Mike Anthony @ engage consultants

Mike Anthony on Shopper Marketing

One Simple Thing That Will Change The Way You Look At Shopper Insight Forever

with 16 comments

I’ve billed this blog with a big title – and to any of you in the research community perhaps I’m over-promising. To some of you this will be obvious, and something that you knew ages ago. I’ve known it for ages too. But here’s the thing, and here’s why it’s worth all you shopper research and shopper insight guys reading on. Most people don’t get this. 

So here it is.

Shopper research data does not measure the number of shoppers who do something. It measures the number of shopping trips.

See? Simple right? But most researchers or marketers don’t get it. Most quantified data is presented with a headline that would read something like “45% of shoppers visited the category”. Pretty straightforward. That means that 55% did not, and that’s an opportunity.

But, that isn’t the whole story. Actually what was measured was not shoppers, but shopping trips. People do not shop the same way every day. So the ones that visited the category today, might not visit the category on their next shopping trip. When research companies produce their segmentation model for example, “24% of shoppers are frugal hunters” that is misleading. What it should say is that on 24% of shopping trips shoppers behave like frugal hunters. This is intuitive – we know from our own behavior that we can behave completely differently on a different day – but the consequences of this are not.

Let’s take an example. I go to a supermarket twice a week. About half the time I do a big shop, buy lots of stuff, stock up a little. And around half the time I pick up a smaller number of items. Now imagine you work for a juice company. You’d survey shoppers and you’d see 50% of shoppers buying juice, and 50% of shoppers not buying juice. Your report from your agency might read “50% of shoppers don’t buy juice”. But that is simply not true. 100% of shoppers buy juice: they just only buy it on 50% of shopping trips. 

The shoppers in a survey are merely a sample. They represent a larger group of people. The 55% of shoppers in the sample who did not visit the category do not necessarily represent a different group of shoppers. They represent a different type of shopping trip.

Whilst in our sample clearly they are different shoppers, that does not mean that the population they represent is necessarily different. It could be the same shoppers, just shopping in a different way.

So why is this important? Am I just playing with semantics here? Absolutely not. If genuinely only 50% of shoppers buy juice ever then we need to understand why they don’t buy juice at all. Is there a consumption barrier? Is there a reason they really don’t like juice? Or is juice far too expensive?  But if 100% of shoppers buy juice, but only 50% of the time, then the reasons for not buying might be very different. It might be that I already have stock at home; or perhaps on this trip I am picking up a few items on the way home from work, I’m on foot, and I simply can’t carry bit cartons of juice? Different barriers would imply that different solutions are required, and that leads to more effective shopper marketing activities.

For the shopper marketers out there who hadn’t thought like this, I urge you to go back and review your data and your conclusions: rethink the data as shopping trips rather than shoppers. What if these trips were actually the same people? Use the demographic data to see if there is a good match in terms of the profile of the shoppers. Consider using some qualitative research to understand how shopping patterns change over time, to see if people will be in different modes on different trips.

For those shopper research professionals who know this, one simple thought: Never assume your audience does!


Written by Mike Anthony

November 29, 2012 at 9:04 am

Posted in shopper insights

16 Responses

Subscribe to comments with RSS.

  1. Hi Mike,
    I think you have made a great point here and honestly I think you may underestimate the number of people who need to heed your message. For every seasoned ‘Shopper’ professional there must be tens that come into contact with shopper data. It’s a lot easier for them to accept assumptions than it is to identify the truth behind the story. We do it every day when watching mainstream media. A deep dive often costs you time you may not necessarily have.

    The best way forward might be to be aware of the assumptions but nevertheless live with them.




    November 29, 2012 at 2:46 pm

    • HI Kenan,

      Thanks and maybe you’re right – most people I come across struggle to read data properly and take the charts they are presented with at face value – but then I’m talking about a relatively small sample so I didn’t want to be accused of teaching granny to suck eggs.

      As you say, awareness is the first step – to be aware of the assumptions and that there may be danger there..



      Mike Anthony

      November 29, 2012 at 3:29 pm

  2. Mike, your watch out here could be extended to a wide range of research issues. i see discussions in Linked in groups all about how researchers need to “cut to the chase” and present the ” short story” or the just the “headlines”. And indeed clients tell me how little time they have to “get to grips” with all the research they have.

    This places enormous responsibility on the researcher to do the spade work and the thinking correctly (often without getting a second opinion or check…). Reality is with any research the first requirement is to understand exactly what question has been asked, how and of whom. Without that clarity any research ‘take out’ can be risky.

    Being old school, I like to see (indeed insist that) any research debrief start with a proper methodology explanation (complete with an honest appraisal of its limitations and cautions – of course no research is without these) plus always to show the exact wording of the question asked, and the sample size on any quant chart. All too often I see material being used that fails all three of these basic requirements! And don’t get me started about levels of understanding of statistical significance!

    roger Jackson

    November 30, 2012 at 7:42 pm

    • Hi Roger,
      I am sure you are right – that this type of thing happens in other research areas – this one struck me as particularly prevalent in the shopper area, so I chose to highlight this one.

      I agree with your analysis too and it seems a little sad that so much money is invested by clients yet so little time is taken to understand the results.

      At the same time research companies are under pressure to cut costs, and (particularly the large ones) to standardize everything to make it replicable. All trends which are likely to increase the chances of mistakes like this happen.

      All hail then those that take the time and glean real meaning and accurate interpretation and therefore much better value out of research.

      Thanks for reading and contributing.


      Mike Anthony

      December 1, 2012 at 7:52 am

  3. Hi Mike – I’m very new to the topic of shoppers insights so forgive me please if this is a basic question. I was following along with your overall point that shoppers are not equivalent to shopping trips, and shopping trips are really the relevant level of analysis. But then I came to the statement below and was confused.

    “Your report from your agency might read “50% of shoppers don’t buy juice”. But that is simply not true. 100% of shoppers buy juice: they just only buy it on 50% of shopping trips.”

    Seeing 50% of shoppers buy juice leads to a conclusion that juice is purchased on 50% of the shopping trips, but how do you get to the statement that 100% of shoppers buy juice? I understand that the % of shoppers isn’t necessarily going to be 50%, but is it really 100%? (The statement struck me because I personally don’t buy juice.)



    Anne Robison

    December 1, 2012 at 5:46 am

    • Hi Anne,
      Thanks for reading and asking the question.

      The point I am trying to make is that in our sample 50% of shoppers bought. But the sample is a representative of shopping trips, not shoppers. Those same shoppers may buy on a different day or trip.

      Unfortunately many presentations do not make this clear and the interpretation is often to consider that 50% of shoppers don’t buy ever (and by implication that 50% buy all the time). This is not true.
      As tithe point on 100% I was attempting to simplify the situation by merely considering the shopping behavior of one individual (me) who buys juice on 50% of trips. In this case (with a sample of one) 100% of shoppers but juice, but on 50% of trips.

      I hope this helps! Let me know if I can help further.
      Thanks again


      Mike Anthony

      December 1, 2012 at 8:01 am

  4. Agree Mike, an holistic picture is required. We like to look at Homescan – type data (Household penetration, IPI, Frequency, AWOP etc) as big picture context and then when we do instore quant (observations, intercept interviews etc) it’s in that context and the traffic/browse/buy conversions are reviewed both in total and split by trip type. What I would say, irrespective of trip type, is that if if your traffic-to-browse ratio for a category is way higher than your browse-to-buy ratio then you have a likely conversion issue. And if your browse-to-buy ratio is higher than your traffic-to-browse ratio you may have a destination category that is only shopped by a small number of shoppers. If you look at traffic/browse/buy ratios in the context of household penetration and frequency it helps understand what the issue/opportunity might be.


    December 2, 2012 at 5:48 am

    • Thanks Norelle for your useful additions to the debate. Your points about taking an holistic view are completely correct (a great reason for doing effective desk research before starting any new fieldwork, and to review all existing data). Unfortunately Homescan isn’t always available, and even if it is, I still see people taking the wrong assumptions from the data.
      Likewise, your points on ratios are also really useful – the client had already done this work when we got involved. Unfortunately this was what led them to a dangerous conclusion because they assumed those that browsed were different to those that didn’t. Consider that they might actually be the same person and at that point the reason for not browsing/visiting/converting might become much clearer.

      Mike Anthony

      December 5, 2012 at 2:50 pm

  5. Hi Norrelle/Anne,

    I think both of you and Mike are saying the same thing. What Mike has highlighted is a very common data reporting/labeling error that can lead to wrong understanding/interpretation of the business issue and suggested insights/solution. It is pretty straight forward 10 shopping trips can be made by one shopper or 10 different shoppers.

    Anne “Seeing 50% of shoppers buy juice leads to a conclusion that juice is purchased on 50% of the shopping trips” is basically a wrong conclusion, in fact in this example 50% of the shoppers(shopper base) can account for 100% of shopping trips at the retailer. In other words all the juice purchase shopping trips can be generated by 50% or more of a retailer’s shopper base. In fact for categories having mass penetration level like coffee (assuming it has a 100% HH penetration) this could go up to 100% of a retailer’s shopper base. One of the correct ways of evaluating a category’s performance at a retailer is to look at category’s penetration levels among category shoppers at the retailer i.e. what % of category buyers are generating the category purchase trips at the retailer(conversion). Therefore, the goal of a retailer should be the moment a category shopper walks in to a store and is in the shopping mode for the category he/she should be converted. The point to note over here is even though I might be a category X buyer but on a specific shopping trip I might not be in the shopping mode for that category. This is the main disconnect between “Purchase” and “trip data” and many people do not understand it. For example many people at the manufacturer side are more exposed to reports and analysis based on purchase data for example looking at the measure:

    Buying rate = Purchase size(volume or $/occasion) X Purchase frequency(occasions/buyer).

    Please note here “occasions” refers to “Purchase occasions” meaning how much volume/size or $ amount was bought on a “purchase occasion” and how many “Purchase occasions” were there per category buyer. What many people do not see in reports generated from purchase data alone is the trip data meaning how many total shopping trips this shopper makes to the specific retailer in the specified time frame. Looking at the trip data alone you cannot conclude anything about products HH penetration issues. The combination of purchase and trip data permits viewing at a retailer and category level and provides a better assessment of the issue.

    Based on my experience with Nielsen/IRI HH panel data(more so now as both source the data from the same HH panel company called National Consumer Panel-NCP) there are three types of data reported by this Panel. Household Data – Describes panelist’s demographic and geographic characteristics. Purchase Data – Describes purchases by panelists of HH items both UPC and non-UPC(fresh produce etc). Trip data – Describes where HH panelists shop for e.g. grocery store or discount store etc and includes information such as trip date, total amount spent and retailer name etc by trips.

    Expanding on what Roger has said I believe it totally depends on the business issue at hand, what question has been asked, how and by whom. What I have learned from my experience is “Crystallization of business issue” is extremely important. Different functions in the organization face different issues and same issue can be looked by different functions in the organization in a different manner. It is the interactive discussion about the issue that really helps explain what the person is trying to address and what type of data/data sources can address the issue.

    Norrelle, in the end I agree with you 300% that in order to have complete understand of the business issue a holistic picture is required that can help us in delivering meaningful insights for example as follows in a very simplistic manner without going into the details of trip types and Heavy, medium & light buyers:

    Category X has HH penetration of 80%. Retailer Y presents the best opportunity to target the shoppers of category X as it attracts 60% of category buyers (Retailers penetration among category buyers). Retailer Y is also most successful in category buyer’s conversion at 80% (80% of 60%) and trip leadership at 70% (capturing 70% of total category purchase trips).

    Most of the above also applies to retailer based loyalty card data as well, whether from select retailers or from a loyalty card-based shopper panels which recently Mike M referred to in his post “Five Benefits of Using Loyalty Card Data To Enhance New Product Launches”.

    Last but not the least, Norrelle I hope you are not referring to HH panel data in your traffic-to-browse or browse to buy ratios as I have some reservations in this regard which I will try to discuss in another post.


    December 3, 2012 at 2:01 am

    • Dear Humayun,

      Thanks so much for adding so much value.

      Your points are well made and the point about the business issue especially so. Too often we find clients attempting to answer the “wrong” question with the data – or trying to use an incomplete data source to answer a question it can’t possibly answer. Specifically we see a lot of channel specific data (be it research or layalty data) being used to understand a shopper in total – without recognizing that those shoppers nearly always use other channels and retailers.

      Thanks as always,


      Mike Anthony

      December 5, 2012 at 2:53 pm

    • Great post, Humayun. What do you see as the 3 biggest weaknesses of the National Consumer Panel data in support of your work?


      December 11, 2012 at 2:34 am

  6. great points from others – thanks Humayun for a detailed discussion. Norelle, I wonder about browser data as in our eye tracking studies, many people ‘browse’ a category from a distances some up to 5 metres away thus a ‘visual browse’ and then may either do a ‘physical browse’ or move on. Trip data (%buyers who generate trips) is an interesting one but no one has mentioned unplanned versus planned – I only ask because this concept is becoming important to some of our US clients and I wonder if it is not considered important enough generally speaking – maybe I misread the posts.


    December 5, 2012 at 11:07 am

  7. Hi Norelle/Jared/Shane,

    Thanks for your comments. I am assuming that you are referring to weaknesses of longitudinal HH consumer panel data as national consumer panel data is simply the joint venture of Nielsen and IRI to form a new company where they have merged their respective panels to create a one big pool and enhanced coverage of panelists. My experience of using HH panel data (both Nielsen/IRI) is prior to this joint venture. Well, household panel data despite being a great resource for insight generation and informed business decision making sure does have short comings. However, I will share my point of view for both types of short comings one which are most common and most quoted and others as well. The three common ones are:

    a) Difficulty in recruiting certain households – Minorities, low-income hhlds, mobile singles
    b) Panel represents household-based purchases – No small businesses or civic organizations
    c) Not all purchases are recorded by panelists – Immediate consumables, consumed elsewhere(Parks, movie theaters Friends place and on the go etc.).

    Please note I read sometime back that either Nielsen or IRI has tried to address this issue by providing a new type of a scanner to the panelists which they called as “U-scan”. Well, with the help of technology the future is promising for e.g. iphone, some smart shopping carts that automatically captures purchases can integrate shoppers loyalty card details and above all the advancement in the RFIDs

    Other issues are:
    1)Impact of dealing – Panel dealing & pricing includes impact of “perceived” deals 2)HH panel data is not always going to be equal to store data due to – store data records millions of transactions whereas HH panel data records few thousand of transactions by panelists, 3)panel data includes non-cooperating stores and simply due to a different sample design for store vs HH panel data. I will not go into the details of SPECTRA store-trade-areas which is a whole different ball game.

    Purchases made through the vending machines, through internet and direct mail etc are also not captured through HH panel data.

    Last but not the least under reporting/low reporting of personal care categories by panelists.
    I do not consider regular vs irregular panelists an issue as it is taken care through “static”.

    Well, so far this is pretty simple 101 stuff. In my view there are bigger structural issues at hand. I think Mike has done a great job in highlighting one issue for which I have provided the details that is the differentiation b/w “purchase” and “trip” data. This is a very big differentiation and has huge implications in understanding, addressing and providing solution to the business issue. I am not sure what type of “browse” data was referred to in the ratios traffic/browse and browse/shop. I hope it is not panel data because HH panel data cannot provide in-store “browse” data plain and simple period. It does not have the capability to report as to how much time a shopper spent in the aisle and in front of the shelf. Please let me know what data sources you guys have used to develop these ratios and if in-store video observation(video minning or shopper gauge) data was used for these ratios then how it was integrated with HH panel data. I also have issue with the word “traffic” as it is very misleading and has different meanings for both manufacturers and retailers. In my view retailers consider “traffic” as “footfall” both terms referring to trip data whereas most of the manufacturers understanding of traffic are “shoppers”.

    Please note HH panel data captures purchase behavior that is simply what is being put in the basket and checked out and not motivations behind the purchase, dwell time in the store/in the aisle/at the shelf.

    Another structural issue – We all are very familiar with loyalty(through panel data-share of requirements). Let us consider two brands in a category one a renowned global brand and the other a private label. Please note that I am not referring to some upscale private labels like being offered by Target or Kroger. For example if we see share of requirements of +70% for both the brands then what are we going to conclude for private label are its shoppers loyal to price or brand? This becomes even more complicated for some of the upscale private labels that have recently gained traction during and post-recession period.

    Planned and unplanned are also an issue. This can only be captured if panelists are prompted for this at the time of scanning their purchases at home or through qualifying the trip either planned/unplanned through a customized survey. I know Nielsen had a service called “panel views” which is a customized survey executed through selected panelist to understand the “why behind the buy” or to understand the purchase behavior in detail. Thanks and Regards


    December 14, 2012 at 9:42 pm

  8. Hello Mike, very interesting article! We faced a similar challenge for a client and solved almost every question, except the correlation with the trip to the shop: http://www.websignage.eu/en/popchannel_en.pdf
    This will be the next step of our research projects and we are on our way to integrate a check in mechanism via smartphone. We are looking for guinea pigs to nail this down!


    December 18, 2012 at 2:35 am

    • Hi Blochin,
      Thanks for sharing – do come back when you have completed the research and share!

      Mike Anthony

      December 18, 2012 at 1:49 pm

  9. I will! If someone wants to participate to such a research project, we would be glad to offer the technology for free for a pilot project. Our project is involving universities and research centers, therefore there are enough mad scientists to support the craziest solutions!


    December 18, 2012 at 4:33 pm

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: