Snowball sampling is a technique in market research used to gather responses to a survey. The survey is shared with a small pool of respondents who participate, and each of those participants shares the link with their network or pool of respondents.
Those respondents then participate and share again, and so on. The pool gradually snowballs and grows larger.
Sounds pretty great, right? Our market research company utilizes different techniques to collect survey responses for our clients - snowball sampling not being one of them.
In this blog post we'll dive into:
- The definition of snowball sampling
- Why snowball sampling in dangerous in market research
- Examples of snowball sampling effects
- Better survey sampling techniques
But first...let's talk about lasagna. Yes, lasagna.
Let’s say you wanted to launch a new Lasagna Food Truck...
A very random idea but one you feel very confident in. You’ve been cooking lasagna over the past 20 years regularly. It’s a staple in your household and something you often make, nearly once a month. If life had a calling for you, this is undoubtedly it. You are serving lasagna.
You decide one night that you are ready to take the next step. You want to get a food truck, serve lasagna to the masses in the Memphis TN market, and turn this gift into a profit train.
But you’ve realized over the past 20 years, the only people who have eaten your lasagna are your husband and two children.
Research is needed to test this concept and how good your lasagna is.
You’re ready. You develop a short one-page survey asking about taste, heartiness, portion size, and so on.
You are going to use this data to help drive a six-digit business launch, which includes the purchase of a truck, equipment, payroll for staff, and marketing. The survey will be crucial to help you understand if this investment is worth it or not.
You launch the survey and need respondents.
So you cook a large batch of lasagna and invite these respondents over for dinner. You invite your parents, uncle, two nephews, your niece, and your three best friends. It’s a lasagna rating party.
You serve everyone the dish and then hand out the survey and ask them to rate it for you.
Scores are off the chart! Your parents love it. Your extended family loves it. Even your friends love it.
You think to yourself: this is a slam dunk.
Everyone who ate the lasagna loved it, so this will undoubtedly result in the entire general public loving my food and the food truck. I am going in head first, no turning back now. The survey data confirms it.
See any problems with this?
Would any sane person believe the sample size of respondents is representative of the Memphis market area?
Although this example is drastic, we see far too many consultants and even market research firms doing this too regularly. It’s terrible research and is concerning.
As referenced earlier, snowball sampling is a technique used to gather survey responses by sharing the questionnaire with a small group of people. This group of respondents will then share the survey link to their network of friends, family, colleagues, and so on.
Then these new groups of survey respondents share the survey link again to their own personal network. The pool grows larger, like when adding snow to a snowball.
Keep in mind, all survey respondents are connected to in some fashion to the person or persons who originally started sharing, which creates sampling bias in market research.
Those respondents likely know each other through shared traits, characteristics, working relationships, family relationships, or hobbies.
It is far from a probabilistic random sample, which is often recommended in survey sampling.
When running a survey, the goal is to ensure your survey sample is representative of the population you are surveying. Snowball sampling will never accomplish this and will very likely steer your results in the wrong direction.
It is an extremely risky form of sampling.
It’s very rare top market research companies would recommend or use such a sampling approach.
If pursued, everyone reviewing the results needs to understand the inherent bias and extreme caution that should be taken when examining the results.
The data will not be representative, and as mentioned above, could very possibly steer you in the wrong direction.
Why is Snowball Sampling Dangerous in Market Research?
Birds of a feather flock together. Take Facebook, for example.
Your friends on Facebook are the conglomeration of people who have something in common with you. Likes, sports, work, family, etc. Friends often share the same attitudes, perceptions, and feelings about topics.
So if you feel strongly about a topic, answer a survey about it, and then share that survey with your friends, you are merely compounding and multiplying responses among people who feel the same as you.
That is not good for probabilities and top market research companies to implement.
Your sample will not be reliable and will be composed of friends of friends of friends. This makes it extremely difficult if you are looking to obtain exact measurements from a market.
The data will be inaccurate if you used any form of snowball sampling for the following key performance indicators (KPIs):
- Awareness
- Sources of awareness
- Perception
- Image
- Word Association
- Familiarity
- Satisfaction
- Likelihood of success
- Concept testing
- Pricing
None of these measurements would be genuinely accurate if your survey were shared organically on social media platforms such as Facebook, LinkedIn, Instagram, or Twitter.
Examples of Snowball Sampling Effects
The example below highlights a study for ABC Diner. The diner has commissioned market research to help understand its brand equity in the market.
It hears a lot from customers that the reason it is not busier is that no one in the area knows about it. It is a hidden gem.
The diner decides to hire a market research consultant to send a survey out to the community to understand:
- Awareness
- Awareness of competitors
- Perception
- What it would take to get customers to try the restaurant for the first time
The local market research consultant writes the questions and posts a link on its company page.
Six people click on the link from the research company and participate.
Three are aware, four are unaware. One of those unaware (John) decides to help a local business.
He knows the survey is for ABC Diner from just taking the survey and remembers his friend Pete went there two weeks ago. John shares the survey on his feed and tags Pete.
Pete sees the share the next day and takes the survey. On that visit two weeks ago, Pete went to the diner with his entire hockey team after a late-night game.
Pete also wants to help, so he posts the link to his hockey team page, and 25 of his friends participate. All having been to the diner and aware of the restaurant.
This results in 31 respondents being aware of the diner and only seven unaware of the diner from the initial results. You can see how this data has been skewed so drastically from this organic sharing and lack of control over the sample.
They learn from the survey response that “long wait times” was the biggest reason customers would not return.
Using these results, the diner spends another $10,000 a month in marketing to focus on boosting awareness of new fast-check in and seating process. On top of this, the restaurant also hires two new hosts to help speed up the process to capture incoming customers and speed up seating times.
They invest a ton of money in revamping their marketing and operations, which is all wasted.
Why? If they had surveyed a more random sample, they would have found that only 10% of the market is aware of the diner. The real issue with attracting customers is not the long wait times but rather just a lack of awareness of the diner itself.
By learning this, they only need to spend $5,000 a month to boost awareness of the diner and do not need any additional staff or operational changes.
The feedback they received on those issues was not representative of the community. In reality, with a truly random sample, only 1% of the population complained about wait times.
This is a theoretical and unfortunate example of how snowball sampling can steer a business in a completely wrong direction.
What is a Better Way of Sampling?
If you are looking to obtain a sample for your survey to test a new concept, awareness, familiarity, image, or any number of KPIs, there are better options for sampling.
Here are two options to consider, all of which offer more reliable forms of feedback more scientifically.
Online Survey Panel
These online survey panel companies like Drive Research offer access to millions of participants to leverage for market research.
The pool of respondents is much more random and scientific than sharing a survey with your friend list. You can reach out to an online survey company who can help you estimate a cost per response for your study if you offer the following information:
- Market area to be surveyed (DMA, ZIP Codes, states, etc.)
- Length of the interview (LOI) or survey (3-minutes, 5-minutes, 20-minutes)
- Incidence rate (what percentage of the general population will qualify for the survey)
Paid Social Media Advertisements
There is a significant difference between organic snowball sampling and paid social media advertisements. With a paid social media advertisement you publish an ad with a link to a survey to a chosen audience (geography, demographic groups, etc.)
The ad is then served to a much more random audience than what would be shared organically to your friends. The ad reaches out to social media users across the board in your market geography at random to collect feedback.
If an online survey panel cannot offer enough responses, and you must use social media, always opt for the paid route for the goodness of the market research.
Contact Our Market Research Agency in NY
Drive Research is a market research firm in New York. Our team of experts works with products, services, businesses, organizations, and brands on a variety of market research topics.
Our suite of services includes both qualitative and quantitative research, including approaches like online surveys, virtual focus groups, mobile ethnography, and phone interviews.
Questions about how we can help? Need a proposal?
- Message us on our website
- Email us at [email protected]
- Call us at 888-725-DATA
- Text us at 315-303-2040
