Moment marketing and how predictive AI algorithms can help


Ensuring that products are not over or under exposed compared with demand is a key success factor in online selling.  Our experience tells us that both manual and rules driven approaches are unable to achieve the ideal balance.

Take the example described in the chart below, here we see that products are often under exposed at the start of their life and over exposed towards their demise in popularity.  This example is for just one product, the problem is magnified in line with the size of a product catalogue.

Fashion retailer - ecommerce data on sales and exposure

Above: data collected for an advertised product from a fashion retailer. X axis holds time units – see them as days or hours. The purple columns stand for sales volumes; the orange ones represent the volume of digital product exposure done with manually set digital campaigns.

Digital merchandising

The physical world of retail campaigns and transactions had an inherent direct shop-client contact but the move to the internet brings a new set of challenges to merchandisers.

Whilst the physical world meant that an experienced merchandiser with a nose for products in demand could adjust exposure ‘on the fly’.  This cannot apply in the online world as the shop approach does not automatically apply here.

Data processed in the Apptus Cloud subtleties in product demand that go beyond human capacity for data resolution and analysis. Retailing is continually swept by microtrends -- the hits-of-very-short-moments, and because ecommerce enhances speed of uptake and wider social influencers products go viral faster in the digital world than in offline commerce. In addition, commerce and campaigns are permanently affected by unpredictable phenomena not directly connected to the specific market, for example a summer music-hit, an unplanned VIP, an abrupt change in weather, a sudden political change or event…

Best seller of the month, or best seller of yesterday – which one to focus on?

To exemplify further let’s look at book retailing. There is a historic rationale behind the traditional period-based best seller list (the Top 10 of the year, or of the month):

  • People buy what is popular
  • Showing the best sellers makes people buy more
  • Best seller lists make people buy from the head of the list and not from the long tail.

However, as bookstores went online the game changed.

Best seller approach has changed with digital marketing

People got now easy access to the long tail. Consequently, making people buy more now requires the retailer to:

  • Expose more products
  • Make the list more volatile
  • Accept that the last month's best seller may not be the same as yesterday’s best seller.

The last point brings the dilemma: What should you expose more: the consistent seller, which confirmed on a long-time frame (say a month), or the hit of the day?

Moment marketing and how to stay relevant at any given moment

There is an emerging concept of moment marketing which can answer the question of staying all-time relevant. “Moment marketing is the ability to take advantage of an event to deliver relevant and related, seemingly spontaneous, and often fleeting interactions with customers in real time” [1].

To cite Randy Wootton, an advocate of this new field who is soon to publish a book [2] on the topic: “moment marketing requires companies ingest lots of data in near real-time and take action in less than 20 milliseconds. That’s where AI comes in. You need insight engines to mine all this data or you’ll get overwhelmed and drive your customers crazy. AI can help us discern the relevant data and then deliver advertising at the right time, in context” [3].

Product exposure in Apptus eSales algorithms

The Apptus machine learning algorithms have brought this moment marketing approach into reality. The recommender system we offer goes beyond what is easy (i.e. determining the best seller for an interval), to solve the complex: finding the right balance between long and short term best sellers at any given moment. The algorithm is valid beyond bookstores and applies to any retailing segment, including sectors as fashion, DIY, electronics and furniture.

Find the right balance for digital product exposure between microtrends and long-time frame trends

Listening to data coming in real-time, the Apptus algorithms are devised to react and trigger a product exposure that stays at the highest level relevant to the moment, and furthermore to the individual - the product search history and the data coupled to the visitors' behaviour on the site.

According to Wootton “we are still in early innings of understanding the possibility of data. […] We are going to see computers make decisions more and more often. Marketers will have to surrender control and trust. Otherwise they’ll stand in the way of progress – and higher conversion. […] The opportunity to stitch online and offline together thru predictive marketing is radical”.

Learn more on Moment Marketing

  1. William Grobel et al., Webinar on Moment Marketing, WARC Toolkit 2016
  2. Randy Wootton, Eric Duerr, Ken Rufo - Moment Marketing: Optimizing Every Step of the Customer Journey, 2019
  3. The Right Ad, Just When you Need It: Talking “Moment Marketing” with Randy Wootton