How Habits Influence New Product Demand

Table of Contents

• The Persistence of Past Behaviors in Product Forecasting
• How Past Behavior Shapes Future Behavior
• How Big a Role Does Habit Play in Predicting Future Behavior?
• Habits in the World of Healthcare
• Balancing Intention, Habit & Non-Habitual Past Behavior
• Take-Home Points

The Persistence of Past Behaviors in Product Forecasting

If you have been involved in new product forecasting or primary research for any length of time, I will wager you’ve encountered something like the following situation.  You are on a launch team that is bringing a truly novel, useful effective new therapy to an underserved disease population.  The current therapy alternatives are dominated by one product – a medicine that offers some level of effectiveness but a host of “baggage” characteristics.  This should be a commercial slam-dunk for your product.  The left side of Figure 1 conveys a hypothetical model of how this market might evolve in the minds of marketers, management and investors.  It’s a rosy picture.  When the product comes to market, there is much clamor and enthusiasm.  And of course, it encounters the usual early life-cycle access headaches.  Still, as time goes on, share is growing and your drug is trending toward dominance.  But the weird thing is that the old competitor with all the problems is showing remarkable tenacity.  Early on, its share loss accelerates, but then levels off even while your own product appears to be peaking.  The actual diffusion of the product ends up more like the right side of Figure 1.  This makes no sense!  Your product is better.  Physicians and patients love it.  So why are they still using the old drug?  A big part of the answer is habit.

Figure 1.  Hypothetical but Realistic Illustration of Share Capture & Attenuation Over Time

How Past Behavior Shapes Future Behavior

Every once in a while, we work with clients who ask specifically about habit formation in the context of demand research, or other PMR studies, but it is relatively rare.  And I suspect that the reason is because this topic once got very little “air time” in the academic sphere.  But since roughly the turn of the millennium, interest in the role of habits and past behavior has increased among behavioral researchers in healthcare, along with a growing recognition of the relative importance of these dynamics.  As we will see, they have a lot of control over the behaviors that we tend to care about in marketing.  In the scientific study of behavior prediction, habit (and past behavior generally) is normally among the most potent statistical factors.¹  For this article, we use the term habit in a more formal psychological sense to refer to repeated, reinforced patterns of behavior that are activated by situational cues with minimal active consideration.^2,3 The interaction between behaviors caused by habit and behaviors that are driven by careful, willful consideration is worth understanding.  Often a single repeated behavior can be under the influence of both, and this will be important when we get into discussions about demand research. 

Here are the basics of how this works.

• As a rule, novel behaviors first emerge through some kind of evaluation and intention-formation process.  In some cases, that process may be perfunctory and in other cases may involve substantial deliberation, but in virtually all cases that matter to life science marketing, our customers’ behaviors first occur based on active cognition, often involving effort and analytical thinking.  Let’s imagine the first instance of a behavior, such as using a newly-approved drug.⁴  Along with coming to a reasonable understanding of what the product can do, the physician decision-maker might consider factors such as the age and frailty of the patient, genetic markers, lab markers, patient social networks and lifestyle, the relative value of alternative therapies, etc.  The loose shorthand term, “System 2” thinking is a fair approximation of what’s going on in this case. 
• Over time, two important processes emerge.
  • First, we experience the downstream consequences of our decisions – specifically rewarding and punishing outcomes.  If you are a physician, this might mean getting blowback from network pharmacist for using an expensive drug.  It might be comments from your staff about how easy or hard it is to administer.  It might be a patient thanking you for the improvement in their condition.  Some of these consequences will reinforce the behavior and some may tend to extinguish it. 
  • Second, if the behavior persists, we begin to associate it with specific cues that co-occur in the environment.  For example, in routine meetings with patients, a physician may eventually come to use the new drug when patients simultaneously present with symptoms A and B.  These cue associations eventually take on the important property of serving as cognitive heuristics.  Heuristics are often maligned in popular literature because they are associated with occasional cognitive errors (as popularized by the late Daniel Kahneman and other behavioral economics luminaries).  However, in the daily life of humans, heuristics are nearly always productive because they sharply reduce cognitive burden and increase the speed of decision making without much loss in decision accuracy.⁵
• Over time, the above dynamics will often turn into the environment-cue-behavior patterns that psychologists refer to when they use the word habit.  We are now squarely in “System 1” territory, where the behavior arises fairly effortlessly and with little active thought whenever the correct combination of cues appears in that setting.

And so, a habit is formed. 

With habits now defined, we should also note that many of our past behaviors and decisions are not mediated by habits.  Instead, they were primarily driven by our intentions, careful thought, and situation-specific evaluation.  For the remainder of this article, I’m going to simply use the term past behavior to delineate the role of non-habitual past behaviors in the prediction of future behaviors.

But what do habits and past behaviors have to do with demand for new products?  At the most basic level, a demand study is an exploration of future behaviors relating to one or more new products.  Since habits and past behaviors are strong predictors of future behaviors, we would naturally want to consider them in our models.  When thinking about demand studies specifically, I find it is helpful to think of habits acting on future behaviors in two ways – which I refer to as antagonistic and agonistic.

• Antagonistic:  Habits can decrease the likelihood that people will engage in novel or intended behaviors.  For example:
  • A patient believes that it would be best to be more adherent to a regimen of exercise but finds themself devoting their spare time to social media.
  • A physician expresses interest in using a newly-approved product, but when interacting with a candidate during an office visit, she defaults to her current preferred product.
• Agonistic:  Habits can amplify product use when product selection is repeated, rewarded (in any number of ways) and gradually becomes associated with specific situations and contextual markers.  The example of the physician honing in on Symptoms A and B above illustrates this type of action.   

For marketers and forecasters, this means habits are either your nemesis or your ally.  Most of the time, in the context of new product demand, our main concern is with antagonistic habits – those that make customers persist in old, familiar patterns rather than new ones, but which one matters the most will have a lot to do with how your brand compares to therapies that are already available.  However, there are plenty of situations where simple intention-based demand may underestimate future utilization because of the potential for use of a particular product to become habitual.  Since demand studies are models of future behavior, the effects of habit and past behavior are crucial contributors to model accuracy.  If we don’t pay attention to them, we may find ourselves with a demand study that presents a narrative similar to the “Expected” version of the scenario in Figure 1 (left side), when it ought to produce data the look more like the “Observed” version (right side).     

Some readers may wonder how we can emphasize habit strength as a driver of demand when we have also advanced the role of intention or reasoned action as the single largest consistent predictor of future behavior in prior articles.  This is not a contradiction, because both drivers have meaningful independent predictive value – that is, they predict unique behavioral variance in studies where both are measured.^6,7 So, it’s not one or the other; it’s both.  From a cognitive perspective, I like the analogy of conceptualizing intentions to engage in new behaviors and “antagonistic” habits that sustain old behaviors as competing horses.  The faster, stronger horse wins the race – and, in this case, gets the behavioral outcome.  Not only does the analogy reflect the subjective experience of everyday life (e.g., I want to be leaner, so I intend to eat salad for dinner, but default to eating pasta by habit), but we’ll see later that it conveys useful ideas that relate to demand methodology.    

How Big a Role Does Habit Play in Predicting Future Behavior?

If intentions and habits/past behaviors are horses competing to capture future behavior, we might want to understand their relative predictive power with future behavior.  We already know from our prior work that stable intentions tend to be the single largest predictive factor in studies of behavior prediction, so while we can assume habits/past behavior are weaker predictors, their magnitude could still be meaningful.  As we mentioned, habits and past behaviors were not always prominent in behavior prediction, as researchers had long been focused on cognitions like attitude and intentions.⁸  However, a 1998 meta-analysis by Oullette and Wood illuminated just how powerful these dynamics can be.  In their synthesis of experimental work on the relationship between past behavior, intentions, and future behaviors, they found two clear predictive roles for habit and past behavior:

• Habit/past behavior predicts intentions for future behavior:  r_ES = .43 (k=33)
• Habit/past behavior also predicts future behavior directly (as in, we do more of what we have been doing):  r_ES = .39 (k=16)

Let’s unpack these.  To the first point, if we are setting intentions for our future behavior, we would be foolish to ignore our own routine patterns of behavior, or to ignore the decisions that worked well in the past.  So, what we have done affects what we plan to do.  But to the second point, habit and past behavior have predictive power even when they are measured independently of intentions.  This means they can influence behavior despite contrary intentions.  According to their analysis, habits would predict 18% and 15% of behavioral variance – quite a bit lower than what we see for intentions, but still substantial.  This widely referenced meta-analysis set off several decades of experimental investigation on this topic.  About 25 years later, in the largest, most comprehensive meta-analysis of this topic to-date (267 studies and over 107,000 human participants), Hagger et al (2023) found a meta-analytic correlation between habit and future behavior of r_ES = .34, representing about 12% of total model variance.  This is a slightly smaller effect size, but given the scope of the meta-analysis, I think it is probably the correct best-estimate for the degree of predictive power that habits will have on generalized human behavior.   

Habits in the World of Healthcare

Now let’s see if these relationships hold up when we turn to the world of health-related behavior.  For our first data source, we’ll turn to 2011 meta-analysis by Gardner et al, which focused on habit and intention as predictors patient behaviors in nutrition and physical activity.  They found a slightly larger overall effect size in this setting – r_ES = .45, or about 20% of total behavioral variance.⁹  I find this interesting because (as discussed in our post entitled, “Can People Accurately Predict Their Own Behavior?”) intention tends to have a slightly lower predictive strength in health-related versus more generalized behavior.  Perhaps habits are picking up the slack?  Gardner et al also point out that in almost all of the included studies, habit and intention displayed interaction effects relative to behavior – specifically, the intention-behavior correlation is diminished when habit strength is high, and vice versa.  This also squares with common sense:  We expect that some behaviors are easier to come under habitual control (e.g., consider habitual use of alcohol vs. habitual vigorous exercise), while others are more “habit-resistant.” 

Habits may feel like a perfectly logical thing to consider with patients, but what about treaters?  Aren’t they more rational, aloof and prone deliberate action?  In an interesting meta-analytic review of how habits impact the behavior of healthcare professionals, Potthoff et al (2019) found very similar overall effects for habit.  They analyzed studies that measured habit formation for behaviors like the use of diagnostic tests and making therapeutic recommendations for various conditions.  They found an overall meta-analytic effect for habits versus future behaviors that was nearly identical to those found by Oullette and Wood back in 1998, r_ES = .35.  This value is in the normal range for generalized human behavior, but it is smaller than the effect size associated with patient health behavior. 

To contextualize both sets of findings, let’s consider a set of experiments published in 2019 by Sheeran and Conner, two of the most widely cited authors on behavior prediction.  They found that habitual control of behavior is powerfully moderated by several types of intention-related cognition.  Their core findings were as follows:

1. The more thought-through and reasoned the intended behavior change, the more likely the participants would be to follow through, even when the behavior ran contrary to a well-established habit in a stable environment.
2. The stronger the intention (greater conviction; greater stated motivation), the more likely the behavior change would win out, again despite well-established habits in a stable environment.

Boiling this down, they found that careful thought, conviction, and well-structured beliefs tend to neutralize the obstructive effects of established habit.  Physicians are trained and incentivized to develop defensible, evidence-based intentions.  In contrast, patients may approach their health-related behaviors in more of a “set it and forget it” way.  Additionally, there is more opportunity for repetition with many patient health behaviors, which in many cases will occur more frequently and consistently compared to physician behaviors.  All this is to say, it is not hard to understand why habit would be a weaker (but not weak) predictor of physician behaviors compared with patient behaviors. 

The scientific findings discussed so far in this article are summarized in Table 1 below.  Looking at this, I hope it will afford two basic lessons:

1. Habit is a reasonably strong predictor of future behavior, albeit only about a third to half as powerful as intention, which tends to account for about 45% of everyday behavior, and about 35% of health-related behaviors.
2. When people tell you that no consistent effects are ever observed in behavioral science, don’t assume that they know what they are talking about.  These effects are impressively consistent. 

Table 1.  The Predictive Power of Habit on Future Behavior – Meta-Analytic Effects Through Time

As mentioned above, some past behaviors are non-habit-based.  They are simply normal intention-based behaviors that happened to occur in the past.  This category of past behaviors appears to have useful and distinct predictive power for anticipating future behaviors, and this may be especially true in healthcare contexts.  A 2016 meta-analysis of health-related behaviors (including a wide array of preventive, screening, and therapeutic activities) tells an interesting story. This analysis includes 237 individual studies or experimental conditions, so it was powerful enough to allow for sub-analysis for different types of behavior.¹⁰  As you review the findings in Table 2, keep in mind that that habitual behaviors are not disentangled from other past behaviors, so habit is playing a supporting role in these impressive effect sizes.  It is worth noting that some of these effects are nearly as large as those we see for deliberate intentions.  What this means is that our past intentions are nearly as predictive of future behavior as our current intentions. 

Table 2.  Past Behavior Correlations with Future Behavior in Health-Related Settings

Balancing Intention, Habit & Non-Habitual Past Behaviors

But we can further disentangle these relationships.  The above-mentioned 2023 meta-analysis by Hagger et al specifically explored the dynamics between intention, habit and non-habitual past behavior on downstream behavior.  This synthesis had sufficient power to allow for meta-analytic structural equation modeling, and their model isolated the following independent path coefficients for intention, habit and past behavior, as shown in Table 3 below.  Based on this model, the ratio of explanatory power for intention vs. habit looks fairly consistent with what we’ve seen so far (i.e., habits are about half as powerful as intentions).  What’s interesting is that non-habitual past behavior is a potent independent predictor of downstream behavior.  Not surprisingly, past behavior was also very strongly correlated with habit, underscoring the inherent relationship between these two categories of behavior.

Table 3.  Isolated Meta-Analytic Effects for Intention, Habit, and Past Behavior on Future Behavior

A final useful perspective comes to us from a longitudinal study of deliberate habit formation.  Lally et al (2010) studied the formation of habits associated with intention-based behavior change.  Participants made decisions about new behaviors (e.g., diet change, reduced drinking, etc.) they wanted to engage in, and the researchers documented the process the participants went through in undertaking the change.  Along with success/failure rates, they also looked at the degree to which the new behaviors came under habitual control.  One of their major findings was that the degree of new behavior attainment and the degree of habit formation for the new behaviors was massively idiosyncratic across the participants.  For example, the range of time needed for new behaviors to achieve automaticity (a strong habit marker) ranged from 18 days all the way to 254 days depending on the person.  This means that individual difference also plays a role in the interplay between intention, habit, past behavior and future behavior. 

Take-Home Points

Looking at this collection of accumulated data, I hope that it is now clear that the persistence of past behaviors (as in our opening scenario) is not a surprising thing.  It simply reflects the natural human tension between intended behavior change and the repetitional pull of our past.  This has been a lot of information, so let’s recap the core points. 

• Point #1:  Just like intentions, habits are strong predictors of future behavior.  Overall, they are about one-third to one-half as strong as intentions and account for 15% to 20% of behavioral variance. 
• Point #2:  Non-habitual past behavior is also independently predictive of future behavior.  When you put habitual and non-habitual past behavior together, the predictive power is basically equivalent to intentions.
• Point #3:  Habits/past behavior operate on future behavior in two ways.  First, they primarily work as impediments to the adoption of new behaviors (as in Figure 1).  Second, in some cases amplify the use of new products that eventually take on habit-like qualities.   
• Point #4:  Further, including measures of habit/past behavior increases the precision of future behavior prediction models above and beyond intention-based models, such as those described in our post entitled, “Can People Accurately Predict Their Own Behavior?”
•  Point #5:   Some of the variability in the “habit-vs-intention” tug-of-war also comes down to individual differences in humans.

For those readers interested in understanding how we can update our demand study designs to accommodate the role of habit and past behavior, we will provide detailed recommendations in a separate post.  So, check back soon!

To learn more, contact us at info@euplexus.com.

About euPlexus  

We are a team of life science insights veterans dedicated to amplifying life science marketing through evidence-based tools.  One of our core values is to bring integrated, up-to-date perspectives on marketing-relevant science to our clients and the broader industry.