Constant improvement is a mantra here at ShopLocal. To that end, the team knew there was room for improvement in our popularity ranking algorithm that we call “Wisdom of the Crowds” (WoC).
The current system for determining product popularity only counted those products that a shopper specifically clicked on. There are a number of product level clicks that are counted in terms of a shopper expressing interest such as a product detail page, a buy online click or an add to a shopping list click (which are cumulatively called Expressions of Interest). Not any more. The ShopLocal team has begun collecting and counting another key piece of consumer interest data – product rollovers.
By factoring in both product rollover data AND expression of interest data, the overall Wisdom of the Crowds ranking data is all the richer and accurate. This process of collecting and using rollover data for SmartCircular sites is underway and will complete over the next month or so for all clients, SC sites and versions of SmartCircular.
At a macro level, what we have learned so far about rollover data is:
- Rollovers are on average about 5x as common as Expressions Of Interest. This means by adding them to the Wisdom of the Crowds calculation, the overall popularity ranking data should become much more accurate and more truly reflect actual consumer interest.
- Rollover usage varies significantly by retailer. No surprise given all the variation in UI/UX out there within weekly ad sites. Also some sites don’t even use rollovers. Or on other sites, the rollovers are down played. In any case, we are now tracking them all so as to extract the most useful and complete set of consumer interest data as possible.
Illustrated here is a specific product level rollover that has been triggered by a user rolling over a predefined hotspot region that is overlaid on top of the circular page view images. When talking about rollover tracking, what is actual being tracked is the number of times, for a specific product, a rollover is triggered and left open for at least 2 seconds. If it meets that criteria, it would be counted as one (1) rollover for the product.
Here is a quick snapshot of some early returns from a 6 day sample set of data across 20 retailers:
- # of total product level exposures: 1,006,078,269
- # of total product rollovers: 25,620,887
- # of total product specific indicators of interests expressed: 4,532,430
- Overall ratio of interest divided by rollovers: 18%
- Rollover trigger rate (of overall total product exposure) : 2.55%
- Interest trigger rate (of overall total product exposure): 0.45%
- Total CTR (rollovers + interest): 3.00%
UPDATE 1/21: The ShopLocal analyst team just today cranked out a nice little additional stat on this topic that is worth sharing. The correlation of interest/exposure ranks compared to the rollover/exposure ranks, on average overall is 0.6636 Meaning that rollovers are for most sites a good measure of interest. Maybe think of it this way. This correlation helps prove that the items a user rolls over are likely the very ones that they end up taking other subsequent action at a later point in their site visit.
Not all sale offers are created equally. Out of the roughly ~200 to ~300 items that are featured in most weekly ads, there are some that are more important to a retailer to promote and/or sell. This may be driven by many factors such as:
- Supply issues (overstocked)
- Profit margins
- Store traffic (loss leaders)
- Vendor co-op funds (retailer is paid to promote certain items by the manufacture)
To help bring focus to this common problem, ShopLocal has created the ability for a retailer to create a subset of weekly ad items (typically called ‘Featured Items’) that can be uniquely promoted at a higher level. The main visualization mechanism that ShopLocal uses to push additional user awareness and engagement is a SmartStrip which is an interactive horizontal or vertical carousel that displays these top items, on a localized basis. There are three main methods for populating the specific items that will appear within a SmartStrip:
- Manually: In this case, the retailer specifically indicates which items should appear and the order in which these featured items should appear in
- Business Rules: This allows a more automated approach that draws the content for the featured items area from for example, the front and back covers of the print promotion.
- Wisdom of the Crowds: By choosing this method, the list of featured items that are displayed within the SmartStrip are constantly changing, due to the ebb and flow of what the overall user community is deeming popular by how they express their interest to the retailer’s sale offers
Walgreens uses a vertical SmartStrip that is manually populated by the retailer.
Kohl's uses a horizontal SmartStrip that is populated via the twenty-five (25) most popular items overall (across all categories and brands) for the user's closest store location.
Wow. Talk about coming full circle. Here is what this new reverse run of press (ROP for short – these are the ads that are published in-line within newspaper content and not inserted like a circular/flyer/FSI) ad publishing product is all about.
- Step #1: Print Circulars All ShopLocal content begins first as print circulars. Typically this equates to starting life as a multi-page, multi-layer, HiRes Adobe PDF file. The same asset that is sent to a commercial grade printer to mass produce say 50,000,000 (yes million) copies on any given Sunday.
- Step #2: Digital Circulars From here, this print PDF is broken down (or decomposed) into many unique individual parts and stored in a relational database that ShopLocal hosts. At this point, the local promotional (e.g., circular) content is fully digital and easily distributed via whole hosts of digital mediums including search, display, mobile, widget, microsite, etc.
- Step #3: Crowd Sourced Popularity This digital circular content is displayed to hundreds of thousands or up to millions in some cases of in-market local shoppers within the advertisers SmartCircular (or online weekly ad) site. The result of all this exposure and consumer interaction is amazing insights into what products and offers are the most popular. ShopLocal calls this ranking data the Wisdom of the Crowds. This crowd sourced popularity data is then used to filter out all of the products that the large audience did NOT find engaging or interesting. Rather, only the most engaging circular content makes it to the last step of this process.
- Step #4: Print ROP Ads The most popular circular content (as determined by the advertiser’s own customers) is then inserted into a pre-defined and pre-approved print ROP ad template. The combination of these “best of the best” local promotional offers combined with the great reach of the newspaper is what makes this a recipe for success.
So in summary from print (circular) ads all the way back to print (ROP) ads. That is essentially the life cycle of the circular content for this product.
This is the core problem that reverse ROP publishing solves. A print Sunday circular ad quickly gets tossed in the trash can / recycle bin by Monday or so which leaves a real hole as far as ongoing reach and distribution for the advertiser. By inserting targeted and timely ROP ads back into the newspaper, the advertiser is able to reach these shoppers at the right time when shopping activity is starting to pick back up on the days leading up to the weekend - Thursday or Friday.
Enough talk. Time to show everyone a few samples of what this type of advertising is all about – see the two (2) below. To better define reverse ROP ads, they can be:
- Geo-targeted by unique newspaper and print version. So in Gannett’s case, there are over 80 unique local newspapers to choose from. This helps ensure that the same print versioning (eg price and product assortments) that are created and applied to the print circular content can and will be re-applied to the creation and placement of these reverse ROP ads
- Flexible in the number of products shown within the ROP ad template (the template however needs to be consistent from market to market and week to week)
- Flexible in the ROP ad dimensions (height & width)
- Flexible in the placement within the local newspapers
- Able to be distributed on either the Thursday or Friday following the typical Sunday circular ad drop / newspaper insertion
- Circular products for inclusion into these reverse ROP ads can be chosen based upon:
- Wisdom of the Crowd (WoC) popularity
- Front circular page / back circular page location
- Category of the items (which would allow for some level of contextual targeting – combining let’s say the category content within “LCD / Plasma TVs” with the “Sports” section placement of the ROP ad)
- Brand of the items
- Additional filters that can be applied to the above content selection logic include
- Item price (so for example exclusing all the items under $5.00 regardless of popularity)
- Item sale end date (no sense advertising products that will for example not be on sale very shortly)
Here is a real reverse ROP ad that Gannett created for Best Buy and ran across a handful of it's top local newspapers. This is an ongoing process on a weekly basis to turn the digital circular content into a print ROP ad
Here is a standard reverse ROP "template" that Gannett uses to showcase the overall product concept
So in today’s WSJ there was a rather fitting article titled,
Look at This Article. It’s One of Our Most Popular
Top-Ten Lists Abound Online, but Following the Herd Can Make You Wonder About the Wisdom of Crowds
By CARL BIALIK
In it the author makes the point that when a site publishes a “most popular” list, that this type of presentation actually distorts the overall ranking, because the perception of popularity alters users’ behavior. For some entertaining examples, read the complete story. This in general makes sense. By showing lists or result sets ranked by popularity, the site is to some extent creating a self-fulfilling prophecy. The items at the top of the popular list will ONLY continue to become even more popular and thus remain at the top of the list by the very fact that they are first.
Companies need to be careful and take the necessary precautions when creating and using crowd sourced popularity so as to not create bias and affect consumer preferences
However, ShopLocal’s own Wisdom of Crowds methodology does not suffer from this problem. This is because the presentation of content within the SmartCircular sites is not affected by the Wisdom of Crowds ranking (e.g. it is not used currently for things like result set sorting order), that this ShopLocal generated popularity data (WoC) is an unbiased measurement of user preference.
Since one of the current near-term road map items for SmartCircular 4.1 sites is to begin consuming and using this popularity-based rankings in areas such as category, brand and keyword search result pages, the ShopLocal team will make sure to exclude any user responses to featured offers from the Wisdom of Crowds ranking so as to keep this problem highlighted in the WSJ article above from occurring.
Special thanks to our own Loch Rose for pointing this fun read out and doing the initial analysis.
(4.00 out of 5)
Loading ...
(4.50 out of 5)
Comments: None