Which CEPs to build?

Category Entry Points (CEPs) are retrieval cues that category buyers use to access useful brands from memory. In this way, they act as ‘distribution outlets’ in the mind. Just like retailers, CEPs can be judged by the amount of traffic they get. This mental traffic via a CEP can be thought of in two dimensions:

• CEP Penetration – how many people encounter each CEP
• CEP Frequency – how often people encounter each CEP

Most categories have a wide variety of CEPs, which leads to the question: how do I select which CEPs for my brand to message? In a previous report (Report 124: Category Entry Points,To Own or not to Own: Is that the question?), we highlighted the futility of trying to only own a single CEP – but obviously we also can’t message all CEPs. One criterion to narrow down the list is to consider how often a CEP occurs in the lives of category buyers. Is penetration alone sufficient? Or is frequency also important to gauge? Is it typical to have CEPs with the same penetration but widely different frequencies?

CEPs could show the same patterns as brand sales metrics, which is a Double Jeopardy (DJ) pattern. In a DJ pattern, (purchase) frequency varies in-line with brand penetration; smaller brands have far fewer buyers than larger brands, who buy slightly less often than buyers of larger brands (see Report 26: Double jeopardy revisited again). Because of this relationship, penetration is the primary growth lever for brands.

However, CEPs reflect buyer’s lives, and so frequency could be unconstrained by penetration, which means Double Jeopardy patterns may not be evident. Two CEPs could have the same overall incidence, yet one may have high penetration and low frequency, while the other may have low penetration and high frequency. If such instances are typical, then the selection of CEPs requires quantification across category buying occasions rather than just category buyers.

We therefore address the following questions:

• Are CEPs that are encountered by more (or fewer) category buyers also encountered more (or less) often by those category buyers?

• How often do we see deviations to this pattern? Are there any common characteristics about the CEPs that are exceptions?

Research Method

Data

The data set comprised of the aggregated results of commercial research studies carried out by the Institute. The overall data set used for this research covers 2,115 CEPs across 64 categories (including CPG, B2B, service and durable categories) over 12 countries.

Analysis approach

As part of the prioritisation stage of an Institute CEP study, each of the CEPs are converted into specific scenarios or events. For example, in the context of social media, the CEP ‘For someone who wants to keep up with news/current events’ is converted into the event/scenario ‘Used social media to keep up with news/current events’. This links the CEP with each category buyers’ actual past behaviour.

Respondents are asked if they had encountered that event/scenario in a past time period; and if so, then how many times they had encountered that CEP in the same period. Data captured is then used to calculate CEP penetration and frequency. We report these as:

• CEP penetration = n category buyers encountering CEP / total sample
• CEP frequency = mean frequency of category buyers encountering the CEP

The process is modified for categories where discreet events are harder to capture and/or the past is not a good indicator of future demand. In such cases, we use the Juster Probability scale (Juster, 1966; see also Romaniuk, 2022 for use in CEP context):

• (Probability of) CEP penetration = % category buyers who have a 6 or more chance of encountering the CEP in the future time period
• (Probability of) CEP frequency = average probability among of those with a 1 or more probability

Results

We use the appropriate statistical techniques to quantify the relationship between CEP penetration and frequency¹. Overall, the results show strong positive correlation between CEP penetration and frequency. Some categories showed very high correlations (e.g., Personal Banking in Figure 1), while others had weaker correlations (e.g., Spirits in Figure 1).

However, the categories with low correlations are simply due to very little variation in CEP frequency². These categories are therefore an extreme form of Double Jeopardy where the frequency is largely the same for all levels of penetration, and CEP penetration is even more important for prioritisation as it is the only source of variation.

While Double Jeopardy relationship extends across all category types examined, it was found to be stronger in B2B and services, and slightly weaker in CPG and Durables (see Figure 2). The difference is simply due to different numbers of data sets that have low correlation due to lack of variation in frequency.

Figure 1: Example of excellent fit (Spearman’s Rho = 91%) and poor fit due to lack of variation in frequency (Spearman’s Rho = 6%)

Figure 2: Average relationship between CEP penetration and CEP frequency by category type

And the CEP deviations?

We classified a CEP as having a deviation in frequency if it scored two or more standard deviations from the category average CEP frequency³. On this criterion, only 5% of CEPs deviated significantly (n=108). Therefore in 95% of cases, high CEP frequency aligns with high CEP penetration.

We can classify deviations into two groups: ‘Specialist’ and ‘Special Occasion’ CEPs (see Table 1). Including these types of CEPs in broad, wide-reaching marketing communications is likely to result in reaching a lot of category buyers for whom the message has no relevance either at that time or ever, and so a under-utilised opportunity.

Table 1: Incidence and reasons for CEP deviations

Summary and Key Implications

Category buyers’ CEP behaviour largely follows a DJ pattern: CEPs used by more category buyers are also used more often by those category buyers. This means that CEP penetration should be the first consideration when deciding which CEPs to prioritise in messages. The higher the penetration, the more valuable the CEP to the brand. It will generally be more beneficial in the long run to focus on higher penetration CEPs, which have frequency in line with their size, as the higher frequency means a greater chance that message receipt will align with category buying.

Dealing with ‘Special Occasion’ CEPs

These are CEPs that are lower value than their penetration would suggest. While the high penetration of these CEPs means that buyers will be easier to find, these CEPs are drawn on very infrequently, and so the challenge is timing. Some of these CEPs can be predicted due to seasonality (e.g., Christmas, vacations) and so can have limited messaging runs at that time. Others, such as breaking down or getting sick, might need to be linked to an adjacent behaviour to reach the right people at the right time – for example, buying a plane ticket means you might need travel insurance.

However, if it is a desirable message, the reach of marketing communications for CEPs is even more important than frequency of messaging. People only evoke a few brands when making purchase decisions, and brands with existing mental availability have an advantage. This means that these CEPs are best used in high reach media but at low spend levels (i.e., spread out to avoid paying for close by exposures), and only sparingly in media where frequency caps are hard to implement or unreliable. 

Low penetration CEPs

This DJ pattern also means that we can use CEP penetration to identify less valuable CEPs, that are rarely encountered by category buyers. This can be used to cull messaging options and make narrow the decision making field. Knowing what is less useful to message, is also valuable information.

The exceptions to this would be low penetration CEPs that are:

• Getting you additional buyers you would not get otherwise, and so providing incremental penetration (e.g., offering a dairy free option for vegans).
• Linked to a growing subcategory, in which case there can be an advantage in getting in early, but being careful not to over invest in case the growth predictions are incorrect. 
• Opportunities to make a great deal of profit per occasion, for example for people paint their entire house is a rare, but potentially very profitable context if you are a paint brand. 
• Specialist CEPs which are low penetration CEPs that are more valuable because they have an abnormally high frequency of occurrence.

However, the challenge with these low penetration CEPs is reaching relevant people. If there are media or moments where these people congregate more readily, then it might be cost effective to message in these more targeted contexts. However trying to ‘catch them in the wild’ is likely to lead to the message reaching many category buyers for whom it has no relevance and therefore wasting opportunities. If you do want to message these in a mass media, then this is best kept in low rotation as there is always an opportunity cost when you promote a less commonly encountered CEP instead of a more commonly encountered CEP.

______________________________________________________________________________

¹ Due to the widely differing scales, we used Spearman’s Rho, which is based on rank order, as the correlation co-efficient.

² The median range of CEP frequency was 3.4 (IQR = 1.9-5.2).

³ This decision was based on visual inspection of the data, which revealed that the convention of three std deviations would result in under-classification and miss CEPs that looked notably different from the general pattern. However, using only one std deviation difference led to over-classification of random error. We calculated a line of best fit, and used this to determine the expected encounter rate based on the penetration. 

______________________________________________________________________________