such as Whole Foods have made an industry out of this preference, as we evolve to loathe the cheap, synthetic and toxic ingredients of yesteryear’s food products.  Natural food is grown, prepared and consumed the way nature intended.  It tastes better and it’s better for you – a simple win-win.  In your pursuit of Marketing Technology, it’s time to take the same leap.  It’s time to leave the synthetic, Artificial Intelligence behind and move on to a more natural solution.  Biological Intelligence enables just that, with a natural approach based on real behavior too complex to be calculated by artificial means.  Because in an Ad Tech community that has as many PHDs as marketers, real, human approaches are just what the doctor ordered.

Pleased to announce a new Locomizer Advisor!

We are pleased to announce that Peter Baldwin has joined Locomizer as our advisor. Peter brings with him over 30 years of experience in the mobile and adtech industries and extensive startup mentorship expertise. Peter’s hands-on experience and deep knowledge of adtech market will greatly help with shaping up the Locomizer’s product roadmap and overall business strategy. As an advisor Peter also brings a vast network of business contacts and prospect leads. We are looking forward to working together. Welcome to the team, Peter! It’s great to have you with us.

How Twitter’s geo data fuelled a fast food ad targeting campaign

Summary: Locomizer used geo-enabled tweets to pinpoint user segments with a high affinity for eating/drinking and fast food in order to extrapolate them on a whole population of central Madrid for a fast food mobile ad targeting campaign. As the result, Twitter data proved to drive the click through and conversion rates by 40% and 30% correspondingly. 

At Locomizer, we experiment with all kinds of data that contain a geo element such as a set of lat/lon records. When we had a chance to run a live targeting trial with our DSP partners for a fast food brand, we decided to give Twitter data a try. Our concerns about using Twitter were mainly down to two points: 1) geo-enabled tweets are still a tiny fraction of 500 million tweets generated daily worldwide; and 2) Twitter user base is not fully representative – it skews more male and younger age brackets. However, the second point had turned out to play more in favour of the campaign objectives as the results showed.

Campaign Objectives

Réseaux sociaux

Drive mobile coupon ad click-through and download rates by pinpointing audiences on the map with eating/drinking and fast food interests or intents that make them receptive to fast food ads.

Step 1: Aggregate geo-enabled tweets from central Madrid area

We gathered almost two million tweets from over 70 thousand users who sent at least 10 geo-enabled tweets during a two-month period of time. 
geo-twitter-ad-targeting-01

Step 2: Identify user behaviour patterns based on the location of historic tweets

We pinpointed tweets on the map for each and every of 70 thousand users we had in our data set.

geo-twitter-ad-targeting-02

 

Step 3: Translate that data into user geo-behavioural interest profiles

Using our proprietary database of points of interest as a complimentary input, Locomizer’s algorithm translated each user’s location history into a distinctive user interest profile by calculating an affinity score for key activities, including eating/drinking, fast food and coffee shop categories.

geo-twitter-ad-targeting-03

Step 4: Form user segments with high affinity scores for key categories

We matched user profiles by their similarity to form distinctive target user segments that had high affinity for eating/drinking, fast food and coffee shop categories, collectively named “fastfood” sample.

Step 5: Extrapolate “fast food” sample on the whole population of central Madrid area

After aggregating and extrapolating the “fast food” sample on the whole population of central Madrid area, we developed API to integrate with our trial partner’s hyper-local self-serve ad targeting portal. The API was feeding lat/lon records of 500mX500m polygons with an affinity score for “fast food” categories. To visualize the API, a heat map was created showing polygons with varying affinity scores – the darker polygon is, the higher its affinity for fastfood. We also enabled a filtering option that could show how the affinity score changes by hour.

geo-twitter-ad-targeting-04

Step 6: Campaign launch

Fastfood brand’s marketing manager used our partner’s hyper-local self-serve ad serving portal to plan and run a targeted ad campaign by making data-driven decisions of WHEN & WHERE to send mobile ads based on Locomizer’s extrapolated view of footfall by fastfood interest and time. Our partner bought audience (any mobile audience available in that areas, not limited to Twitter users) in the specified polygons and delivered the ads.

Campaign results

Overall, the campaign was a great success as Locomizer has outperformed the industry standard CTR benchmarks for similar location-targeted campaigns. Locomizer pinpointed areas and time slots with audience highly receptive to fastfood’s ads, driving up CTR by 40% in comparison with CTR in areas blindly targeted by fastfood ads.

Locomizer analytics drove the coupon conversion rate up by 30% among customers who clicked on the ad, that’s an incremental increase in footfall of 7,000 customers in fastfood restaurants in one month.

Conclusions

  • Twitter geo-data can be successfully monetized despite its small share out of all tweets (indirect monetization)
  • People interested in fastfood are not always or necessarily can be found in close proximity to fastfood venues: knowing how footfall interests change with time can significantly increase the effectiveness of your targeting campaign.
  • Locomizer is breaking the traditional location-based targeting to move to a more effective dynamic geo-behavioural model.

For questions and enquires please email us: info [at] locomizer.com.

Part 2: What is a geo-behavioral user interest profiling?

Creating user interest profiles, as discussed in Part 1, is just one side of the story, the second part is that we can aggregate and anonymise those profiles to extrapolate on the whole population. We call it the Geo-Behavioral Interest Graph (The Graph) or Footfall API when we want to monetize it as a product. Imagine getting a detailed, rich, contextual knowledge of any given place by people interests and day part. For example, if an advertiser wants to know when and where to target people interested in buying a car in New York area, the Footfall API can answer this question by querying our interest graph. The Graph can be visualized through the creation of interest heat maps as shown in the video below.

How do two million tweets look on a map?

madrid_tweets_sm

We build people interest profiles based on any kind of location data, including geo-enabled tweets. This week I am going to share a mobile targeting use case locomizer did based entirely on tweets with location feature switched on. Ever wondered what will be the map look like if you put a dot for every geo-enabled tweet sent within a two-month period in a large city like Madrid? Well, that was what we did – we pinpointed the location of each and every tweet out of two million in order to build an interest heat map for Madrid. Stay tuned to learn about what kind of result we got from this trial. It will be posted this week.