Market segmentation

Origin

Wendell R. Smith set the idea out in July 1956, in the Journal of Marketing, volume 21, number 1, pages 3 to 8 [1]. The title says what he was doing: “Product Differentiation and Market Segmentation as Alternative Marketing Strategies”. American firms in the mid 1950s faced imperfect competition and a market that no longer behaved like one homogeneous mass. Smith’s contribution was to name two ways of responding, and to argue they were different strategies rather than the same thing under two labels. The distinction was still muddled thirty years later: Dickson and Ginter opened their 1987 paper by noting “considerable misunderstanding about their meaning and use” [5].

His definition, on page 6: segmentation means “viewing a heterogeneous market as a number of smaller homogeneous markets” [1][2].

The idea entered the mainstream in two steps. Daniel Yankelovich’s 1964 Harvard Business Review article argued that demographics had lost predictive power and that segmentation should also consider “buyer attitudes, motivations, values, patterns of usage, aesthetic preferences, and degree of susceptibility”, across ten consumer and industrial markets [3]. Yoram Wind’s 1978 review in the Journal of Marketing Research codified the method: problem definition, research design, data collection, analysis, interpretation and implementation [4]. That is a review of how to segment, not a test of whether segments predict brand choice.

The paper is six pages long and is still cited every year, seventy years on.

Philip Kotler’s Marketing Management, first published in 1967 [6], turned the idea into the sequence every marketing graduate learns: segment, target, position. Kotler and Keller’s current editions list four bases for consumer segmentation, geographic, demographic, psychographic and behavioural. David Aaker’s Strategic Market Management places segmentation inside customer analysis, the first question a strategist asks about a market [7]. Robert Cialdini did not address segmentation. Influence describes principles of compliance that operate across people and offers no view on whether different segments need different principles [15].

What the evidence shows

The record is uneven. Segmentation is one of the most used tools in marketing and one of the least tested at the level that matters to a brand owner. The literature from Yankelovich to Wind is methodological [3][4]. It tells you how to find segments. It rarely asks whether the segments a brand targets contain different people from the ones it does not.

Two things are well supported.

First, categories are heterogeneous. Buyers differ in usage rate, occasion and need, and Smith’s point about “divergent demand” [2] stands. Nobody in the debate below disputes it.

Second, breadth of targeting predicts business results. Les Binet and Peter Field analysed the IPA Effectiveness Databank, which at the time held 996 campaigns entered for the IPA Effectiveness Awards from 1980 to 2010, covering more than 700 brands in over 80 categories [13]. Campaigns that targeted the whole market reported on average 1.8 very large business effects and 6.7 very large effects in total. Campaigns aimed at new customers reported 1.7 and 5.4. Campaigns aimed at existing customers reported 0.6 and 2.4 [13, Figure 18].

Very large business effectsVery large effects, total

Existing customers0.62.4

New customers1.75.4

Whole market1.86.7

The step from existing customers to new customers is the big one; adding existing customers back in, to reach the whole market, adds a little more.

Source: IPA, Binet and Field, The Long and the Short of It, 2013, n=996 campaigns, 1980 to 2010, Figure 18. Redrawn by Brave New Media.

Unit: very large effects per campaign

Show the data

Campaign target	Very large business effects	Very large effects, total
Existing customers	0.6	2.4
New customers	1.7	5.4
Whole market	1.8	6.7

Efficiency moves the same way, only more sharply. Measured as annualised market share growth per ten points of excess share of voice, whole-market campaigns scored 2.6, new-customer campaigns 0.3 and existing-customer campaigns 0.2 [13, Figure 21]. Binet and Field’s own summary: “the benefits of broad reach considerably outweigh the benefits of tight targeting: a finding that directly contradicts much of the current orthodoxy” [13].

Existing customers0.2

New customers0.3

Whole market2.6

Effectiveness rises step by step with reach (Chart 1), but efficiency jumps only when the campaign stops choosing between new and existing buyers.

Source: IPA, Binet and Field, The Long and the Short of It, 2013, n=996 campaigns, 1980 to 2010, Figure 21. Redrawn by Brave New Media.

Unit: annualised ESOV efficiency

Show the data

Campaign target	Annualised ESOV efficiency
Existing customers	0.2
New customers	0.3
Whole market	2.6

The loyalty finding is the sharpest. Only 12 percent of IPA campaigns that set out to increase loyalty achieved a very large loyalty effect, against 47 percent of acquisition campaigns that achieved very large penetration growth [13]. Their 2017 update, Media in Focus, re-ran the analysis on 2008 to 2016 cases and found the same ordering for annual market share growth: campaigns addressing both new and existing customers grew share fastest, new-customer campaigns next, existing-customer campaigns least [14, Figure 14]. By 2014 to 2016, pure loyalty campaigns had fallen to 3 percent of IPA entries [14].

One caveat is ours, not theirs. The Databank is built from award entries. Whole-market campaigns may also be the ones with the biggest brands behind them. The share-of-voice measure controls for spend, but the cases are not a random sample of marketing activity.

The challenge

The strongest published critique comes from the Ehrenberg-Bass Institute, and it is narrower than it is usually reported. It does not say segments do not exist. It says brands do not own them.

Hammond, Ehrenberg and Goodhardt put it to the test in 1996 in the European Journal of Marketing. Using consumer panel data across more than 20 grocery categories in several countries, they found “little brand segmentation”: the profiles of competing brands’ buyers differed little on the usual socio-demographic measures, and brands in the same category “tend to be bought by similar kinds of people” [8]. They also noted that “few, if any, examples of marked brand segmentation are cited in the literature” [8]. Kennedy and Ehrenberg gave the finding its blunt title in 2001, “There is no brand segmentation”, in Marketing Research [9].

The finding has been re-tested rather than left alone. Uncles, Kennedy, Nenycz-Thiel, Singh and Kwok re-examined it in 2012 with newer data, extending the scope to brands in emerging markets, private labels, variants and composite segments, and reported that “despite attempts by marketers to differentiate brands and provide customized features for distinct target audiences”, user profiles of directly competing brands seldom differ [10]. Anesbury, Winchester and Kennedy added the time dimension in 2017: profiles seldom differ between brands and seldom change over years [11].

Byron Sharp gave the finding a chapter title in How Brands Grow, “Our buyers are different”, and a second chapter on competition, “Who do you really compete with?” [12]. His examples run from Ford and Chevrolet owners in the 1950s to UK car owners in the 1990s, whose agreement with attitude statements barely moved across brands [12]. The institute’s later B2B work, on infrastructure-as-a-service, business intelligence and CRM software, claims the same law holds there: “the only meaningful difference in the customer base is its size, not its composition” [16]. The institute’s business intelligence data, published with the LinkedIn B2B Institute, show what that means in practice: 80 percent of Amplitude’s customers also use Google, and on average 68 percent of any brand’s customers also buy AWS, so the buyers a small brand calls its own are mostly the big brands’ buyers too [19].

Penetration, percentAverage sharing, percent

Amazon (AWS)62%68%

Microsoft Azure49%52%

Google (GCP)46%61%

IBM46%59%

Oracle Analytics22%34%

SAS17%25%

Microstrategy11%26%

SAP BI suite10%21%

Tableau6%15%

Amplitude5%14%

Mixpanel5%14%

The sharing column follows the penetration column down the table. A small brand's "target customer" is mostly a customer of the big brands already, which is the duplication of purchase law showing up in B2B software.

Source: LinkedIn B2B Institute with Ehrenberg-Bass Institute, How B2B Tech Brands Grow, 2024, n=293 IT decision makers, Figure 4, p. 29. Redrawn by Brave New Media.

Unit: percent of IT decision makers

Show the data

Brand	Penetration, percent	Average sharing, percent
Amazon (AWS)	62	68
Microsoft Azure	49	52
Google (GCP)	46	61
IBM	46	59
Oracle Analytics	22	34
SAS	17	25
Microstrategy	11	26
SAP BI suite	10	21
Tableau	6	15
Amplitude	5	14
Mixpanel	5	14

What is contested, and by whom. Segmentation’s defenders hold that the Ehrenberg-Bass tests use the wrong variables: demographics and broad attitudes were never the bases a good segmentation would use, and needs- or occasion-based segments would show differences these tests cannot see. Uncles and colleagues acknowledge “potential boundary conditions” in their abstract [10]. The Ehrenberg-Bass reply is that whatever the basis, the brands end up with the same buyers, so the segment describes the category, not the brand. The literature has not resolved this. The two sides measure different things: who the brand wanted, and who the brand got.

Where practitioners get it wrong

Three misapplications recur.

Personas become media plans. A segmentation built to sharpen the offer is handed to the media team, who buy only the audience the persona describes. The persona was a design tool. Used as an exclusion rule, it cuts reach, the variable Binet and Field found matters most [13]. The narrowing also fails on its own terms: in Nielsen’s benchmark of more than 44,000 digital campaigns, about half of the impressions aimed at a broad age and gender group reached it, and only about a fifth of those aimed at one gender in a narrow age range did [17].

DesktopMobile

Broad audience53%52%

One gender, medium age range (spanning 16 to 30 years)28%33%

One gender, narrow age range (under 16 years)20%22%

A persona handed to the media team as a buying rule does not even buy the persona: the tighter the demographic, the larger the share of paid impressions that land on someone else.

Source: Nielsen, Digital Ad Ratings Benchmarks and Findings, 2016, n=more than 44,000 campaigns in 17 countries, Q2 2016, press release of 15 December 2016. Redrawn by Brave New Media.

Unit: percent of impressions on target

Show the data

Target definition	Desktop	Mobile
Broad audience	53	52
One gender, medium age range (spanning 16 to 30 years)	28	33
One gender, narrow age range (under 16 years)	20	22

“Our buyer is different.” Brand teams compare their buyer to a stereotype of the competitor’s buyer rather than to the category’s buyer. When the comparison is run on the same variables, on panel data, the difference is usually within a few percentage points [8][10]. The belief survives because nobody runs the comparison.

Loyalty first. Budgets tilt toward existing customers because they are cheaper to reach and easier to measure. The Databank says these campaigns produce the smallest effects and the lowest efficiency of any targeting choice [13][14].

Our own observation. A European supply chain services group asked us to open two markets where the category itself was unfamiliar. We ran one message identically in both, so the variable was the market and not the copy. Two countries, different languages, different competitors, different maturity, and category search presence landed within a point of each other inside sixty days. The consistency was the finding, not the level. Had those buyers needed two segment-specific arguments, the numbers would have said so. They did not.

Buyers do ask to be understood: in Salesforce’s 2022 survey of 13,020 consumers and 3,916 business buyers, 73 percent expected companies to understand their unique needs, up from 66 percent in 2020, while 56 percent said most companies still treat them as a number, which is a case for segmenting the offer and the service rather than the audience [18].

20202022

The experience a company provides is as important as its product or services80%88%

I expect companies to understand my unique needs and expectations66%73%

I expect companies to anticipate my needs56%62%

I feel an emotional connection to the brands I buy from the most53%62%

I expect offers to always be personalized52%56%

Most companies treat me as a number65%56%

Buyers ask to be understood, which is the case for segmenting the offer and the service. Only 56 percent expect every offer to be personalised, and 56 percent still feel treated as a number, so the demand is for relevance, not for narrower advertising.

Source: Salesforce, State of the Connected Customer, 5th edition, 2022, n=13,020 consumers and 3,916 business buyers in 29 countries, p. 10. Redrawn by Brave New Media.

Unit: percent of customers who agree

Show the data

Statement	2020	2022
The experience a company provides is as important as its product or services	80	88
I expect companies to understand my unique needs and expectations	66	73
I expect companies to anticipate my needs	56	62
I feel an emotional connection to the brands I buy from the most	53	62
I expect offers to always be personalized	52	56
Most companies treat me as a number	65	56

What to do with it

1. Segment the category on behaviour first: usage frequency, occasion and need, from purchase or usage data. Record each segment’s size as a share of category buyers. A segmentation document with no sizes is a persona deck.
2. Run the user-profile test once a year. Put your brand’s buyer profile next to the category’s on the same five to ten variables. If no variable differs by more than a few percentage points, remove the phrase “our target customer” from the brief and replace it with “category buyers”.
3. Write the reach target as a share of all category buyers, with a number, in every brand-building brief. A brief with no reach figure is a brief that will be narrowed by default.
4. Split the budget by task. Brand building goes to the whole category through broad-reach media. Activation goes tightly targeted to people in market now, the line Binet and Field draw in Figure 01 of Media in Focus [14]. Keep the two on separate lines in the plan so one cannot absorb the other.
5. Use segments to change the offer: variants, price points, pack sizes, distribution. Do not use them to decide who is not allowed to see the advertising.
6. Re-check profiles every two to three years and expect them to be stable [11]. A profile that moves a lot is a measurement problem before it is a strategy result.
7. Report penetration before loyalty in every quarterly review, and report both against the category, not against last quarter.