(3&4) Repeat-buying patterns PDF

Preview

Summary

Lecture notes...

Description

(3& (3&4) 4) Rep Repeateateat-bu bu buyi yi ying ng pa patter tter tterns ns Tuesday, 10 November 2020

11:54 AM

Polygamous share loyalty towards the brands we buy i.e. Repertoire market

Repeat-Buying /noun/ People who buy an item more than once in given time-period. ➢ Emphasis on building non-cognitivist Empirical Generalisations on buying behaviour ➢ Repeat buying is the incidence of repeat-purchase from one time period to another i.e. we can compare the repeat purchase rate from period to another to see changes in consumer behaviour Average Purchase Frequency = can be considered a measure of repeat-buying for a specific time period (e.g. A month, quarter, year)

Purchase Probabilities = Views buying behaviour as an aggregate expression ➢ 'How often' and 'when' people buy e.g. Brand X was purchased 3 times in the past month ➢ Likelihood to buy again given the number of past purchases ➢ Varies across people according to past purchases ➢ Not uniform across consumers - heterogeneity Purchase propensities

Arry - Has higher probability of purchasing Coca-Cola than the other brands Magda - Has an even probability of purchasing Coca-Cola & Pepsi /noun/ Describing how likely a consumer is to purchase one brand within a category Some people buy the category more frequently than others in any time period ➢ (e.g. I like chocolate bars and I buy it daily, whereas I buy canned soups maybe only once a month) Many people may not buy in a particular time period at all; this depends on the length of the time period ➢ (e.g. Within a week not many people buy canned soup but within a year they have higher frequency) ➢ Relationship between length of time analysing and number of people purchasing from category and purchasing opportunity Measure this amount using penetration = The proportion of people who buy at least once Andrew Ehrenberg (1950) His research focused on the way people bought product categories at different rates i.e. The CB Page 1

distribution of purchase frequency (e.g. Shopper A bought from category x5). He found that the distribution followed a common pattern in all categories. ➢ Knowing the average purchase rate for the category and the proportion of people buying at all, enables you to generate figures for how many people will buy once, twice, three times, etc. ➢ Calculate the expected proportion of sales to classes of buyers e.g. Light or heavy ➢ In order to make predictions, the market must be stationary in the medium term

Near-stationary market Key assumption of NBD-Dirichlet model ➢ Many established brands and categories show little to none change in the medium term ➢ If market is changing = dynamic market ➢ Brand level = similar market share within several years ➢ Category level = similar number of brands competing within the category and penetration ➢ We see stability in the middle and long term at both category and brand level ➢ This is because consumers buy out of habit and not much changes occurs - stationarity This makes near-stationary markets predictable, either through empirical (past) patterns or mathematical models that mimic the market (theoretical).

NBD-Dirichlet Model

The Negative Binomial Distribution predicts single brand data and Dirichlet predicts data on all brands in a category. Applicable across a wide range of brands - empirical generalisation. Gamma Distribution: Describes people buy infrequently, purchase weight and how much, lots of light and few heavy buyers. Rate or weight that people purchase, how much people buy.

CB Page 2

Poisson Distribution: Describes how people buy randomly, purchase timing and when people buy. Timing in irregular but predictable. Purchase timing, how often people come into the category Multinomial Distribution: Describes people buy from repertoire. Multivariate Distribution: People show heterogeneity in their purchases, consumers have different types of repertoires i.e. Different combination of brands

Purchase timing is like raindrops, we can calculate the aggregate amount that rain will drop per second but not specifically when the rain drop will fall.

Zero and First-Order Effects

Zero-Order effect: Last purchase doesn't influence future purchase, we just have brand preferences. Dirichlet Model assumption. There is no order in the allocation of purchases across

CB Page 3

brands; the chance to purchase a brand is determined by how many times the brand has been purchased in the past. The most preferred brand just has a higher chance - based upon how times brand has been purchased in the past. First-Order effect: Last purchase will directly influence future purchase in category. There is some order in the allocation of purchases across brands; there is a higher probability to purchase again the last brand that was bought.

Light Buyers There are non, light, medium, and heavy buyers. There are more infrequent (i.e. Non and light) buyers than frequent buyers. There are heavy buyers but there is only a small percentage of them in comparison to non and light buyers. Regularity = The average amount bought per purchase occasion does not vary much from brand to brand. The APF also doesn't vary much within the category.

Light buyers matter 80:20 Rule, Heavy half principle The notion that 20% of our buyers i.e. The heavy buyers, account for 80% of our sales and 80% of our buyers i.e. Light buyers, account for 20% of our sales.

80% (Light buyers) = 20% sales 20% (Heavy buyers) = 80% sales In reality: 80% Light buyers = 50% sales 20% Heavy buyers = 50% sales

CB Page 4

Implications: Because majority of the buyers buy infrequently and are light buyers, the amount purchased by this buyer class amounts very close and even to the amount that heavy buyers purchase. This highlights that light buyers shouldn't be ignored as they account to a large percentage of the sales Leaky bucket theory ➢ Many brands are bought infrequently by their buyers - irregularity is the norm ➢ Light buyers may miss quarters and be classed a 'lapsed' ➢ Lapsed buyers are replaced by new buyers i.e. Light buyers who missed the reference period ➢ Imagine a bucket with holes with continuous flow of water ➢ This is why sales remain fairly stable Implications: Majority of buyers buy infrequently therefore LIGHT BUYERS MATTER, they are an important part of the customer base. NON BUYERS matter too, they might purchase in the next period.

Patterns in Repeat purchasing: IMPLICATIONS 1. Light buyers MATTER - Light buyers still manage to account for a lot of your sales - simply because there are so many of them - So many of your customers are light because buyers buy other brands, and because light buyers are the norm for the category too 2. 80:20 Rule - Generally the top (heavy) 20% of customers generate about half of your sales - heavy half, light half - This concentration is slightly less for larger brands. - So growth seems to come more from light customers - Can't force buyers to purchase more as they would have already saturated their needs, they can only buy so much 3. Less than 100% Repeat-buying is the NORM - Customers aren't being disloyal or dissatisfied, just repertoire buying - Period-by-period repeat buying is fairly low but much the same for the various brands For example, typical quarter by quarter repeat buying is around 40-50% 4. -

Non-customers matter too They are potential buyers, they might buy in the next period. Leaky bucket theory and zero-order effect i.e. They have potential to purchase NBD shows this

Using the theoretical values provided by NBD-Dirichlet model, we can match them to real-life observed values - we find that the theoretical and observed fit well. ➢ Makes metrics meaningful ➢ Benchmarks and know what to expect - the norm ➢ Use it to review brand & strategies

From theoretical benchmarks we know: ▪ The timing and rate that people buy the category (NBD model) ▪ That our customers buy other brands

CB Page 5

▪ What this means for the distribution of timing and buying rates for our brand

Repertoire buying: Polygamous Loyalty Why does this happen: Various reasons ➢ Complete awareness is impossible ➢ Buyers seek variety ➢ Buyers buy for different occasions/purposes ➢ Price promotions and other point of sale ➢ Availability of brands ➢ Household include many buyers ➢ Etc...

The rate at which a brand gets bought matches a statistical distribution called the ‘NBD’ ➢ People buy infrequently

Double Jeopardy Empirical Generalisation - Marketing Law Bigger brands (higher market share) have more buyers (higher penetration) and are bought slightly more often (higher purchase frequency) Smaller brands (lower market share) have fewer buyers (lower penetration) and are bought slightly less often (lower purchase frequency) - Smaller brands "suffer twice" Purchase Frequency only slightly varies but is in line with brand size Averages allows to see deviations & which brand is big/small Know what to expect i.e. Benchmark Main difference between brands is brand size metrics Brands with fairly equal market share also have fairly equal purchase rate Loyalty varies little between competing brands and when it does, it is in favour of bigger brands ❖ Niche brands are RARE and when they are, they are tiny ❖ ❖ ❖ ❖ ❖ ❖

CB Page 6

Difference is larger in penetration than purchase frequency. Smaller Brands have buyers who have attitudes towards the brand similar to other brands i.e. Same "love" for other brands. ➢ They have "better quality" customer base i.e. Heavy category buyers as they're more involved in category in order to know about small brands THEREFORE - Big Brands attract light buyers as they don't have much brand awareness. ➢ Light buyers just purchase the more prominent brand i.e. Mental and physical availability i.e. Natural Monopoly Effect, seen through Category Buying Rate ○ Big Brand = Low CBR ○ Small Brand = Higher CBR ➢ There's more light buyers than heavy DJ Pattern can be used for loyalty benchmarks: brand tracking, new brands & brand growth, positioning, loyalty initiatives Natural Monopoly Effect

CB Page 7

Brands have lower CBR because light buyers don't purchase frequently from category Deviations Deviations from Double Jeopardy pattern are rare and often fail to persist over multiple time periods. The following typical instances of deviations can be observed in buying behaviour: • Niche brands • Change of pace brands • Large brands • Private labels Why DJ? All comes down to popularity rather than differentiation - most brands compete head on and are similar i.e. All 'capable of doing the job', satisficing. More people know about big brands than small brands - people who buy small brands ALSO know about big brands whilst buyer of big brands may not know of small brands

Duplication of Purchase Law All competing brands within a category share customers and how brands share occurs in line with brand size (i.e. Penetration and market share) - Repertoire market - 100% loyals are rare - SCR usually below 50% - Lots of cross-purchasing of brands, related to polygamous loyalty A brand will share MORE of its customers with brands that are larger than it, and LESS of its customers with brands that are smaller than it ➢ Partitions (Identify, describe, and reason for occurring) ➢ Know how to read a DoP table ➢ Implications

❖ ❖ ❖ ❖

Brand share customers in line with market share Every brand has the same proportion of its customers who also buy a 2nd brand All brands share a larger proportion of their customers with larger brands The main determinant of competition between brands are their relative sizes NOT positionings

Most brand customers buy different types of brands and products e.g. Luxury chocolate buyer also purchase cheap chocolate CB Page 8

Similar percentage of brand buyers purchasing from competitors Duplication of Purchase Table

Deviations Partitions = Cluster of brands/SKUs compete more closely. More likely to sell to the same people ➢ Associated with functional differences/qualities ➢ e.g. Gluten-free, vegan, dietary products ➢ Product offerings are similar ➢ Share more or less than expected To be partitioned, both brands are sharing more/less than expected, not just one brand E.g. A large brand is sharing more of their base with smaller brand than expected - due to partitions and functional attributes PARTITIONS IS DIFFERENT FROM SEGMENTATION - Segm. Is more about the consumer's qualities than product qualities. Practical Application 1. To develop understanding of competitive structure i.e. the different markets, categories, and subcategories available 2. Identify partitions - check if you have full category coverage i.e. Have offerings in appropriate partitions if its important and related to category 3. Benchmark for cannibalisation - Avoid stealing buyers from your own brand when releasing a new one i.e. Avoid having similar offering. There will be some level of sharing but not excessively 4. Assist in analysis to determine which brand or SKUs to get to rid off to ensure there aren't

CB Page 9

too many that confuse consumers. For when retailers demand rationalisation Implications ➢ Brands compete in a largely un-partitioned market: the battle is to be considered by more customers more often, NOT to change your customer base (more loyal, more passionate) but to enlarge it !! ➢ To grow a brand more customers are needed (mass marketing, ‘here I am’) ➢ About competing dynamics: brand customers are occasional and they buy other brands even more often; to be successful, it is necessary to capture more light buyers and to maintain buying propensities -> vital role of advertising (reach!) ➢ No need to convince buyers the brand is best or different, no point in trying to make customers ‘solely loyal’, passionate, advocates etc. ➢ When DoP is evident, so is DJ Pattern

NBD - Dirichlet Model Inputs and outputs ➢ Use calculations of BPM ➢ Number and % of buyers who buy x1, x2, etc. ➢ Sales importance of buyers ➢ Repeat-buying from one period to another ➢ Growth over time in penetration, purchase freq. Etc

If markets are stationary, BPM will stay relatively the same for each period

CB Page 10

Assumptions: ▪ Stable/Near-stationary market conditions ▪ Consumers have steady long-run purchase propensities ▪ Poisson distribution - irregular purchase distribution but steady in the long-term ▪ Zero-order effect - our future purchase is independent from past purchases ▪ Compounded by a set of statistical distributions that capture some key aspects of how people buy - Poisson and Gamma Distribution Implications ➢ Buyers have steady propensities to buy (which brands and how often) ➢ These propensities are independent of what they bought previously (zero-order)

Dirichlet Distributions - Brand Choice Multinomial Distribution Multi-brand buying Zero-order effect in brand buying. Past purchases doesn't effect future purchases. Multivariate Distribution Heterogeneity. Variation in the brands we have within out repertoire. Different for each individual.

Regularity in Brand Choice 1. Loyalty levels vary little from brand to brand (e.g. similar purchase frequency or repeat purchase rate, as well as defection rate) 2. Slight gaps (excess or deficit loyalty) are linked to the SIZE of the brand (e.g. their market share or market penetration), which influence also competing dynamics (‘sharing’ of customers between brands) - biggest difference between brands is their size 3. Brand ‘switching’ is quite the norm, but the level of switching changes across categories In addition to the earlier NBD assumptions, the Dirichlet assumes that: Zero-Order effect 1. Consumers do not devote their category purchases solely to a brand (multi-brand buying) 2. Consumers have steady propensities to buy particular brands, but a different propensity for each brand (heterogeneity) E.g. 40% propensity to buy Special K, 20% to buy Weetbix, 10% propensity to buy Rice Bubbles etc. The Zero-Order effect shows that there is an "as-if random" pattern occurring i.e. Brand purchases are zero-order events.

Implications of DJ and DoP Marketing ▪ Brands compete in a largely unpartitioned market ▪ Big brands are main competitors, main battle to grab customer's attention ▪ Focus is to enlarge customer base NOT make them more loyal

CB Page 11

▪ Don't need to 'convince' customers you're better than other brands, focus on just making sure customers consider your brand ▪ When DoP is evident, so should the DJ Pattern ▪ Brands compete head-on, not much differentiation between brand-profiles ▪ Brand loyalty is predictable but outside manager's direct control - increasing penetration size indirectly effects loyalty i.e. DJ pattern Growth ▪ To grow, must increase penetration ▪ Mass marketing is path to growth - appeal light buyers Maintenance & competition ▪ Customers are light buyers - they don't think about brand frequently ▪ Marketing's focus is to remain in memory and as a choice i.e. Mental and physical availability - Brand Salience ▪ Repertoire buying behaviour Advertising ▪ Successful advertising helps maintain loyalty i.e. buying propensities ▪ All about reminding than persuading ▪ Encourage habitual buying i.e. Remind to purchase our brand ▪ Reach/create awareness for non, light, and lapsed buyers - they MATTER ▪ Brand-profiles are similar in consumer's view therefore differentiation isn't super important ▪ brand salience, mental availability, awareness

Patterns in Repeat Buying - Tutorial Questions 1) Most repeat purchase markets can be described as stationary in the medium term. What does this mean and how do we know this is the case? What are the implications of this finding for managers? When repeat purchase markets are described as stationary, it means that changes in the purchase frequencies of the market are very minimal in the long term or in the short-term there will be some fluctuations in sales due to promotions but generally do not have a long-lasting effect. This stationary pattern is mainly because consumers are habitual meaning consumers have a natural loyalty within their repertoire of brands and any defections are levelled out by acquisitions in the market because majority of buyers are light buyers. Light buyers of brands in the stationary markets are replaced by new buyers periodically who have the same purchase rate as the previous light buyers which is what helps create the stable pattern. The boundary condition of this finding is that stationary markets can only be found in the medium term because the regularity of these purchases seem near random being that the timing is irregular but at the same time show a pattern that is somewhat predictable in the medium term. Another buyer behaviour pattern found in stationary markets is the zero-order effect where recent purchases don’t determine future purchases and is mainly based on preferences of brands. The predictions of the likelihood of a future brand purchase in zero-order effect is obtained from the ratio of past brand purchases. These findings are found using the NBD-Dirichlet model but more particularly the

CB Page 12

NBD model as it computes the repeat-purchase and new purchase rates for different periods therefore helping marketers identify patterns in the market. The implications of these findings for managers are that light buyers matter as they account to majority of the sales and customers in the market. As most buyers are light buyers and purchase from a repertoire of brands, repeat-buying is low which is normal, buyers aren’t necessarily disloyal or dissatisfied with the brand they previously purchased from particularly because of the zero-order effect. This also implies that just because shoppers of a category aren’t a buyer of your brand, doesn’t mean they won’t purchase in the future. From this information we can see that managers should foc...