supermarket research/loyalty cards 3

[lionelhill]

Anybody got any thoughts on the future of loyalty-cards as a method of market research amongst supermarkets?

Although I'm a big fan of data-mining I'm a bit scared they are getting too much attention. After all, they can only tell you what your customers are buying. They can't tell you:
(1) why your not-customers aren't there.
(2) what your customers would love to buy, but you don't sell.
(3) why your customers aren't buying something (i.e. you sell it, your customers want it, but buy it elsewhere because there's a problem with yours)

These seem quite big weaknesses to me, but although loyalty cards/data mining have had several big articles in the quality press in the UK recently, I've not seen these problems mentioned.

Data-mining has also led to conceptualising "typical shoppers", with typical shopping patterns. But my wife chooses some items from Marks and Spencer (sometimes on grounds of cheapness!), and others from Lidl (sometimes on grounds of better quality), which any UK resident will know is a total social no-no, and completely the opposite of predicted behaviour of typical shoppers. Where do we fit in?

 

[Dimandja]

lionelhill: They can't tell you:
(1) why your not-customers aren't there.
(2) what your customers would love to buy, but you don't sell.
(3) why your customers aren't buying something (i.e. you sell it, your customers want it, but buy it elsewhere because there's a problem with yours).
These seem quite big weaknesses to me,...

Those are not weaknesses at all. I think there are too many assumptions made here. The least of which are:
(1) How long will you wait before deciding this? Just because you started issuing cards does not guarantee that the whole world will join, does it?
(2) You are assuming that data from the competition is safe, right? Think again; companies that collect data about you, make even more money selling it. When was the last time you received credit card offers based on information gathered elsewhere?
(3) See (2).

Dimandja

 

[jrbarnett]

They also don't tell you the purchases that are made and not registered. For example, there is a small branch of a well known UK newsagent/stationer near where I live, based in the railway station.
This particular branch does not have facilities to process the loyalty card claims, so while I use this to buy a lot of IT magazines, as it is the best local newsagent for doing so around here, none of these have been registered on my account.
For branches further away, I tend to buy CD's and books at other branches.

John

 

[RiverGuy]

1.) I agree with you there.
2.) I think this can be discovered, to an extent. You can mine out what types of products your customers are most likely to like based on related products that did well, and make a pretty good guess of what to sell.
3.) I think you can figure this out as well. If a product has been getting a lot of returns for defects, then obviously the product has a problem. Also, by comparing the price of competitors versus yours, if yours is routinely higher, then that would more than likely be the cause of it.

 

[MDXer]

Remember datamining doesn't tell you what people will do but rather the likely hood that people with a common set of attributes will follow the same pattern.


In the case of customer loyalty cards at a supermarket. Datamining won't tell you that 500 of your customers will buy a new brand of beer. But it may tell you that 60% of Men who are single with an average income of 30K-40K a year who buy ligt beer that is less than $10 a case might but your new light beer that retails for $8 a case. Datamining also can tell what types of products a store may not want to carry. In the same scenario if that demographic is 80% of your customers you may not want to sell French Wine that is $250 a bottle.

As for the data being collected you would be suprised what is collected without peoples knowledge. In the US a large portion of Peoples Prescription data is captured, warehoused and patterns reported to pharma manufacturers. This is all done within the HIPPA compliance guidlines as well.

"Shoot Me! Shoot Me NOW!!!"
- Daffy Duck

 

[Dimandja]

I must confess: I don't get it.

Are you guys worried that these companies are not collecting enough information about you?

Also, tracking what customers buy is not the only reason for the existence of those systems. There are so many business reasons to issue cards! For example, the very act of issuing cards enhances the image and exposure of a company considerably - whether or not they are interested in collecting your personal data.

Also, collecting data comes with its own challenges such as storage and processing. Your shop may only be interested in tracking a small portion of their inventory, if at all.

Dimandja

 

[RiverGuy]

The system is designed poorly, in the U.S., anyways.

For example, one can acquire a card and use it, taking home the sign up sheet to fill out and return later. But turning in the sign-up sheet is not a requirement. Therefore, there are customers in the "no man's land" zone, contributing to patterns and statistics that cannot be linked to particular areas or age groups, etc.

 

[lionelhill]

Dimandja,
I'm worried that supermarkets may be wasting a lot of (my)money developing schemes to find out (approximately) what any checkout operator probably knows already (fairly accurately). If you're going to use IT, you might as well use it well, and it seems to me the supermarket approach is full of holes.

Mostly that worries me because I'm gradually running out of useful supermarkets! They keep introducing loyalty card schemes, and six months later dropping all my favourite products (and the vegetables turn manky). Maybe I'm just a mutant who doesn't fit into a convenient group.

RiverGuy,
I think you're right that guesswork can fill in the blanks a bit, but it's not the whole answer. For instance "returns" are a poor measure of quality of fresh fruit and veg. (1) most people just won't buy a bad lettuce, and (2) if you get home and find a 54p cabbage is rotten inside, you usually swear a bit and throw it away. It's not worth returning!
Also if you don't sell parsnips, how can you work out I wanted one from the fact I bought an extra potato to roast?

I find this an interesting area where data-science and psychology/sociology meet...

 

[MDXer]

linonel,

They won't know that you wanted a parsnip instead of the potato.

As for the rotten cabbage, the best way a store can track this is buy bringing other systems into a data warehouse and report off these systems as well. Most purchase or inventory systems carry fields for Quality, Rejections, Expiration date. If the produce manager is trashing cabbage in 3 days when he knows it should be good for 7 then by tieing these systems together he can see that maybe they were of poor quality.

Data Mining won't give someone all the answers, but Data mining as a component of a solid Business Intelligence System Will. Companies that have Datamining typically have a OLAP system already in place.

"Shoot Me! Shoot Me NOW!!!"
- Daffy Duck

 

[johnherman]

lionelhill
Of course you are right that data mining is not the next killer application, although some in the media and those who do marketing for DM companies would disagree for obvious reasons.

Additional kudos for your ability to think outside the box and determine what metrics and indicators would provide excellent ROI for the market scanners.

If the media HYPES it, it MUST be good, RIGHT?

Vendors, to have your product hyped by an independent consultant, send cash, gift certificates, or stolen credit care numbers to:

OnTheTake@BigMedia.com

and we will speak highly of you until the next collection cycle.


Sometimes the grass is greener on the other side because there is more manure there - original.

 

[lionelhill]

A potentially scary thought process goes:

... if your competitor's got data-mining and they're doing well commercially, can you afford not to follow suit? And once you've got it, can you afford to tell the share-holders (and schadenfreude popular newspapers) that you wasted an awful lot of cash setting up a system that didn't give you any useful information...

So my feeling is that even where it doesn't work (or not very well), we're doomed to see more and more of it. But I note a few stores have quite openly gone the other way (the "We know what our customers want, and they don't want gimicky plastic" approach).

 

[MDXer]

lionellhill,

I'm willing to wager that more Datamining and BI implementations fail than succeed. Most common reasons I have been exposed to is
1) Data Quality or Lack of Data
2) Unrealistic expectations of the result will bee
3) Poor Planning.

1 and 2 are issues that you touched on. You can't work or extract answers from data you don't have or is so dirty or sparse it is unuseable. Also many people don't fully understand what it is they are investing in. They think that because in a demo or discussion they saw the ability to do X then it will automatically do Y and Z.

"Shoot Me! Shoot Me NOW!!!"
- Daffy Duck

 

[octothorpe]

Let's not forget that the loyalty cards and data mining projects are not always linked. The loyalty card allows them to see repeating customers and handle direct marketing (if you fill out the form). But most supermarkets use their big-iron data mining for inventory control rather than marketing.

It is certainly quite difficult for data mining to predict particular customers' buying habits, but it works great for figuring out how to minimize shelf time for product (which is especially important for perishables). It also works well for figuring out which products to use as loss leaders to bring people into the store (like selling milk at a loss to increase sales of eggs).

Checking the produce section of a supermarket is usually a good indicator of who is in charge of the company. Stores that have consistently fresh (but occasionally sold out) produce are usually run by operations managers who focus on tight inventory controls, while mediocre produce usually points to marketing managers running the show.

 

[sha76]

Any system's going to also depend on the calibre of people using it. A certain well know British chemist chain has all my details linked to my card, including the fact I don't have children. I bought a multi-activity cuddly dinosaur for a friend's baby & ever since I am bombarded with special offers on nappies! Maybe the software's not always the limiting factor on the quality of their market research!

 

[jwenting]

For example, one can acquire a card and use it, taking home the sign up sheet to fill out and return later. But turning in the sign-up sheet is not a requirement. Therefore, there are customers in the "no man's land" zone, contributing to patterns and statistics that cannot be linked to particular areas or age groups, etc.

In the Netherlands (and possibly elsewhere) companies are required by law to not require customers send in any personal information in order to get a card.
Privacy protection.

Card will still collect info which can be useful but can never be traced back to an individual address in order to send advertising etc. or be used for social engineering.

It's the ONLY way in which I ever participate in such programs, no need giving marketeers more info about me than they pay me for in discounts

 

[Dimandja]

I think octothorpe and sha76 articulated fully the argument I broached on earlier.

If all you have is a loyalty card, there is no way to know what the store is really using it for. Therefore it is highly premature to claim that such systems don't work.

I have designed such systems before, and inventory was usually the deciding factor; also marketing to a certain extent. For example, if you wanted to temporarily/selectively change the price of an item, it is far cheaper to do it through a system connected to the customer card, than by manufacturing a steady stream of "sale" stickers.

You may have a idealistic view of how such systems should work; but, often the reality of it is what companies use to maintain a competitive edge. Translation: even if you supoena the store to tell you what they do the the "scant" data they collect, they may still not tell you.

There is a myriad facets to data mining. Most of which many of us here cannot begin to guess. Often, the stores themselves only stumble upon the potential of the data they managed to collect.

You should rejoice that your store is not collecting more info about you, rather than trash their system because it does not care to know who you are.

Dimandja

 

[lionelhill]

I'm not angry with the store for collecting data as such; I'm sure the big supermarkets know less of my inner private life than would a small shop-owner in a village, and I don't suppose Lord Sainsbury gossips much about my choice of aftershave. I suppose I'm just dubious about it as a panacea. Like sha76, I've had some mis-hits. Like being bombarded in disposable nappies ("we know as a young mother you're very busy..."), which suggests whatever marketing genius thought that one up missed the fact I'm a bloke (not that blokes can't change nappies too... oh dear, I'd better stop typing.)

Thanks every one for interesting insight into how data might be used.

Octothorpe, thanks for comments on stock control. Now I understand why one local supermarket went from never having any milk left on Saturday afternoons to having plenty, but really naff fresh vegetables.

 

[bbarr]

data mining is more about developing a set of probabilities to hopefully increase the likelyhood of a particular set of actions.

for example, if 70% of males age 25-35 buy beer A and the customer is a male age 25-35 then there is a 70% chance that he will buy beer A. Therefore, if you were planning a mass marketing campaign to men in this age range for a couple of beers and the other beer (B) has a 20% probability, then it would make sense to send the material promoting beer A.

Now, if the campaign was internet based and you knew the customer often bought beer B but not A, then you may set up some logic that makes the decision to overide the general rule for this particular situation.

Data Mining is a tool that can be used well or poorly.

And because it is a tool, it has it's limitations. While it may be able to provide a probabilty for future actions, it may not be able to tell you the factors behind that behaviour, unless there is data such as from a market research survey and supplemental regression analysis.

Developing effective systems to provide useful / actionable information is difficult and requires a number of tools and channels.

Hope this helps.

- Bruce

 

[lionelhill]

Yup, agree. Collecting the data is easier than interpreting them. Frankly, bad interpretation of good data is probably more harmful than blundering along without any data at all. But that's just a personal view.

 

[chiph]

For example, one can acquire a card and use it, taking home the sign up sheet to fill out and return later. But turning in the sign-up sheet is not a requirement. Therefore, there are customers in the "no man's land" zone, contributing to patterns and statistics that cannot be linked to particular areas or age groups, etc.

Ever watch "The Blues Brothers?" Elwood Blues had his home address as 1060 West Addison (aka Wrigley Field).

Whenever I sign up for one of those cards I make up a name and address. I like the discounts (it's about the only way to buy anything on sale anymore), but I'd rather keep my demographics private.

Chip H.


If you want to get the best response to a question, please check out FAQ222-2244 first