Which of the following forms of direct marketing has the lowest business expenditures?

DTCA versus Physician Detailing

Although DTCA is banned in most countries, advertising targeted towards physicians – so called detailing – is generally allowed (though regulated). Indeed, physician detailing constitutes the main share of total drug marketing expenditures. This form of drug marketing includes visits by sales representatives to physicians, as well as advertising in medical journals. Because face-to-face advertising is more costly, the likely impact on prescription choices is also higher.

Like DTCA, physician detailing can, in principle, have both market-expanding and business-stealing effects. It has a market-expanding effect if it increases physicians' propensity to choose drug treatment over nonpharmacological treatments, and it has a business-stealing effect if it affects physicians' propensity to prescribe drug treatment A over drug treatment B. Like other types of advertising, detailing can reduce informational inefficiencies and improve the matches between medical conditions and drug treatments, if the informational content of this type of marketing is sufficiently high. However, it would be naive to disregard the possibility that there is also a substantial persuasive element to physician detailing. In fact, empirical studies showing that detailing reduces the price elasticity of demand suggest the existence of a significant persuasive effect.

An interesting question is whether DTCA and physician detailing are complement or substitute marketing strategies for pharmaceutical companies. Although detailing clearly affects prescription choices, empirical evidence suggests that DTCA has a larger effect on physician visits than on prescription choices, implying that DTCA mainly has a market-expanding effect. If the effect of detailing is mainly business stealing, while the effect of DTCA is mainly market expansion, this suggests that detailing and DTCA are complement strategies: More DTCA leads to a higher number of physician visits, which increases the profitability of spending resources on physician detailing to influence prescription choices.

Thus, if DTCA and physician detailing are complement strategies, an unintended side-effect of allowing DTCA is that it would lead to increased levels of physician detailing as well.

Get competent

The good news is that you can become competent and thereby become more persuasive, more trustworthy. You can read, talk to experts, research organizations that offer classes and certifications, watch online tutorials and attend seminars.

If the issue with your children is nutrition and exercise, you can learn about healthy eating and the benefits of supplements by reading books and published research and then share your knowledge with your kids. When they see that your diet has changed, you take vitamins daily and you’re going the gym several times a week, you’re chances of persuading them increase dramatically.

If you want to direct marketing for a document management solutions provider, you can learn about industry associations that offer courses designed to improve your skills. You find an organization such as AIIM (Association for Information and Image Management) and take an online certification course in electronic content management. You study and pass the exam and are now a certified electronic content management practitioner (ECM Practitioner). The job interviewer now perceives you as credible in document management.2

If you want to advance your career in human resources by volunteering on projects in which you have minimal expertise, you get competent by reading. You study manuals and books at night and during lunch breaks. You ask questions of those more experienced. You consume all the information you can about the topics of the projects you’re working on. You build your personal credibility by making sure you know your stuff. Eventually, you become vice president of human resources (case history spotlighted in Allgeier, 2009).

Conclusion

Pharmaceutical promotion, and in particular DTCA, has emerged as a marketing force in the US healthcare system. Although the debate surrounding such promotion is unlikely to be resolved anytime soon, pharmaceutical promotion should be evaluated both in terms of its costs as well as its benefits. Welfare implications can be indirectly gleaned from the extent to which such promotion affects demand, competition, and prices.

Several studies have suggested that consumer-directed advertising provides information content regarding treatment options, induces physician contact, and expands treatment, at least for certain undertreated or chronic conditions such as depression and high cholesterol. The benefits of DTCA derive from improved health due to increases in the initiation of drug therapy and adherence with drug therapy. Detecting and treating health conditions at an earlier stage, through primary care, may also be cost-effective relative to treatment at a later stage through acute care. Many health conditions are especially undertreated for disadvantaged groups; for instance, Blacks are significantly less likely to receive Rx drug treatment for high cholesterol. Thus, if DTC advertisements provide useful information and induce patients to visit their doctors, then their potential educational benefits may help reduce health-related disparities.

There is limited direct evidence on the competitive effects of pharmaceutical promotion. Though, the few studies that have been conducted seem to indicate that, if anything, promotion may be pro-competitive. Promotion aimed at providers can facilitate entry of other products in the drug class and also positively impact the number of new products entering into clinical development. There do not appear to be strong deterrent effects on generic entry. These results are consistent with the informative-view of advertising, and studies that find advertising-induced market expansion effects generally interpret these findings as welfare improving.

One of the costs of DTCA and DTPP includes potentially higher drug prices and increased use of more expensive drugs in place of equally effective lower-priced drugs. Although there is no direct study of this latter effect, Kravitz et al. (2005), in a randomized setting, find that for actors portraying adjustment disorder where antidepressants confer little or no benefits, 37% of actor-patients requesting Paxil received a prescription for the drug, compared to 10% of those who made a general drug request and none for those who made no request. Higher drug and healthcare expenditures can raise insurance premiums, increasing taxpayer and individual costs, and lead to a larger prevalence of the uninsured. Cost-ineffective treatments also impose opportunity costs for public and private resources. Here too, the evidence is limited and hampered by measurement error in drug prices. However, the few studies in this area suggest that promotion may have a small positive effect on the AWP and reduce retail pharmacy margins. There is no strong evidence that DTCA or other promotion substantially raises retail-level drug prices.

Evidence from physician surveys and a randomized control study (Kravitz et al., 2005) does suggest that there may be some DTCA-induced overuse and overtreatment, particularly in cases where there are no structured clinical guidelines for treatment. That physicians prescribe a certain drug in response to patients' request suggests a persuasive brand-switching response to DTCA in addition to a market-expansion component. Some econometric studies confirm that DTCA affects selective demand, which is often viewed as less benign relative to promotion that affects primary demand. However, these brand-specific effects generally tend to be small in magnitude. In contrast, both the US-based and international studies consistently find that the brand-switching effects are far stronger for physician-aimed promotion.

Market expansion and shifting brands for nontherapeutic reasons also raise the concern of a suboptimal patient–drug match for marginal patients, carrying the risk that the drug is prescribed inappropriately and leading to a worsening of the drug's average safety profile. As shown in David et al. (2010), increased levels of DTCA are associated with increased reporting of adverse medical events for certain conditions. Because newer drugs generally tend to be more heavily promoted, especially with consumer-directed advertisements, a popular proposal among critics of DTCA in Congress is to impose a moratorium on such advertisements during the first 2 years of a drug's launch. In response, a group of leading pharmaceutical firms (Merck, Schering-Plough, Johnson & Johnson, and Pfizer) have agreed to a voluntary 6-month moratorium on DTCA for new drugs. This would give the FDA, providers, and patients time to learn about new safety issues for new entrants. The benefits of such a proposal also need to be balanced against the need to convey information regarding new drug therapies, which may be especially important in the early stages of a drug's launch. Optimal use of DTCA may therefore require further structured guidelines.

In summary, pharmaceutical promotion has effects which can be strongly health-promoting and welfare-enhancing, but may also have adverse effects through potential overtreatment, cost-ineffective substitutions, and potential misuse. In cases where physicians can effectively perform their role as mediators, the concern about promotion-induced inappropriate use is mitigated. However, for conditions where the diagnosis or risks may be difficult to assess, there may be a need for greater oversight and investment in postmarketing surveillance by pharmaceutical firms.

2.2.7 Use case “virtual power plant”

6 Results and Evaluation

The present work is based on a bank's direct marketing data set, which is collected from different web sources of the University of California at Irvine (UCI) machine learning repository. This data set is used to implement and estimate the performance of the proposed anonymization technique in terms of preserving the data privacy for publication in the cloud. Earlier, this data set was collected and arranged by Moro and Laureano [38] and also utilized by Elsalamony [39]. The dataset (52,944 records) is associated to different marketing campaigns of a Portuguese banking institution based on phone calls. The data was collected during the period of 2008–13 [40]. Regularly, more than one contact was required with a single client in order to analyze if the product (bank term deposit) has been (or not) subscribed. Bank.csv is shown in 10% of the examples (4521), randomly selected from bank-full.csv.

The bank's direct marketing data set includes 300 data samples with 17 attributes, without any missing values [41]. The data set is comprised of two different attributes; nominal and numeral, as listed in Table 1. Three types of attributes are depicted in Table 1, including numerical (age, balance, day, duration, campaign, pdays, and previous), categorical (job, marital, education, contact, month, and poutcome), and binary, that include yes or no in their classes such as default, housing, loan, and output [39]. Fig. 5 illustrates the data before anonymizing.

Table 1. Description of Datasets

This technique is performed on an anonymized dataset after anonymizing them in Phase 1, based on the heuristic anonymization technique. This technique is achieved by selecting groups from these anonymized datasets, which are lower than K values, as shown in Table 2. After testing the groups of quasi-identifiers that are lower than K, the complement groups, with respect to the K value, are united by data relocation, as mentioned in Section 4.

Table 2. Creation of BelowKList

As mentioned before, the main aim of the hybrid generalization technique is to minimize information loss by reducing the size of groups to further populate small, equal groups of tuples. The results of the hybrid generalization technique are displayed in Tables 2–9.

Table 3. Labeling the BelowKList and the Complementary Groups by A, B, and C

Table 4. BelowKList After First Step of Data Relocation

Table 5. BelowKList New Groups After First Step of Data Relocation

Table 6. BelowKList After Second Step of Data Relocation

Table 7. BelowKList New Groups' First and Second Step of Data Relocation

Table 8. BelowKList After the Third Step of Data Relocation

Table 9. BelowKList New Groups After the Third First Step of Data Relocation

Table 2 shows the creation of BelowKList. After inputting the anonymized dataset, the groups of quasi-identifiers are selected. These groups are checked to get the list lower than the K value, which is called BelowKList. Then, it should check the percentage of reduction of BelowKList to achieve the data relocation method. If the percentage is less than 80%, the data relocation is required, as shown in Fig. 3. Otherwise, it should be achieved by anonymization.

Before achieving the data relocation n, a list of groups that complement each other with respect to the value of K will be created. So, the complementing groups are labelled with A, B, and C as listed in Table 3.

After the first step of data relocation, the complementing groups are integrated under one K value, as shown in Table 4. For example, the first row (1A) is integrated with last row (5A) under one K value, so, the first row becomes (Retired, Single, Unknown). Same is for all the remaining complementing groups, for example, the sixth row (2B) is integrated with the tenth one (4B), and the seventh and eighth (3C) are also integrated.

To create the equality groups that contain similar tuples with respect to the K value, the complementing groups are united by achieving the data relocation method under one K value, as shown in Table 5. For that, the first row (1A) is united with the last row (A5). The sixth row (2B) is united with tenth row (4B), and seventh and eighth rows (3C) are also united. Table 5 has outperformed in the preserving utility of data compared with Tables 1–3. It reduces the size of groups compared with other tables. With transformation, the number of groups becomes eight, while the number of groups for other tables is twelve.

To create more equality, groups should be achieved by another step of data relocation. So BelowKList is checked to identify if it still has rows whereby the process will proceed for the relocation method until BelowKList becomes empty. But before achieving the data relocation, it should be checked that the percentage of data relocation is not more than 10% according to the percentage of data modification in the table. If it is more than 10%, then a rollback is called, and a new K-anonymity iteration is performed. Otherwise, the second step of data relocation is shown in Table 6.

Data after application of the second step of data relocation is shown in Table 6. The remaining groups are integrated by applying data relocation according to their RowCount. More specifically, the groups with the lowest row counts are integrated with the groups with highest row counts. The first two groups (1A) that have the lowest rows are integrated with the sixth group (4A), which has the highest rows, by applying the data relocation method to create one group (second group in Table 7). Furthermore, other groups with the lowest row counts are integrated with the ones having the highest row counts. So, the third and fourth groups (1B) are integrated with the fifth group (3B) under one group (first group in Table 7). This operation is executed iteratively until the BelowKList is empty.

After achieving the second step of data relocation, the number of groups has become four, as shown in Table 7. Comparing with Table 6 and other tables, Table 7 is superior in minimizing information loss by reducing the size of the groups to half. The percentage of data relocation will be checked again before performing the third step.

If the percentage is not more than 10%, the third step of data relocation is achieved as shown in Table 8. Also, the remaining groups are integrated by applying data relocation according to their RowCount. So, the first group with the lowest row count is integrated with the second group with highest row count under one K value of the highest row (blue-collar, divorced, and tertiary). To achieve the data relocation method, the first two groups are integrated under one K value, as shown in Table 9 (the last group).

In comparing Table 9 to the other tables, we see that it has outperformed the other tables in preserving the utility of data by reducing the size of groups to be only three. This signifies that Table 9 is superior compared with the other tables, especially Table 2.

The experimental evaluation of hybrid generalization data relocation has been presented. The truthfulness and utility of the hybrid generalization data relocation technique has been evaluated using UTD Anonymization Tool Box software.

Truthfulness evaluation: Fig. 6 shows the percentage of relocation, which is calculated to measure the data truthfulness of the proposed hybrid generalization data relocation technique. The data relocation is found to affect the truthfulness. The percentage of data relocation attained a maximum of 6%, regardless of the value of K, as illustrated in the shaded blue band in Fig. 6. This signifies that the truthfulness of the data is more than 94% (dark grey). The measure of truthfulness is observed to depend on the percentage of data relocation. In other words, it is the complementary percentage of data relocation, as depicted in Fig. 7, which is considered a good indicator of the capacity of the hybrid generalization data relocation technique to maintain the data truthfulness relocation percentage.

Utility evaluation: The expression for utility has been described by Eqs. (1)–(4). The metric of utility information gained per privacy loss (IGPL) is written as [42],

(1)IGPL=IGspec /PLspec+1

(2)IG=Rc−∑c∈childpRc/RpIRc

Rx represents the set of original records having attribute values that can be generalized to x. IR is the entropy of Rx given by,

(3)IR=−∑s∈sensitiveRxsv/Rx⋅log2Rxsv/Rx

Let Ap (spec) mean the anonymity before performing spec. Ap represent the anonymity after performing spec. Privacy loss caused by spec is calculated using the expression,

(4)PL=Ap−Ac

It is found that to preserve the data utility, the relocation technique is used, but with a limited percentage. However, it is essential to tradeoff between data utility and truthfulness. The utility is measured in terms of two metrics based on the average group size. Fig. 8 displays the K-value dependent average group size as an indicator to data utility. The average group size is observed to reduce significantly using the hybrid generalization data relocation technique. Thus, the application of the hybrid generalization data relocation technique minimized the information loss. Conversely, too big of an average group size in K-anonymity enhanced the information loss. This is equivalent to an increase of the utility of the resulted anonymized data in the hybrid generalization data relocation technique, and acts as a good performance indicator.

Direct to Consumer Advertising

The United States and New Zealand are currently the only two countries that permit “Direct-to-Consumer” (DTC) advertising. DTC advertising is traditionally understood to refer to the marketing of FDCA-regulated drug and device products to patients and their families rather than HCPs. Groups such as the American Medical Association and various patient advocacy groups have argued that DTC advertising should be banned in the United States. Nonetheless, the practice continues, and the relationship of DTC advertising within the family of promotional activities must be considered when marketing a product in the United States.

DTC advertising is not considered to be “promotional labeling,” but rather “advertising” under Federal law. FDA has developed standards for such advertising and will enforce those standards when it considers advertisements already released to the public. However, under current law, they cannot require drug manufacturers to submit commercials for review or approval. Most DTC advertisers will nevertheless submit those materials voluntarily to FDA. DTC ads must satisfy specific elements when making any claim about the drug’s use. The advertisement must list at least one approved use for the product, the generic name of the drug, and perhaps most famously, a list of all known risks associated with the drug. Some exceptions to this final requirement do exist. For example, if listing all known risks would be overly time-consuming, the ad may run an “adequate provision,” which provides the most serious risks in greater detail but only a summary of known “less serious” risks. The advertisement must also have a statement advising consultation with a HCP, a toll-free telephone number for patient questions, and/or a website address that can be used by patients to obtain additional safety information.

Not all drug products are given free rein to engage in DTC advertising. Limitations that are strictly enforced include, for example, “reminder ads” (ads which mention the drug by name, but do not discuss the drug’s use), which are not permitted for any drug product that carries a “black box warning.” Drugs with boxed warnings are those that carry serious potential risks, such as death or serious injury. The restriction is considered important because patients may not have the appropriate capacity to assess those serious risks for themselves.

Enforcement actions against DTC advertising are common. They can be directed at different types of violations. Enforcement action can be taken if a manufacturer fail to satisfy the FDA requirements for specified elements in the advertisement or engages in off-label promotion. Regulators also intervene if products violate the “fair balance” requirement by failing to discuss all known risks or introducing other misleading statements when advertising a product. Typically, the FDA will first notify the drug company that the advertisement is inappropriate and must be withdrawn or corrected. If these notifications do not spur sufficient action, the FDA may seek a court-affirmed injunction against the advertisement, may seize drug product in the stream of commerce, or may seek criminal action against the company.

Sales and Marketing

Describe your overall marketing strategy and sales tactics. Will the business require advertising, direct marketing, and sales promotions? In the early stages of a business, the founder or CEO will likely be engaged in sales. Higher valued products sold to a few large customers will be a better use of top management’s time than a sales effort aimed at low-value, high-volume items.

The portion of your plan that deals with channel strategies requires a thorough understanding of distribution. How your product reaches the market is one of the most important aspects of your business plan, and your ability to effectively articulate this strategy is critical. Resist the normal temptation to cover all bases by listing every imaginable channel possibility. “We will market our product via Internet, catalogs, distributors, value-added resellers, infomercials, wholesalers, direct mail, agents, direct field sales, telemarketing, and retail outlets.” All this tells the investor is that you do not have a channel strategy.

Privacy Statements to Customers

Today, most companies issue privacy statements to assure their users of confidentiality. The Direct Marketing Association (DMA) strongly encourages their members to disclose their cookie practices and provides a model statement for which their members may use or amend. Several examples of these agreements are presented at the DMA's Web site at www.the-dma.org.

However, these privacy statements are only the companies' “voluntary best-effort” promises. They actually promise very little and have no real teeth if the promise is not kept, since no damages are available to the injured parties. An example of this was a privacy study completed by the California HealthCare Foundation (CHCF). The study reviewed 21 popular health-related sites to discern how each site handled the privacy of the information it received. Of interest to the CHCF was not the promises made, but what actually happened with the information. To its dismay, the CHCF found that most of the sites did little to protect the actual privacy of the information, despite their claims. Instead advertisers were able to obtain the names, addresses, and significant private health information of most site visitors. Details of the study and its findings can be found at www.chcf.org. For this reason, corporations, health-care organizations, insurers and other large collections of users frequently negotiate a contract, written with specific nondisclosure language and agreed-upon damages if the information is revealed.

11.3.1 Privacy Concerns and Targeted Advertising

Since the advent of the newspaper, newspapers have depended on purchases, subscriptions, and advertising for revenue. A major factor in the reduced circulation of newspapers has been a fall in advertising revenue, as advertisers shift from the relatively untargeted advertising environment offered by newspapers to the highly targeted ads available online in connection with specific search terms (as with Google) or specifically stated consumer preferences (as with Facebook). However, the new targeted ads pose their own set of analytical problems relating to stated privacy concerns. Advertisers select, on an anonymous basis, the consumers who will receive an ad impression, based on the advertiser's information on that set of consumers’ personal tastes and actions. Consumers’ “click-streams,” their click-by-click footprints as they navigate across the web, are utilized to construct detailed profiles that can be used to target advertising more accurately than ever before, and much of this data collection occurs without consumers’ knowledge or permission.

Collecting advertising data online is often argued to be effectively harmless on an individual level because it typically involves a series of actions linked only by an IP address or otherwise anonymous cookie-ID number. Advertisers cannot garner much about an individual's web surfing habits, say, from a family computer shared by four people of different ages, genders, and tastes. However, attempts by advertisers to use such information has still met with resistance from consumers due to privacy concerns. In a well-publicized survey, Turow et al. (2009) found that 66% of Americans do not want marketers to tailor advertisements to their interests—and this holds true even in situations like Facebook where consumers have voluntarily and affirmatively disclosed things that they like and do not like. This customer resistance to tailored advertising is a major problem for advertisers. Fear that users may react unfavorably because of privacy concerns has led advertisers to limit their tailoring of ads. A recent survey suggested that concerns about consumer response have led advertisers to reduce the targeting of advertising based on online behavior by 75% (Lohr, 2010).

Therefore, early work on Internet advertising suggests that when asked their preferences, consumers spurn tailored advertising. However, when we turn from consumers’ stated preferences to their empirically measured behavior, a more complex picture emerges.

Goldfarb and Tucker (2011a) investigate how consumer privacy concerns affect what kind of advertising techniques work and, more crucially, do not work. Faced with consumers who simply ignore online banner ads, the advertising industry developed two different techniques. First, advertisers improved their targeting of ads—for example, by making sure that ads matched the content that the web user appeared to be seeking. Second, advertisers developed obtrusive ad features using video and audio, including the ability to make an ad float above or take over content the user is seeking. We explore the effectiveness of these two techniques, both in isolation and in combination. To measure advertising effectiveness, the paper used a very large database of field tests for online display advertising campaigns. In isolation, these two techniques each significantly increase advertising effectiveness. However, surprisingly, attempts to combine both targeting and obtrusiveness nullify the positive effects that the two techniques have in isolation. The same obtrusive ad works far better on websites with irrelevant content than with relevant content. One potential explanation of this result is that both of these advertising techniques impinge on users’ privacy. Even weak forms of targeting rely on advertisers collecting and using more data about the user and what they are looking at online than is the case for mass-media advertising. Intentionally obtrusive advertising—ads designed to compete with the content next to it—also intrudes on users’ privacy by interrupting their online experience without permission. Privacy concerns appear to be the underlying behavioral driver in our results because the combination of obtrusiveness and targeting diminishes ad effectiveness more for privacy-sensitive items (such as healthcare and financial products) and for people who exhibit privacy-sensitive behavior while taking the survey. Acquisti and Spiekermann (2011) reinforce this finding by showing that obtrusive pop-up ads online negatively affect willingness to pay for the advertised goods. This again emphasizes that though much of the privacy debate may be focused on data-enriched ads, there are many other negative ways that advertising can intrude on users’ privacy beyond behavioral targeting. A more comprehensive approach to user privacy would therefore consider how obtrusiveness also hinders the user experience, and how that can be traded off against non-obtrusive but personally targeted ads.

White et al. (2008) show that privacy concerns are greatest when advertising is most personalized. Expanding on this theme of personalized advertising, Lambrecht and Tucker (2013) examine one of the most data-intensive and personalized forms of online advertising—dynamic retargeting.3 Dynamically retargeted ads display products that consumers have viewed previously on other external websites. The paper uses a field experiment conducted by a travel firm and finds, contrary to accepted marketing wisdom, that ads personalized using a consumer's previous product exploration are not usually effective. The only time when they become moderately effective is when a consumer's browsing history indicates that they have reached a stage in their purchase process where they are ready to buy.

Given this work, it is interesting to contemplate how users’ ability to control what ads they see affects outcomes. In particular, it would seem likely that privacy concerns may give rise to advertising-avoidance tools (Hann et al., 2008). Johnson (2013) shows that the increasing ability of firms to target their ads influences market outcomes when consumers have access to advertising-avoidance tools. Consumers may gain by absorbing more-relevant ads; two negative consequences of improved targeting are that the number and obtrusiveness of ads received may increase, and rather counter-intuitively, that improved information accuracy may lead consumers to receive ads that they say they prefer less, like ads from niche firms when they say they would prefer an ad from the mainstream firm.

ISSUE 5: MARKETING AND DELIVERY OF PRODUCTS

How products are marketed and sold is one of the most high-profile and persistent sets of ethical issues facing the bioscience industry. What forms of direct-to-consumer advertising are appropriate? What information can be given to doctors relating to off-label use of drugs? What benefits or medical education is it fair to give doctors to encourage use of a particular treatment? Participants from several companies explained the challenge their sales representatives face when competing against other companies that are using lavish inducements, such as sending doctors on expensive vacations. Said one marketing manager, “The reps will come back and say, ‘My competitors in that country take doctors out and spend money, and I know it's not legal in the United States. but it is here, and I'm not going to make my quota if I don't do this.' And we just have to say, ‘No, that's not how we do things here'.”

Industry efforts at self-regulation in this area have not proved very effective: “Generally speaking, product promotion is not a level playing-field,” said Merck's Vice President of Economic and Industrial Policy, “since not enough firms adhere to the voluntary codes of conduct and guidelines of the American Medical Association and of industry associations in the United States and Europe.”

Sciona and Interleukin were entering less-charted ethical water, attempting to build an effective and ethical business model that entailed analysis of individuals' genetic information on which to base dietary recommendations. Sciona used to market its services directly to the consumer and Interleukin is considering the direct-to-consumer approach through independent sales representatives who will sell the service from their homes (the Amway sales method). Interleukin's in-house ethics experts acknowledged that this form of sales needs to be thought through carefully.