Try this offline experiment this holiday season:
Walk into a Tiffany’s with a bag from JC Penney’s and see if anyone will talk
to you. Next, go to JC Penney’s with a bag from Tiffany’s and watch the
staff fawn over you.
Online personalization is an evolving and very dynamic space. The algorithms
behind personalization sift through reams of data and supply us with tailored
recommendations, advertisements and the most personally relevant and appealing
results.
Because online personalization is on the cutting edge, we sometimes forget that
we have been targeted for decades while shopping in the brick and mortar
stores. Old-school targeting faces a similar set of problems. Old-school
retail targeting is about visually sizing up customers:
Vivian, played by Julia Roberts in the movie, “Pretty Woman,” enters the
posh shop where a day earlier, dressed in her “working girl” clothes, she was
snubbed by the shop assistant. Now, wearing a beige dress, black hat and white
gloves–the very image of elegance–she is loaded with shopping bags. The shop
assistant smiles and asks if she needs help.Vivian: I was in here yesterday, you wouldn't wait on me.
Shop assistant: Oh.
Vivian: You people work on commission, right?Shop assistant: Yeah.
Vivian: Big mistake. Big. Huge. I have to go shopping now.
The shopkeeper’s targeting mistake was fictional. But every day, millions of
dollars hang on sellers properly sizing up their prospects.
Never in my life have I been referred to the women’s department when entering a
clothing store. The employees quickly surmise that I’m not looking for a dress.
They might be wrong. I could like wearing dresses, or, I may be buying a gift
for my wife, mother, or daughter. But they play the odds and refer me to the
men’s section. Additionally, because I’ll never be confused with a fashionista,
they are likely to recommend a new pair of khakis rather than the luxury
cashmere hoodie made from the soft wool of Mongolian goats.
Online personalization is similar. When you visit that same clothing
retailer’s online site, the algorithms attempt to determine something about you
and customize the content. It’s all about probabilities – the better the data
and the better the algorithm, the better the chance they will hit the mark and
show you something you like.
Of course, just as I can cloak my identity online using a cookie blocker, I
could also cloak my identity offline. If I do not want a store clerk to
know I am a man, I could wear a large parka and mask (though they might call
security to escort me out). If I cloak my identity online, I’ll get a less
interesting experience or just be ignored.
Old school targeting happens everywhere. I know a single woman who wears a
wedding ring when she travels because she gets better service from the airline
and the hotel. Sadly, society still discriminates based on appearance. In
old-school targeting, we sometimes receive wonderfully customized service, but
are also vulnerable to unpleasant experiences due to our looks, what we wear,
our gender, or the color of our skin.
Humans naturally react to others based on many factors – some legitimate and
some not. Store clerks need a filter to triage purchase intent because they
can’t spend equal time with everyone who walks into the store. They need
to allocate time to people they believe are going to buy, so they develop
heuristics, based on their own biases, to help them interact with customers. If
you walk into Tiffany’s with a JC Penney’s bag, you might be a billionaire, but
my guess is that the store clerk isn’t going to take the time to find out (unless
it’s a great sales clerk who notices your brand of shoes or the wristwatch you
are wearing).
Old-school personalization is supposition based on our five senses. When
practiced well, it can be accurate. Online personalization lacks many of
the cues one gets in the offline world, and, like store clerks, there are good
and bad online systems. Good systems unify online and offline data and
work on enhancing the experience of the customer.
Computers, like humans, have the same potential to do great good or cause harm.
But unlike humans, computers can quickly digest millions of different
interactions to make decisions. For instance, BestBuy.com can query its
database to check if you are an existing customer and what you purchased in
your last visit. It would be impossible for a store clerk to remember the faces
and purchases of millions of customers.
Stores, both brick-and-mortar and online, want our business and try to seduce
us for it. Except now, in the age of “Big Data,” it’s easier for the ones
online to be masters of seduction. One hundred years ago, John Henry
beat the machine at a severely high cost. Today, he wouldn’t even stand a
chance.
Vivian tells Edward: “The stores are not nice to people — I don't like it.”
This is after she was snubbed.
Edward responds: “Stores are never nice to people. They're nice to credit
cards.”
Special thanks to Ken Treske, John Battelle, Caitlin MacDonald, Brad Justus, and others for their inspiration and advice on this piece.
I find that when I shop in stores, the clerks are often sizing me up and judging me. It was a great relief when I started shopping online ten years ago. When online, I feel less pressure to make a purchase, can take my time to make the right decision, and generally have a happier and more pleasurable experience.
What’s your point?
People and machines are different. Machines do some things very well; much better than people and vice versa. It is when they work together, that great things happen.
For example, at one hotel chain, the doorman will get your name off a luggage tag or similar method. He will call that in to the front desk so they can greet you by name and have you customer record in front of them so your preferences and any special needs are confirmed and they can speed you to your room. That is personalization, uh, what we used to call “Personal Service”. Whatever. It makes you feel better.
We do the same things to the stores. We don’t shop for luxury goods at J C Penny, and we don’t go into Tiffany’s looking for bargains.
When I shop online, I am looking for low prices and a fast transaction. When I go into stores, I am looking to hold and see what I am buying, and I expect the immediate gratification of walking out with it.
Auren, great point on how profiling has always worked. However, I’m not sure machines can eat people on this front yet.
Big data gives retailers an opportunity to recommend products and categories, but it cannot yet read my reaction in real time (note a click is a result of my reaction, not my initial reaction) and respond appropriately. Online retailers will continue to get better, but I think it will be some time yet before a site can seduce a potential buyer into purchasing more than a good sales person can.
It will be an interesting thing to watch for sure.
Some stores are installing tablets (well, iPads really) throughout stores to help consumers personalize their shopping. Instead of interacting with a sales assistant, the consumer interacts with a tablet computer which provides suggestions based on previous sales history. A good personalization system would also be of use to first time shoppers at a store. Sales assistants can then focus on other areas of the shopping experience.
It should also be possible for consumers to connect to a store’s personalization system with their phones and tablets while visiting the store.
Most people prefer to go shopping at physical stores as it is a social experience that is usually mixed-in with other socializing such as having something to eat or a coffee.
One thing we see offline but not quite online is negative feedback on a product. If I am at JC Penny, the salesman takes cues not only from what I like but also from what I don’t like (common things being common, most likely at the beginning negative signals will dominate the positive ones). Online, for some reason, that is less prevalent. I have yet to see an Amazon, Wal-mart, etc add an “x” to let people now I don’t care about the product. Not buying, leaving the page to quickly, deleting from shopping cart and other proxies are not IMO particularly telling. Then again, I am not in this business and don’t know if they indeed work or why online retailers are not making more efforts to gather what people Don’t like.
Am I correct in my assessment? If so, why do you think this is the case?
Another good, informative read. Thanks Auren.
Happy Holidays!
The unwritten point is not what JC Penney or Best Buy can do with their customer database information they have collected on our past purchases. But, really how Big Data and Predictive and Prescriptive Analytics will enable businesses to be better prepared – well ahead of time, for our planned purchases.
This will stretch one’s thinking well beyond what any retailer is doing today, but it’s coming and soon. What we tweet about, where our car drove, what pictures, videos and images we have uploaded to Facebook and Instagram, local news stories, what people around us are doing/tweeting/emailing, websites we’ve visited, and perhaps even our overheard conversations will drive what JC Penney and Best Buy (or their successors) present to us when we walk in. Even the manufacturers of goods in those retailers will be changing, in real time, and mass customizing goods and services for us as we walk in the door.
This new field is not being driven by retail, or the usual consumer analytics companies. But “mood” or “sensitivity” analytics is coming to marketing – and it’s anchored in Big Data – REALLY BIG Data (not just the retailer’s database).
Great article.
I had a helper who colored his hair green and wore an earring. When he parked his car in front of my house, a neighbor called the police.
I was snobbed out of a store, showing up in ragged jeans when I was actually ready to spend money for a gown to wear at your wedding…
Most people purchase things within a narrow range. Most people have an insatiable appetite for new things but only when it falls into their comfort range. The real value for retailers is in their existing hard data particularly loyalty card data. I suspect that in most cases, modern machine learning methods are not being applied to the data that retailers already have. The external data mentioned is available to all retailers and so ultimately offers no significant competitive edge.
The assessment is correct. It is the notion of relevance feedback ie. shown a list of things, select the ones I like and show me more like those (positive feedback) or, point out the ones I don’t like and remove things like those from the current list and future lists (negative feedback). 😉
Hi Auren,
Store clerks don’t follow you out of the store.
The place where “Big Data” and adtech are most
starting to set off people’s creepiness alarms is
in retargeting.
I just put up this piece on related subjects.
http://zgp.org/~dmarti/business/adtech-the-end-is-near/
Don
Went to the same high-end retail store to demo some new watches. Wore same outfit both times.
Visit #1: Wore no watch – waited 10 minutes for salesperson to acknowledge me. They finished a phone call 15 minutes away from me and walked away from me mid-demo to help another customer with a refund. After the demo I have to ask for product info.
Vist #2: Wore Breitling watch. Two salespeople immediately jockey for my attention. The winner immediately introduces himself and acts as if we are old buddies. I try on several different watches with the salesperson’s full attention while other customers wait their turn. When I am done, the salesperson offers to e-mail me detailed product information and photos (including a photo of the watches on my wrist) with no further follow-up unless I want to proceed. He then escorts me to the exit.
Using the watch example from Sparks, I’ve had the same situation occur in a watch store, and, if offline data were married to online data (assuming, in a consumer file somewhere, it is noted that I own a Montblanc and Baume & Mercier watch) the online personalization experience could be a great one.
Marketing to me online, knowing I have several watches and showing me similar styles of watches or complimentary goods without me having first to go to Ashford.com or Torneau’s website to drop a cookie for retargeting purposes is they key…just the same as if I walked into Tourneau and the store clerk remembered me from my last visit, even if my coat was over my watch.
Another other interesting thing to ask in relation to your article Auren is this: how does online marketing and targeting, with or without offline data matching, account for multiple users using one account, browser session / logged in user session on a computer?
For example, my girlfriend often uses my iPad to check something out, and I rarely delete any session data, cookies etc. So, data is being collected about both of our browsing habits, stores we’ve visited, products we’ve purchased etc but we may be buying two very different types of things.