Capturing “Up for Grabs Shoppers” is an Important Key to Downtown Retail Success

By N. David Milder

Who Are They?

Many downtown retail growth strategies are doomed because they try to avoid some key facts. One is that, except in the very rarest of rare situations, downtown retailers, be they new or old, large or small, must compete for and win sufficient market share to prosper. Another, and closely related fact,  is that beneath the venerated “leaked” sales to merchants located beyond the downtown’s trade area, and the 45% of GAFO sales now being e-leaked to online merchants, is a group of shoppers who are either weakly bonded or completely unbonded to merchants in either the downtown or its larger trade area. They are “up for grabs shoppers” who are very likely to buy fewer things, or to be won over by strongly magnetic brick and mortar merchants located beyond the trade area, or by online merchants, or—and this is very important – by new retailers opening in the downtown or elsewhere in the trade area.

Some Implications

The existence of such shoppers has important implications:

  • The up for grabs shoppers are always there, though their numbers may vary across retail sectors and over time.
  • For new and expanding downtown retailers, it means that there very often will be between 15% to 60% of the shoppers in their retail sector who are up for grabs and likely to give them a look. That indicates the local competition is weak.  If the new/expanding retailers are capable, they will have a very good chance of winning the dollars and loyalties of these shoppers.
  • For many existing retailers, the up for grabs shoppers can indicate – if they learn about them — that a good percentage of their customer base may be prone to desertion and signal a need for the merchants to improve their operations.
  • For downtown economic strategists and leaders, it means that any successful new retailer brought into town is likely to win customers away from merchants located beyond the trade area, or from online merchants, and/or from brick and mortar merchants currently located in the downtown or elsewhere in the trade area. The existence of substantial numbers of up for grabs shoppers also is a sign that downtown EDOs need to create effective programs to help existing merchants improve, or to be prepared to recruit more capable merchants who can better satisfy consumer needs and wants.
  • Just looking at the shoppers leaking their retail expenditures to beyond the trade area merchants is rather myopic – and a denial of reality. This myopia is understandable given that it seems to allow for the ill-conceived assumption of immaculate retailing that any new or expanding downtown retailer competing for the leaked dollars will not take any sales away from other downtown merchants. The existence of any sizeable number of up-for-grabs shoppers in the relevant retail sector means that is a highly unlikely prospect.

Some Examples

DANTH, Inc. first addressed up-for-grabs shoppers in a number of telephone surveys we did back in the 1990s when we asked respondents  whether various types of retail stores they could visit within  a 20-minute drive from their homes, were excellent, good, fair, or bad. Responses of fair and bad were treated as indicators of weak bonding with the relevant retailers. Their retail expenditures consequently may be considered as up-for- grabs and more prone to being captured by new or expanding retailers, be they brick and mortar or online.  Above are two tables showing the responses to surveys done of the shopperss in the trade areas of Rutland, VT, and Carlisle, PA.  For example, about 44% of the expenditures for suits or dresses by shoppers in Rutland’s trade area were up-for-grabs, as were about 43% of those expenditures by shoppers in Carlisle’s trade area.

For all the retail store types, the average number of loosely bonded shoppers in Carlisle’s trade area, 27.3%, was somewhat lower than that in Rutland’s trade area, 33.7% — see the table above. This may be because Carlisle is in a denser region, with higher household incomes, and with many more retail choices. Downtown Rutland is located in the Rutland Micropolitan Statistical Area that is composed of Rutland County. The median household income in 2017 in the county was about $52,000, and about 19% of the households had an annual income of $100,000+.  The county has a population of about 61,000, and Rutland City is by far its largest retail center. In contrast, Downtown Carlisle is on the western edge of the  Harrisburg–Carlisle MSA that had a population of about 560,000. Carlisle is located in Cumberland County where the median household income in 2017 was over $82,000, and about 27% of the households had an annual income of $100,000+. Moreover, back in 1997, in the downtown Carlisle trade area there were 12 major malls occupying a total GLA of about 3.5 million SF.

It is also interesting to note that, even with all that retail within an easy drive, on average, 27.3% of the shoppers in Carlisle’s trade area were up-for-grabs. Moreover, that number was even higher for some important markets segments: shoppers with children and those with annual household incomes over $50,000 (about $80,000 in 2019 dollars). The same pattern among market segments was even stronger among Rutland’s shoppers.

Some Types of Up-for-Grabs Shoppers

Up for grab shoppers can be present in many market segments and to varying degrees. For example, the numbers/percentages of loosely bonded shoppers in the upper income 4th and 5th quintiles are of particular interest because they account for a very disproportionate amount of consumer expenditures across all sectors, especially retail. As can be seen in the above table, nationally, shoppers in the highest income quintile (the 5th 20% group) accounted for about 38.9% of all consumer expenditures in 2017, about equal to the combined total of the 3rd and 4th quintiles. The 5th quintiles shares of all expenditures on food away from home, home furnishings, and apparel were at about that level. However, they also accounted for 52% of all entertainment fees and admissions, making them an absolutely critical market segment for most downtown entertainment niches. 

In rural towns and cities, such as Rutland, VT, Scotts Bluff, NE and Laramie, WY, where trade area populations are not large and household incomes are relatively modest, one might expect the more affluent shoppers will be among those most detached from local merchants. These downtowns usually do not have a strongly varied retail environment and local merchants are prone to catering to the more numerous middle income shoppers. Underserved, and possibly ignored, these more affluent consumers tend to shop in distant towns and cities having more robust retail assets, and they are increasingly buying from online retailers. 

Very often, a large proportion of leaked retail expenditures come from the 20% to 30%  of the households with the highest incomes in the trade area. Unless a sufficient bolus of the types of retail they prefer open in the downtown or trade area, it will be very difficult to recapture those leaked dollars. Traditional leakage analyses, by themselves, cannot identify such situations. However, an analysis of the up for grabs shoppers can help  answer the critical question that  leakage analyses raise, but cannot answer: how many of the leaked dollars can be captured by new or improved local merchants?

Lower income shoppers also can be up for grabs. The local retail structure also may not have the stores with the price points and/or merchandise they need. Evidence of this comes from the enormous growth in recent years of dollar store chains and their ability to take significant numbers of low-income shoppers away from huge, well-established retailers such as Walmart, as well as from local small merchants. 

It should be noted that an important element in the discussions of upper and lower income shoppers presented above is the existence of what might be termed a gap between the types of stores these shoppers need and/or want and those that exist in the downtown or trade area. A useful estimate of the monetary values of such gaps can be made by multiplying the number of dissatisfied shoppers by sector in the relevant income groups with estimates of the retail expenditures by sector of households in those income groups. However, such estimates do not carry along with them the assumption that all of the potential gap expenditures are being leaked to beyond the trade area merchants. Shoppers might also spend online, or simply reduce their spending levels.

The discussions of these two income groups also helps spotlight a frequent deficiency in downtown market analyses: the primary focus on statistical means and medians.

Millennials, now our largest generation, seems very prone to being weakly bonded to product brands. One might reasonably hypothesize that also will probably be the case for retailer brands.  For example, in 2017, a study found that “67 % of millennials changed brands in the last year” and called this “a clear lack of brand loyalty among 18-34 year olds.” The two major factors driving disloyalty were product quality (49%) and product availability (44%).  These findings suggest that the number of up for grabs shoppers is likely to grow in importance in coming years as the economic importance of the millennials grows. See: “Millennial Research: Factors Driving US Millennials Brand Disloyalty”, Posted on January 20, 2017 by B. Smith to

Here’s the Rub

In my experience, telephone surveys with about 500 to 600 respondents were the best way to obtain useful and reliable data about the up for grabs shoppers in a downtown’s trade area. However, over the past two decades, it has become harder and harder to conduct such surveys. Response rates have dropped significantly as the public became more resistant to answering surveys and responding to telemarketing efforts. Online surveys are not a substitute, since their use really requires a panel of respondents from which a valid sample of trade area respondents can be drawn. Few, if any, trade areas have such panels.

As a result, for many years we stopped doing trade area telephone surveys, yet the need for the types of data they could provide seemed to grow with the upheavals in the retail industry and the need to get a good grip on how many sales were going to online retailers. Today, in the face of that growing need, the best available solution path appears to be one framed by an analytical modesty that recognizes we will have to deal with survey data that is far less accurate than we might like. For example, it may be necessary to accept a 5%  or 10% estimate error at the 85%  or 90% confidence level. These can be maximized when the population being surveyed can be treated as finite.  Furthermore, the solution path might utilize several of these research tools:

  1. Shopper Intercept Survey. The value of these surveys depends a lot on where and when the interceptions are made and the number of interviews that are completed. The more completions the better. That number will be determined by where the interceptions are made, the length of the questionnaire, the ease of answering the questions, and the respondents interest in revitalizing/improving the downtown. Given the need for brevity –- say 10 minutes to complete the questionnaire – it will be essential to carefully select the most important questions. In the past, we limited our use of shopper intercept surveys because they seemed limited in their ability to gather all the information that a telephone survey could. Furthermore, they could not reach the trade area shoppers who did not shop downtown and obtain information from them that might help explain why. That said,  the need to get some useful data about these up for grabs shoppers has grown to the point that we are faced with the choice of either rejecting the use of any survey data or using surveys that may not have the error and confidence levels held as the acceptable standards in the past. One can argue, that if the conclusions drawn from a survey with a 7% or 10% error factor at an 80% or  90% confidence level are carefully structured, they still can be very useful analytically. The analyst is certainly in a better situation having access to such information than not having it.   
  2. Online Surveys. In a number of instances, some market segments may be known to be more important than others and merit special attention. The size of such a market segment and viable ways of contacting its members also may be known. That means that huge proportions of the relevant population, possibly even every member, can be invited to participate in an online survey. In these situations sampling is either not an issue or not a significant one. This is often very true of important segments in a downtown’s daytime population: people employed in the downtown, seniors in downtown housing and senior centers, high school students, patrons of downtown cultural venues, users of downtown transportation centers, downtown residents, etc.  
  3. Nominal Group Process (NGP).  We like this small group process because its structure prevents the discussion being dominated by a few participants and assures a useful information product will be produced at the end of the session. The NGP is able to handle 100 to 150 participants grouped in 10 to 12 tables and then the  results often can be stated in quantitative terms. However, the qualitative inputs generated by participants are usually the primary useful products.
  4. Focus Groups.  These small groups can be useful, but too often are not. They best provide qualitative information, Using them to predict market segment behaviors is ill founded, since the number of participants is usually too small to constitute a useful sample and their characteristics and recruitment are unlikely to be representative of the relevant population. If not well-led and/or are too large , focus groups can be dominated by a few individuals. However, the qualitative information they often can produce can give the analyst an understanding that simply cannot be provided by just the numerical data. They can be invaluable for generating viable explanatory hypotheses.       


GAFO E-Sales

In my retail recruitment experience, I’ve found that there are types of retail stores that clients need and those that they want. The need category generally includes groceries, specialty food shops, pharmacies, etc., while the want category overwhelmingly includes GAFO operations — i.e.,  general merchandise, clothing and footwear, home furnishings, electronics and appliances, sporting goods, book and music stores, and office supply stores. The shops that respond to needs did relatively well through and after the Great Recession, while the GAFO stores have been in consistent decline or weakness since about 2009. Recent research indicates that e-GAFO retailers are now eating the lunch of brick and mortar GAFO merchants.  

An Enormous 45% Hit on B&M Retail Sales Potentials!. One of the most significant trends that has helped define the new normals for retailing and our downtowns is the increasingly significant share of the sales of the merchandise sold in GAFO stores that are being captured by online operations. Obviously, the more sales dollars the e-stores win, the less there are for brick and mortar shops (B&Ms) to capture.

A while back, in another blog posting, I presented the above table, taken from a provocative  study by Hortacsu and Syverson,  that showed  e-store market penetration for a range of retail  categories in 2013 along with estimates of the years in which they each would reach 25%, 50%, 75% and 90% market shares.

A more recent 2019 report by Morgan Stanley suggests that the Hortacsu and Syverson study was pretty sound. It found that while “…e-commerce penetration reached 11% of total retail sales at the end of 2018”  that “e-commerce penetration in the GAFO segment”  was now over 45%.(1) That makes it so much harder for B&M GAFO retailers to survive, much less thrive, unless they are executing or part of an omni-channel marketing strategy.

The Morgan Stanley report also found that “the shift to e-commerce has hit the home-furnishings segment the hardest,” while clothing, linens and other “soft” goods have experienced a significant “e-commerce disintermediation” with a 22% e-commerce penetration expected in 2019. (2)  It was long thought that these two retail segments would be resistant to e-store penetration because one offers large and heavy merchandise and the other offers merchandise that consumers would want to touch, feel and try on. One weakness of such thinking was the failure to recognize that so many of the soft goods we buy are like commodities and we don’t need to touch them, feel them or try them on. For example, lots of people have long bought shirts, trousers, shoes, dresses, swimsuits, parkas from catalogs. They often are buying more garments like the ones they already have – e.g., I have countless blue, button down collar shirts — or replacements for them. Then, too, lots of home furnishings products are not furniture suites or otherwise prohibitively large, while others have been re-imagined – e.g., Casper Mattresses – so they can be shipped “small.” 

How Are the Leakage Analysis Data Providers Dealing With This? Frankly, I do not know the answer to this, but I think the data providers owe their customers a clear explanation of how they are handling this situation. One technique they might be using for estimating consumer demand is to take the sales of retail stores by NAICS code within a certain fairly large geographic area and then divide the sales by the number of households in that study area. That defines demand solely in terms of B&M store sales, ignoring the huge Internet sales and demand. If, instead, they are using extrapolations from BLS consumer expenditure surveys to determine demand, then they must have whopping “leakages” in each of the NAICS codes analyzed unless they also are using data on e-store sales by NAICS code.

The leakages to the Internet for GAFO store merchandise now are probably several magnitudes larger than traditionally defined leakages to B&M shops located beyond the trade area’s boundaries.

Of course, an increasing number of downtown merchants now have both a B&M shop and an e-store. Most of their e-store revenues often come from distant customers and represent “e-surplus” sales. How are these e-sales revenues included in the leakage analysis? How do leakage analysts know which e-sales come from within the B&M store’s traditional trade area from those that come from beyond it?

A growing number of retail sales are “click and collect” transactions that involve ordering online via a retailer’s server that probably is located hundreds of miles away and then picking up  the merchandise at the retailer’s local store. Are those transactions to be deemed leaked or “unleaked” sales? The local store’s involvement may be key to the sales transaction, though it may not logically be part of the monetary transaction. Would the sale have occurred if the local store were not there? If the answer is no, then somehow the role of the local shop has to be recognized in the analysis.

Vacancies, Store Closings and Openings, Changing Functions

A Word or Two About Vacancies. I fear that I’m very much an outlier, a contrarian, when it comes to downtown vacancies. While I don’t like vacant storefronts, my jockeys don’t always get in an uproar when I see them. Too often, they are not viewed from the proper perspective. Rule 1 for looking at vacancies should be to ask: where is the downtown on its revitalization arc? If it’s in the initial very troubled stages, then the prospects for recruiting really good retail tenants are not great, especially with today’s upheavals in the retail industry. Moreover, recruiting crappy tenants would be worse for the downtown’s revitalization effort than the empty shops. Also, at these early points in the revitalization process, an EDO’s scarce resources are probably better spent on working for improving the infrastructure and housing and reducing quality life issues such as the fear of crime,  than paying for very problematic efforts to recruit good retail tenants.

Rule 2 is don’t be snooty — look at pamper niche tenant prospects such as hair and nail salons, yoga and martial arts studios, etc., especially early in the revitalization process when their relatively low revenue needs and desire for low cost spaces can put them among the downtown’s best tenant prospects.

I take vacancies more seriously when the downtown is much further along on its revitalization arc. In these situations, Rule 3 is the locations of the vacancies are far more important than their number. Those that are in strategic locations such as on or near the district’s “100% corner” or near other strong assets will certainly need attention. A cluster of them is also significant and probably indicates the existence of an important underlying problem.

Rule 4 is that the downtown EDO should identify and address such underlying problems, otherwise any “fill the vacancies” recruitment program undertaken either by it or local commercial brokers will most likely yield paltry results.

In the mid-arc downtowns, Rule 5 is to determine if new downtown projects have raised landlord expectations about:

  • Their ability to attract national chains, even though they are looking for fewer and smaller spaces and have become much more finicky about their new locations.
  • Potential rental incomes to the point that their spaces are too pricey for their most likely tenant prospects, small independent merchants.

If either of the above is the case, then there’s a landlord problem, not a tenant prospect problem. This leads into Rule 6: as downtowns revitalize, erroneous landlord estimates of viable rent increases can result in more vacant spaces than diminished consumer retail demand or its associated reduced retailer demand for store spaces.

In the past, I argued that a vacancy rate of about 5% was the sweet spot for mid-arc downtowns. Some vacancies are necessary to allow for the tenant churn that can bring in new merchant blood and help keep the district vital. That still strikes me as an ideal goal. Many years ago, my real estate mentors taught me that vacancy rates above 10% indicated the existence of serious downtown problems that needed immediate identification and remediation. Well, these days, under the New Normal, it seems that a 10% vacancy rate is about average for retail spaces (3). Of course, I am not clear whether that statistic refers to all the spaces in shopping centers and malls or just to those allocated for retail tenants. Given that so many malls and shopping centers have saved themselves by bringing in non-retail tenants, I would say it probably is the former. One disturbing implication for downtowns is that, these days, a 10% storefront vacancy rate may not be all that bad, comparatively speaking. Even more unsettling for me have been the reports I’ve seen of downtown vacancy rates in the 10% to 20% range in some of our small and medium sized communities,  Another implication is that downtowns must look more to nonretail tenant prospects to fill their vacancies, but ones that are able to stimulate and reinforce pedestrian traffic on nearby sidewalks.

Because of Omni-Channel Marketing, B&M Retail is Not Going Away. One might expect that if the addressable retail markets for B&M chain stores have shrunk substantially, that lots of the stores would be closed. In fact, there have been a huge number that were closed –e.g., 7,000 just in 2017.  However, new shops are also opening and an accelerating number of them are by Internet-birthed retailers (4). For example, so far in 2019, there have been 1,674 retail chain store losings, but 1,380 store openings (5).

Today, successful retailers do not see B&M store customers as a different set from their e-store shoppers. Instead, they just see customers who they can individually reach through several channels, e.g., B&M shops, websites, social media, traditional media, etc. They know that while most consumers may still prefer shopping in B&M stores over e-stores: (6)

  • Convenience is an important driver of which shopping channel the consumer will select
  • Unless the B&M store provides an attractive shopping experience, it will not attract as many customers as its management might want.

B&M retail shops, under an omni-channel marketing strategy can play a number of functions, besides being a place where sales transactions occur, that can justify their existence:

  • SONY and Samsung, for example, have had important store locations that are nothing more than showrooms. Many other retailers use their shops as places where customers can experience the use of their merchandise. You can, for example, book a nap at a Casper Mattress Sleep Shop.
  • More and more large retailers are offering “click and collect” purchasing, e.g., Best Buy, Walmart, Amazon.
  • Some retailers are developing special store formats, e.g., Nordstrom Local, where they can provide extremely high levels of customer service to shoppers with a proven record of spending large sums in their stores.
  • Almost universally, the B&M store is seen as the venue where the retailer can best provide experiences that will strengthen their relationships with customers.
  • B&M stores also can generate website traffic. For retail chains, a new B&M store in a market area sparks “a 37 percent increase in overall traffic to that retailer’s website” by area residents. (7) “For emerging brands, new store openings drive an average 45 percent increase in web traffic following a store opening, according to ICSC research” (8).  Of course, web traffic does not mean web sales (see below).

Very importantly, B&M stores outperform e-stores in several very critical ways:

  • They have a much higher sales conversion rates (visitors who turn into actual buyers), averaging about 22.5% across all retail sectors, than the less that 3% for e-stores (9).
  • Merchandise return rates for e-stores are three to four times higher than for B&M stores, probably because e-shoppers cannot touch, feel, try on or otherwise experience the merchandise. Returns have become an enormous ball and chain on e-retailer profitability, while bad returns experiences are really ticking off e-shoppers (10).

Bottom Line: B&M retail stores are not going away, but there will be far fewer of them, they will occupy smaller spaces, and perform many new functions that justify their existence besides making sales transactions. How is your downtown planning on dealing with such a scenario?



2) ibid.

3) closures-are-expected-2019

4) Ibid.

5) ibid.

6)  . Pew surveys have had similar findings.


8) Ibid.

9) See:



The Use of the Muddled Immaculate Retailer Concept in Leakage Analyses

By N. David Milder

Looking at retail leakage studies, I am reminded of Coleridge’s famous line: “Water, water, every where, Nor any drop to drink”. Retail leakage studies seem to be de rigueur in the downtown economic development field, but a good one, devoid of fatal errors is hard to find. As I have detailed in earlier posts, leakage analyses have serious analytical and data issues. I want to return to one of these analytical problems because I think our field lacks appropriate  awareness of the muddled conceptual thinking  that too often is being used to make a lot of important program, policy and investment decisions. I call this muddle the concept of the immaculate retailer.

This concept runs along these lines:

  • A leakage is said to exist when the retail expenditures of a trade area’s residents exceeds the sales of trade area retailers. Those dollars that are uncaptured by trade area retailers are said to be captured by retailers located outside of the trade area. So far, all this is analytically simple, well and good.
  • A next and troublesome step is to assert that one or more new retail firms can locate downtown, and their sales will be based on capturing these leaked sales. Consequently, they will not take sales away from retailers already established in the downtown. The new retailers somehow can compete immaculately. Local retailers ostensibly have nothing to fear from new merchants entering their downtowns.

My understanding of how retail markets function and how retailers behave suggests that immaculate retailing is simply impossible. I have little doubt that some merchants, large and small, may appreciate locations where the competition is sparse and/or weak, though all but monopolies and oligopolies must fight for market share whether they realize it or not. How such retailers then would parse their sales to only capture those dollars that would go to distant rivals is never specified and frankly has proven to be beyond my analytical abilities to identify. I have no idea how a merchant could feasibly, in the real world, compete against retailers outside their trade area, but not with those already located in their downtown – unless it is on the Internet. I’m willing to learn, so if you know of such a path, please let me know.

Even in situations where there is no retailer of that type, e.g., a grocery store, the new entrant will likely have to compete with a really powerful retailer located outside of its trade area. The presence of that strong competitor is probably why, for example, there was no grocery store already there. There is an asymmetry in the trade areas among individual retailers and as well as those among retail centers that reflects their relative strengths. The stronger they are, the farther they can reach. As a result, a proper market analysis cannot just look at the competition within a trade area defined by where a store’s potential residential customers are located. That store’s being located in the trade areas of strong competitors must also be identified and assessed.

There is, however, another perplexing face to the immaculate retailer muddle: that somehow, it will be relatively easy, perhaps because of their greater proximity, for new downtown retailers to win back the leaked sales. There is often an unstated assumption that the competitors located outside the trade area are weak or will not compete, that the leaked funds are like lots of coins fallen on a carpet and just waiting to be picked up.  This shows itself most overtly when the question of how much of the leakage can be recaptured. Far too often the question is not overtly addressed, leaving the implicit false implication that all of it can be recaptured. When the question is addressed, some rule of thumb often is used. Most regularly 10% to 15% of the leakage is suggested as a conservative, reasonable  estimate, though no research supporting that suggestion is cited. The rule’s purported general acceptance is what lends it legitimacy.

To the contrary, I would argue that no general rule can be applied, because two crucial variables are not being properly taken into consideration:

  • The strength of the competitors. Too often the bulk of a leakage analysis’ focus is just on the downtown’s trade area, when its major competitors are located well beyond its borders. That is often because the location of these rivals was not taken into consideration when the trade area was defined. This results in the competitive strength of rival centers being poorly researched, and ill-considered in the analysis. By the way, I think most downtown retail market analyses do not pay sufficient attention to the competition. Indeed, most trade areas are defined by where consumers live, but techniques such as gravity models and considering the distances to competing centers need to be more frequently incorporated.
  • The second and most overlooked factor, is the ability and power of the new retailer(s) that would be brought into the downtown. I’ve seen one suggestion that 40% TO 60% of a retail leakage can be recaptured. Perhaps, by a major retail raptor like Home Depot or a major specialty chain, but in smaller communities, that’s probably unlikely for independents who would be happy as lark with annual sales of about $500,000 each, even if there are more than one of them. This, too, should be considered: by definition, half of all retailers are below average.

Who the downtown can recruit matters more than the size of any leakage. My enquiries to retail site selectors indicated that few, if any, use a leakage analysis to determine where they will locate their physical stores. The retailers you want for your downtown are prepared and able to compete. Among these able retail competitors will be chains and independent operators. In almost every downtown I’ve worked in I’ve found small operators who are very savvy merchants and very able competitors.

The identification of those retailers is what we should be focused on. A proper analysis of the downtown’s  various addressable market segments, that includes psychographics,  should indicate which types of retailers will find their types of customers in a that district and its trade area attractive. Those retailers should be targeted for recruitment.

Major retailers, because they can use their data on their stores’ sales and costs,  current customers and potential customers, can generate more reliable estimates about the potential sales revenues and operating costs at a new location and how much space they can afford to lease. The ability of a leakage analysis to address that question pales in comparison!

Leakage analyses have other analytical issues as well as some very severe data issues. The data issues could be resolved if the data providers would detail how their many needed manipulations of various types of primary data have been validated, demonstrating that they are truly measuring what they say they are measuring. For example,  BLS’s surveys of consumer expenditures are national and the data can be presented at the level of multi-state regions. But when a data firm produces estimates for a downtown’s much smaller radii or drive sheds, how do we know that the necessary manipulations of the data produced the correct results? Given  that such estimates can vary from firm to firm, how do we know which are the correct ones?

A retail leakage report from Esri or Claritas may be relatively easy and inexpensive to purchase. Nonetheless, one should not be misled by that fact — the correct analysis of those data will not be commensurately easy and cheap.

Retail Leakage Analyses Should Be Treated With Great Caution by Analysts and End Users: 2 Some Serious Data Issues

By N. David Milder


This is the second of a two-part article on the need to be very cautious when doing or using a leakage analysis. Part 1 focused on the analytical issues associated with this economic development research tool. It can be found at

This article will focus on the problematical data too often used in leakage analyses.

GIGO is an acronym well-worth knowing. It stands for Garbage In, Garbage Out. Too many leakage studies are very problematic because the data they use are very questionable. This is often not the result of data suppliers being slovenly in their data gathering or analysis – although sometimes, it is – but because getting information about small and medium sized firms’ sales, workforces and finances can be very difficult. To counter this problem, the leakage market data firms sometimes will try to use business census and related data to estimate the sales of the firms in the major NAICS codes that have been identified in a study area. Their resulting estimates are not based on primary research and have a number of degrees of separation from the original firm level data that is usually unknown to the data purchasers. Different research firms will have different methodologies for making these estimates. Which, if any of them, produce accurate estimates about NAICS retail categories’ sales is a key question. Unfortunately, I have not seen any studies that validate the accuracy of these estimates. This suggests caution and prudence in their use.

A Good Starting Point for the Discussion


The table above is a good starting point for this discussion. It shows the results of the leakage analyses generated by Nielsen and Esri for a city in the West that has a population of about 30,000. These two respected private firms provide a lot of the demographic and business data used in the economic development field and the table is representative of the leakage analysis data they produce. They certainly are not the only ones that generate the data used in a leakage analysis, but because of their affordable prices and easy online access, they are used by many EDOs (mostly by their consultants) looking for such information.

The table’s first column shows the NAICS retail categories being analyzed. Leakage analyses subtract retail sales in the study area from consumer expenditures. If the expenditures exceed sales, there is a leakage. Conversely, if sales are larger than consumer expenditures, there is a surplus. Columns three-five provide the data on consumer expenditures, store sales and leakages/surpluses for each category. The data in columns three and four are estimates – these firms have not directly surveyed the residents of City XYZ about their expenditures or the retail stores about their sales. That’s one reason why their leakage analysis reports can cost $50 and not $25,000. The data in column five are simply the results of subtracting the data in column four from those in column three.

Given that the numbers in column five are supposed to facilitate an understanding of economic growth potentials and possibly an assessment of the market viability of substantial private investments, looking at the table probably leaves the reader disturbingly confused. While there is agreement in direction (i.e., leakage or surplus) and quantitative closeness in some categories, e.g., gasoline stations, furniture and home furnishings, on many others the size of the leakage/surplus differs substantially, e.g., food and beverage stores, or conflicts on direction, e.g., health and personal care stores and eating & drinking places.

Looking at the estimated data in columns three and four, there is an even more divergent pattern. The estimates of both consumer spending and stores sales are in most cases substantially different. These two market data-producing firms are obviously using very different methodologies and/or formulas to generate these estimates.

How, then, do we know which, if any, of these leakage estimates can be trusted?

An Example of the Explanations They Provide

If you call these firms or look online for explanations of how they generate their estimates, they may use quite a few words, but seem nonetheless not very transparent. Here, for example, is how ESRI in “2014 Methodology Statement: Esri Data—Retail MarketPlace,” a 2014 white paper,” described how they estimated retail sales:

“Estimates of retail sales begin with the benchmark, the 2002 and 2007 CRT (Census of Retail Trade) from the US Census Bureau. Trends from the economic censuses are used to update the base along with Esri’s extensive portfolio of demographic and business databases. These include commercial and government sources such as the Dun & Bradstreet business database and economic statistics from the Bureau of Labor Statistics. Supply estimates also incorporate data from the Census Bureau’s Non-employer Statistics (NES) division. Smaller establishments without payrolls, such as self-employed individuals and unincorporated businesses, account for a small portion of overall sales. However, these businesses represent more than half of all retailers in the United States. Their inclusion completes the report of industry sales.

Esri’s model captures economic change by first differentiating employer and non-employer sales growth. Multivariate statistical techniques are used to model data that is subject to disclosure issues in CRT and NES. Sales estimates have been recalibrated against the Monthly Retail Trade (MRT) survey and independent sources to address the disparities that can exist between independent input data sources. This methodological improvement yields a more precise estimate of the retail sales attributable only to domestic households.

Esri licenses Dun & Bradstreet’s business database, which also provides sales for the retail market. Although Esri utilizes this database in the derivation of small-area estimates, the methods differ. Esri estimates retail sales only to households for implementation within the Retail MarketPlace data. Additionally, Esri relies heavily on data from both the CRT and MRT to improve estimation. Furthermore, the Dun & Bradstreet business data file is reviewed and cleaned to improve data content. All estimates of market supply are in nominal terms and are derived from receipts (net of sales taxes, refunds, and returns) of businesses that are primarily engaged in the retailing of merchandise. Excise taxes paid by the retailer or the remuneration of services are also included, for example, installation and delivery charges that are incidental to the transaction.”

This explanation shows that Esri s methodology is a rather complicated, multi-step process, which involves quite a few data sources and data manipulations. Part of those complications is the result if its recognition that the estimates based on one data source or another are in need of improvement.

Dun & Bradstreet (D&B) and MRT data are needed because:

  • The business census data are aggregated to such geographic units as states and counties that are not congruent with the non-legally defined boundaries of many study areas. Addressed based D&B data can be aggregated to the needed geographic boundaries.
  • The business census is done only every seven years and the actual data are only released two to three years after it is taken. Depending when they are used by outside analysts, the business census data can be three to six years old. The census data cannot be depended on to detail what businesses are there and what they are like. The D&B data are supposedly more recent and able to establish who is there now. ESRI’s methodology recognizes the importance of knowing how many firms by size there currently are in each retail NAICS code in a leakage analysis study area. The NES data are used to address the presence of small firms.

On a key question, whether and how Esri uses D&B’s data on firm sales, Esri’s white paper is unclear.

My take aways from Esri’s explanation of how estimates of retail sales are made are that it

  • Reveals the sources of some of the data they use, e.g., the Census Bureau, BLS, Dun & Bradstreet. However, they do not detail the data obtained for each retail firm (e.g., name, NAICS code, sales, number of employees) or, very importantly, how these data may have been statistically manipulated to produce their estimates. This is understandable to the degree that their methodology provides a competitive advantage over their rivals and therefore needs to be kept confidential. Nevertheless, it makes assessing their estimates more difficult.
  • Additionally, and perhaps even more importantly, Esri provides no indication that they have validated their estimates, i.e., shown that their estimates are accurately measuring what they are supposed to be measuring, or that they even made an attempt to do so. How, then, can we know that their methodology is sound? How can we have confidence in them?
  • Is a complicated methodology with many steps. I was taught that every time basic data are manipulated or added to there is a potential for a new error factor to be brought in. Esri has added steps to its methodology to overcome some known errors. Whether they are successful or simply add new errors is unknown.

I have not found any better explanation from Nielsen or its corporate predecessor Claritas about how they estimate retail sales.

A Closer Look at Basic Supply Side Information: Retailers –Their Number, NAICS Codes, Sizes and Sales

Accurate basic firm level data are absolutely essential to an accurate and useful to a traditional leakage analysis. Having directed numerous surveys of business firms, especially retailers and those prone to renting office spaces, and conducting countless personal interviews with business owners and managers in the USA and France, I am well aware of the difficulties involved in getting reliable data from business firms. This is especially true for the sales and finances of small and medium-sized operators.

As the following quote from Mike Stumpf indicates, I’m not the only one that questions the verisimilitude of the firm-level data that is often the foundation of a leakage analysis. From a recent email, here is Mike’s description of how ESRI looked at one area’s business mix:

“ESRI looked at the area’s business mix, then talked about the very large number of establishments in the financial sector. ESRI’s database lists every ATM as a business establishment! A grocery store with a deli, pharmacy, and liquor store can be listed as four different businesses, each with its own estimated sales, but then they do not pick up the restaurant in the Mexican grocery. Their sales estimates for one Mexican bakery were under $100,000, but they actually do more than that in a single month.”

Both Nielsen, ESRI and other leakage analysts obtain basic data about the stores in a study area from other market research firms, such as InfoUSA, Dun & Bradstreet or, its subsidiary, Hoovers, etc. that focus on firm-level data gathering. Nielsen and ESRI combine that info with data from the economic census and other sources to produce their estimates. In a sense they are heaping estimates upon estimates, with each potentially having unknown error factors that may combine through addition or multiplication.

Over the past three decades my company, for our client’s downtowns, has purchased establishment level data from some of the same business market researchers that have supplied firm level data to ESRI and Nielsen. On occasion, we obtained such data from local economic development agencies. Our experience with those data made a deep impression because:

  • There were usually error factors of 20% to 40% in the listings of firms. The errors, as might be expected, were most frequent among the really small firms that have a high probability of failure and turnover, but some of a downtown’s larger retailers were too often also omitted or listed after they had closed. We have seen annual retail churn rates in our clients’ downtowns (they include newcomers as well as closures) that can range between 5% and 30%. Unless these gatherers of firm-level data research businesses annually, which is very expensive, there is a high probability that their collected facts will be out of date. The year-to-year errors can really add up.
  • The Retail NAICS codes often may be wrong. Our discussion with one of the businesses that gathers firm-level data revealed that the downtown firms themselves were surveyed and provided their NAICS code(s). They were aware that errors were frequent. They attributed this to a difficulty in getting knowledgeable executives to respond to their survey. My review of the NAICS codes listed for retailers in the downtowns we bought data for frequently found numerous questionable codes.
  • I have supplied the NAICS code for my firm in the business census and to insurance companies and found out what many of the not large downtown business owners have told me over the years: selecting the right NAICS code isn’t easy, its takes too much time, and consequently, they just put down whatever is easiest to get the task done. Some reported giving the job of responding to the business census and enquires from firms like InfoUSA to their secretary or assistant or salesperson. I suspect that there are even substantial NAICS code errors in the US Business Census returns — especially from respondents who are the types of independent retailers found in our small and medium-sized downtowns and urban ethnic neighborhood commercial centers. DANTH, Inc. has conducted many surveys of businesses over the years, including the largest corporations headquartered in Manhattan, those in the healthcare industry, manufacturers in Nassau County, NY, etc. We quickly found there was a tendency for high-level executives to avoid responding to our survey. They felt it was a waste of their time.
  • Getting accurate firm-level retail sales data is really, really difficult. In addition to the surveys mentioned above, DANTH has also done numerous surveys of downtown merchants. Unsurprisingly, we found that there is strong resistance among most independent downtown merchants to providing such information. Interestingly, they sometimes reported that, off hand, they don’t know that information. The managers of the national chain stores also are loath to reveal their sales statistics. Many downtown managers – probably most – can attest to the difficulty of obtaining such information.
  • One of the business market researchers reported to me a few years ago that they have similar challenges getting information about a firm’s sales, workforce and finances. For the national chains, they use the chain’s average store’s sales. For the independent merchants they ask for the number of employees – which they said small retailers are more willing to provide – and multiply that by the average sales per employee for that NAICS code in the state. D&B is in the credit rating business, so small and medium-sized retailers who are concerned about their credit rating have an incentive to respond to D&B inquiries.

Getting basic information about small and medium sized business is a tough job, fraught with potential errors. Given that this type of information is such a basic building block in the often-complicated process of estimating the retail sales of a study area, it seems wise to treat such estimates with caution and prudence.

DANTH, Inc. has used leakage analyses generated by Claritas Esri and UWEX. In each instance, I created small tests of the data. Sometimes running a leakage analysis on a small of part of the study area enabled a comparison of the number of firms with my visual counts. If a NAICS category was identified as being very strong, I tried to verify that by visual inspection and interviews with local business operators. In a project in Sherwood, WI, a fast growing community, I thought that both the demographic data and the leakage analysis of one provider were badly out of date. Instead, I used the UWEX Gap Calculator and data from the Census Bureau’s American Community Survey. On another project, one of my little tests was to see if the leakage analysis had captured the sales of a food operation inside of a convention center. Finally, there is the “making sense test.” Sometimes I’ve looked at one of these leakage analysis reports and felt it just does not make sense. It just may be that the numbers do not jive in my mind with what I saw touring the study area and what I heard from local business people and economic development officials. Reports that fail the making sense test cause me to take a much closer look at the situation.

Sales Tax Data. In a number of states, retail sales tax information sorted by NAICS codes are gathered and published for the state, counties and sometimes the cities. These data are potential analytical gold for those researchers who find those geographic units congruent with their study area. Even when they are not, the validity of such data may provide a strong foundation for making needed extrapolations for the non-legally defined study areas.

Most of my project experience has been in states that do not provide such information. On the two occasions when I did projects in such states, I encountered some problems that impeded my use of the sales tax data. In one state, the sales tax is applied to only a portion of the merchandise in various NAICS categories and we could not determine what those percentages were. In the other state, the data for the counties were only provided at the two-digit NAICS code level, e.g., that of the entire Retail Trade category.

However, I have been told by some economic development consultants in California whom I respect that the state’s sales tax by NAICS code data enables them to produce more reliable leakage analyses.

Avoiding the Need for Retail Sales Estimates

In the 1990s and early 2000s, instead of the usual leakage analysis, DANTH, Inc. used an approach based on the consumers’ satisfaction with the retail services in their market area. We thereby solved the unreliable retail sales estimates issue by avoiding the subject. Our analysis focused solely on the consumer. Back then, response rates to our telephone surveys were far more acceptable than today’s 30% or so and we asked respondents if they were satisfied with the stores within a 15 minute drive in each of the higher level retail SIC or NAICS codes. Those that were dissatisfied we deemed as probable out-shoppers and open to being attracted by new retailers within the trade area. We also knew from the survey about these dissatisfied shoppers’ incomes, education levels, age cohorts, etc.

Using consumer expenditures potentials generated by ScanUS, Esri or Claritas we also provided estimates of the consumer expenditures for some of the products associated with each retail SIC/NAICS code. The expenditures of the “underserved” shoppers were, in a sense, up for grabs and prone to being spent out of the trade area. Back then we were not able to match completely BLS product categories to NAICS codes as can now be done. Our retail growth strategies were based on capturing these up for grabs shoppers and a sample of their retail expenditures.

This approach is sort of like a leakage analysis, though not the same. However, it enables an analyst to answer the same important retail growth questions, with far fewer questions about the validity and reliability of the data used.

DANTH has not used this approach recently because the cost of a really good telephone survey has become unaffordable and their findings less reliable, while the use of lower cost non-panel online surveys have serious sample issues.

Of course, others have used local surveys to ask respondents about whether they mostly buy various products in town, out-of-town or online. That, too, tells you about the leakage of shoppers and avoids the need for data about retail sales. Here, again, being able to do an affordable survey with a representative sample is a critical question.

My guess is that within the next 10 years online survey techniques will solve the sampling issue and affordable surveys will be possible. Then researchers will be able to answer the questions that leakage analyses are intended to address without needing to know about local retail sales or complicated methodologies for estimating them.

The Demand Side – Consumer Expenditures by NAICS Category

BLS Based Approaches. As I have written elsewhere, the best way of obtaining reliable, acceptably accurate data on how consumers spend their money is to survey them. However, such surveys require substantially more methodological skills and resources than most other kinds of surveys. For many consumer items, the relevant information must be obtained close to the points in time when the expenditures are made, lest memories of them quickly fade. For example, the Bureau of Labor Statistics (BLS) probably sets the gold standard for this type of research and it has half of its 14,000 survey respondents keep diaries of their expenditures and interviews the other half 5 times over 15 months. In contrast, I’ve seen many inept downtown and Main Street surveys that ask about shopper expenditures that are oblivious of the challenges involved, have no where near the rigor of the BLS surveys, have relatively few respondents and their results aren’t worth the paper they are printed on.

While the BLS expenditure data are solid and a good starting point, using them to inform estimates of consumer expenditures in a leakage analysis study area presents a number of barriers that must be overcome:

  1. The BLS uses categories to order its data that are product oriented, about the things people buy not the kinds of stores consumers buy in. The demand side retail sales data are ordered by the NAICS codes that are about the characteristics of aggregations of retail stores. To be useful in a leakage analysis, the BLS data have to be “crosswalked” into the NAICS categories. Such data crosswalks are enabled by a list of the products that are typically sold in the different retail NAICS codes that was produced by the Census Bureau. Using the list can take a lot of effort. Computerizing the list makes things easier, but also takes resources. Both Nielsen and Esri do these data coding crosswalks. This data crosswalking is a possible opportunity for errors to enter the analysis, but I think the probability of that happening are relatively low.
  2. The geographies used by the BLS are far, far larger than the size of any retail leakage study area. The 14,000 respondent BLS sample is still too small to provide enough respondents to be directly used even at the state level of analysis; the sample would average only about 280 respondents per state. Consequently, some methodology is needed to make estimates of consumer spending within retail leakage study areas that is based firmly on the BLS data. Both Nielsen and Esri have developed methodologies to make such estimates based on the BLS data. My research experience says these research tasks have significant opportunities for errors to enter into the analysis.
  3. Different types of people have different spending levels and patterns and different retail leakage study areas will vary in the kinds of people that live in them, e, g., by income, education, race, lifestyle group. The BLS data based estimation process needs to be able to adjust expenditure estimates by taking these population differences into consideration. Both Nielsen and Esri use their demographic and lifestyle segmentation databases, Prism and Tapestry, to make these adjustments. My research experience again says these research tasks have significant opportunities for errors to enter into the analysis.

The data presented in the table earlier in this article about City XYZ shows big differences in Esri’s and Nielsen’s estimates of consumer expenditures. This suggests that while they both are trying to address the problems just enumerated above, their methodologies vary significantly. Again we are posed with the problem of whose estimates, if any, should we trust?

The BLS data covers household expenditures. Used in a leakage analysis they inform only about the expenditures of study area residents. They do not appear to inform about the expenditures of day and overnight tourists and non-residents who work in the area. Yet, accurate estimates of retail sales should include the sales dollars from those types of customers. For many downtowns, the non-resident customer base can be extremely important. There consequently seems to be an imbalance between the customers included in the analysis on the demand and supply sides. Bill Ryan has recently upgraded UWEX’s Gap Calculator to incorporate data on the expenditures of tourists and people working downtown. Others should follow suit.


Also, of increasing importance, the BLS data does not tell us whether consumer expenditures are being made in brick and mortar shops or online. The total of online retail sales now only accounts for about 6.5% of all retail sales nationally. However, as can be seen in the above table,

  • By 2017 it is anticipated, if trends continue, that e-commerce will account for 50%+ of the sales for music, videos, books, magazines, computer hardware and software, toys and games, electronics and appliances,
  • By 2017 e-commerce is projected to account for 25%+ of the national sales of furniture, sporting goods, office equipment and supplies, clothing, accessories and footwear.

 The vast majority of online purchases will be going to retailers based geographically outside of the retail leakage study area, perhaps across the country or even on another continent, yet they are very close because they can be reached from a living room chair almost instantaneously via a smartphone, tablet or computer. The online purchases are, of course, leaked purchases. The important question then becomes do local merchants have a real chance to recapture these dollars? The strong trend now is for online sales to grow at a rapid pace. Younger Americans, especially the Millennials feel very much at ease with making online purchases and they will be growing in number. Older folks are getting more accustomed to buying online, but not like the younger age cohorts, and many of them will soon fade away. Consequently, a strong argument can be made that, for the foreseeable future, it will be much harder for local brick and mortar merchants to win back dollars spent on the internet than for dollars spent in other brick and mortar shops located beyond their trade area’s boundaries.

If this is the case, then knowing about how many consumer dollars in a study area are going online is, in effect, identifying “dead dollars” that now are well beyond the reach of local retailers who really have little hope of recapturing them – unless they, too, are able to become strong competitors on the internet. Unfortunately, this is the type of information that many downtown leaders need to know, but don’t want to hear.

None BLS Approaches. Leaders of EDOs in many small and medium sized communities may want a leakage analyses done for their downtown, but the don’t have a lot of money to pay for a consultant. Some consultants and economic development organizations have constructed leakage calculators that can be used by EDO leaders or their staffs either for free or for little cost. These builders of these calculators do not have the funds that take the BLS data and crosswalk then into NAICS codes, or to generate expenditure estimates at the study area level, or to adjust such estimates through the use of a lifestyle database. Instead, they take a simpler and less expensive approach:

  • They take the retail sales by NAICS code at the state level and adjust them for inflation
  • Then they divide those sales numbers by e the population of the state to get per capita sales per NAICS code at the state level.
  • Then they treat these per capita sales estimates as measures of per capita spending. They identify the population of the study area in question and then multiply the NAICS per capita sales estimates by that population number to estimate total consumer expenditures by NAICS codes for the study area.
  • They then identify the median household income in the trade area and adjust the consumer expenditure estimates accordingly.

In a lot of ways this is a simpler and easier way of estimating consumer expenditures in a study area than how Nielsen and Esrsi do it. I’ve used the UWEX Gap Calculator and can attest that its easy to use. It’s intended users are small and medium-sized downtowns, where local leaders can input their own data about the retailers in their trade areas.

The use of sales data at the state level to estimate consumer expenditures at a study area level may raise a few eyebrows. It did mine. Since the state retail sales data includes tourist sales, those dollars are also finding their way into estimates of resident retail expenditures and inflates them inversely to the actual flow of tourists through the study area. The theoretical explanation for it escapes my understanding. But, can Nielsen and Esri prove that their estimates are any better?

Given that there is no evidence that the gap calculators do any worse than the leakage reports the big private market data firms put out, who’s to say they are not worth considering. The price is right. Moreover, they are methodologically rather transparent: it’s easy to understand how all the estimates are being made.

Some Final Comments.

To properly do our research we need to live in the real world and in that real world we most relay on doing secondary or tertiary levels of analysis on primary data collected by someone else. Professionally, we should have been trained to understand that we must be very cautious about how we do our secondary analyses and transparent about our methodology when we present our findings to others. We need to be even more cautious about how we accept the secondary analyses done by others – especially when their methodology is not transparent and their findings have a significant number of apparent errors and conflict with the findings of other analysts.

That is why I argue that the estimates of retail sales and consumer expenditures generated by most leakage analyses should be treated with caution and used with prudence.