Various methods exist to improve CATI Fieldwork with regard to Interviewer training, Dialer, incentives etc. Often overlooked is the actual sampling: whether it is a business sample, consumer sample, client sample or acquired sample. In this article, we will look at various options that exist to improve fieldwork efficiency by developing a sampling strategy. This is a 3 part series which explores B2B Sample, RDD sample, and Lifestyle sample. The article uses real cases as an example to derive lessons learned and best practices.

B2B Sample

In many cases, clients approach us with a specific SIC code range, employee ranges, and a sample is purchased and loaded into the dialer. In some cases, a few days later, a client would come back and explain that the sample does not meet the requirements. In our case, we would then usually go through the questionnaire and analyze the requirements. In many cases, however, time and money can be saved by looking into the questionnaire and the quotas beforehand: What are the requirements, who is the contact person we need to speak to, what are other boundary conditions? When explaining this to a sample provider, a sampling strategy can be developed prior to the actual fieldwork. Whether it is a small or large scale project, a sampling strategy can be vital for the success of a project.

1. Pulsing

A  fraction of businesses die every year, changes their phone number, moves addresses thus it is important to utilize a regularly updated database.  Even though business databases are updated frequently, we encounter 10-20% of phone numbers not working. Before starting a project it might be worth pulsing these records to remove these nonfunctioning numbers so that no time is lost during fieldwork. Furthermore, it allows better judgment on whether a specific number of interviews can be reached.

2. Enrichment

Going a step further from pulsing, we could enrich the disconnected phone numbers with verified working phone numbers. Additionally, it is possible to add other valuable information like employee count size, location, revenue or contact person. By doing so, we can further narrow down the target market or specify stratifications of our sample so that during the end of fieldwork it is possible whether the target record is still required for any of the quotas.

In the current age, we have access to a plethora of sources of secondary data as well as databases from private or public sources that can be accessed

3. Contacts

Contact names are essential for bypassing a gatekeeper. However, many databases contain only the Managing Director or CEO of a company. The chances that a good gatekeeper will pass you on to the CEO are fairly small. In many cases having a few additional contacts can be beneficial.

Imagine you ask the gatekeeper to connect you to accounts payable – I am sure you will be passed on. From there it is an easy step to ask to be connected to the correct contact person.

Having outdated contact people in the data is a common problem. As much as databases are kept up-to-date, changes frequently occur in our fast-paced business environment.

However, instead of neglecting contact people because of outdated information, the info can be used to connect to the correct follow-up person.

It is essential to have multiple contacts – in most B2B research in CATI research, several attempts are required in order to complete an interview or book an appointment for an interview.

4. Employee Size & Revenue

In many cases, the target audience is not just described by means of an industry but also by employee size or revenue band (minimum, range or maximum).

These extra filters can have a significant effect on the available data and make fieldwork not feasible due to the fact that the desired targets cannot be reached. In the following example we explain how employee size and turnover can be used to increase the available sample frame:

Imagine you would like to survey manufacturing companies in Belgium and the Netherlands with revenue of EUR +10M. However, it seems we cannot get sufficient records from data sources making it impossible to reach the desired targets with the available sources. However, we can use an estimate for the required employment size. From which we then calculate around EUR 100.000 in turnover per employee, thus, we can work with an estimate of 100 employees or more. Many databases will contain a lot more information with regard to employee sizes rather than revenue. Note, this method is of course not completely accurate and will have an influence on the incidence rate. We can improve it by increasing the employee size level or decrease it by lowering it (though increasing the sampling frame).

Note that the ratio between turnover and employee size will vary from country to country, for the case of Belgium and the Netherlands, these will be similar.

5. Fine-tuning your selection

By means of sampling a specific audience, we try to create a dataset which has characteristics similar to the desired audience. In this section, we will look at some cases from the past and see how sampling is approached.

A sampling of VETS: Narrowing down the target selection

Let us take the example of animal veterinarians (we do not make use of veterinarians that work with small pets but larger animals such as horses, cattle and so on). We could start off with defining a category or a SIC code and draw the sample from there. Nevertheless, to further increase efficiency we could add some extra steps:

•       Remove all sample that contains the name “pet” or anything that points to smaller animals.
•       Remove all VETs from urban areas and focus on remote rural areas.
•       Prefer records which contain names like horse or cattle and remove Veterinarians that contain dog or cat care.

Even though the sample will still contain Veterinarians which take care of smaller animals, nevertheless, we have managed to increase the incidence rate significantly.

6. Sample Sources

Selecting sample sources does not only have an influence on the quality of the sample but also on the response rate. The most common databases out there require a lot of work and time to keep the records updated. This is costly and can only be covered by sufficient sales of these profiled records. How often do you think Fortune 500 companies, companies with +500 M revenue or IT decision makers of TOP-X companies of a country are purchased? While quality might be good, these people might have been contacted way too often for the purpose of market research but also for sales.

In many cases it might, therefore, be handy to blend data: Think of a government database, blended with Google information and maybe LinkedIn contacts. While upfront it may seem to be more work than a standard database.

7. A/B Testing

Under the section of “Fine-tuning your selection” we have discussed various options for sampling Veterinarians – for larger studies, it is essential to identify the most effective sampling strategy (= highest incidence rate) for different methods of sampling. Therefore, it is recommended to set quotas on the different sample sources. The same holds true for different sample providers. Different reasoning can be used for sampling an audience, nevertheless, only when you measure it, you can manage it and increase your incidence rate.

“Measure it, to manage it!”

After completing about 10% of fieldwork, take a moment to compare the different sample sources together with refusal rates, IR and other important variables. Based on this comparison, you can make a decision with which sample source to continue using for the remainder of your fieldwork.

Sample suppliers should not charge an extra setup fee or high minimum order if you explain to them your case and want advice for your sampling strategy needed for a successful fieldwork period.

Conclusion

Different methods can be used to achieve higher field productivity. It is important that a sampling strategy is developed which exceeds the standard “Some SIC codes + employees ranges).