How AI is disrupting the need for survey research

Running a business in 2022 is tough. Inflation, supply chain issues and the threat of a recession are making it challenging to determine where a business should invest.

With so many factors impacting consumers’ daily lives, it is vital for companies to stay connected to how consumers are learning about, buying and using their products - yet consumer research can be slow and expensive.

15 years ago, when companies needed to understand what their consumers wanted, there were two main approaches: (1) Run a survey (a “quant” study) or (2) Run focus groups or consumer interviews (a “qual” study).

Most companies didn’t have the capability to do this in-house, so they would hire a full-service market research supplier to run a quant or a qual study.

After spending $50k - $150k and waiting 6-10 weeks, they would get a summary report, often an overwhelming 100-200 slide Powerpoint deck.

I’m exhausted just thinking about it.

Running a study this way was so expensive and slow because it required a ton of human hours. A typical survey run in 2010 would look something like this:

1) Needs Assessment: A senior level project manager would liaise with the client, and determine their learning objectives.

2) Questionnaire Development: A senior or junior level project manager would write a questionnaire, and the senior project manager would align on the questionnaire with the client (often with significant back-and-forth).

3) Finalizing Stimuli: If the study involved showing consumers new product concepts, the project managers would work with the client to get the final stimuli in the right format. Some projects required significant effort at this stage.

4) Programming Survey: At this point, many companies outsource the programming and fieldwork. The senior or junior project manager would then kick off this study with the company handling the programming/fieldwork, share the final questionnaire and explain what the client needs. That company would then take a few days to program the study and share a link to the survey with the project manager. At that point, the project manager would quality-check the survey and send feedback to the programmer. Then the project manager would share the survey with the client and pray there weren’t many additional changes needed.

5) Fieldwork & Data Cleaning: It would then take 4-8 days on average to find people to complete the survey. At that point, someone would be responsible for removing any respondents who were speeding through the survey or answering in a way that indicated they weren’t paying attention to the question (for example - if they selected option A for every question).

6) Tabulations: The raw data would then be sent to another team responsible for manipulating the data so that the project manager could make sense of it. The first step is for the project manager to create a “Tabplan”, showing how they want the data to be presented, and then the “Tabulations” team would spend a few days creating excel tables with this data.

7) Coding: If the questionnaire included any open-ended questions, a “Coding” team would manually go through every response, and indicate the themes that each response contained. The project manager would oversee this process and often require a few back-and-forths.

8) Analytics: If any advanced analyses (Factor Analyses, Driver Analyses) were needed, another team would take the data and spend a few days to a week to do this.

9) Charting: Once the data is ready, a junior project manager would put together a Powerpoint deck with charts, graphs and tables showing the results of the study.

10) Interpretation: Once a Powerpoint deck is put together, a senior project manager would then interpret the results, summarize the findings on each slide, and write an overview explaining the actions that the client should take.

11) Delivery: At this point, the project lead would share the results with the client, and field any questions. The client would often ask for some follow-ups, which usually required more work from the tabulation and/or analytics team, more charting and interpretation.

…So yeah - this kind of thing wasn’t cheap to pull off.

The DIY Revolution

Over the last 10 years, there has been a major shift in how companies run surveys. Do-It-Yourself (DIY) survey platforms have seen major growth, giving many companies the ability to run surveys themselves.

Instead of going through all the steps above, companies can now program the surveys themselves quite easily, rely on the software to handle the fieldwork, data cleaning, tabulations and even some charting without any human involvement.

Taking people out of the picture and relying on technology significantly brings down the cost of running a survey. It’s a massive improvement in our industry.

To take advantage of this improved way of running surveys, companies need enough people with the time and capacity to execute the surveys. While some have hired external freelancers to help, most companies rely on internal employees.

This works well if you have a growing - or at least a stable team. However, tons of companies are down headcount these days. Between the Great Resignation and the Great Layoffs of 2022, a lot of companies have fewer people, but an equal or greater need to understand their consumers.

How AI is changing everything

Out of necessity, many companies realized they needed to think differently.

I believe survey-based research will always have a time and place. However, there are more and more learning objectives that can now be answered without the need for a survey at all. There are tons of ways we can use AI to answer questions that used to require surveys.

Here are a few examples:

1) Trends Research: We can learn a ton about what people want by analyzing publicly available online data (from eCommerce Ratings & Reviews, Blogs, Forums, Search Volume Data, Social Media, etc.). With a good machine learning model, you can predict upcoming trends by analyzing how much growth we see in conversations around certain topics / ingredients / etc., and how rapid that growth is.

2) Brand Equity: The same online data we use to uncover upcoming trends can be used to understand consumer perceptions of your brand and competitive brands. You can also supplement this data with micro-surveys (ie. very short surveys) that are easy and cheap to run, and can fill in the gaps that you can’t get from analyzing online data.

3) Copy Testing: There have been fascinating improvements in how to evaluate new ads using AI. There are algorithms that can evaluate whether an ad has strong brand linkage (ie. how noticeable your particular brand is) without the need for human involvement whatsoever. There are also AI-based approaches that expose consumers to ads, ask them a few open-ended questions, and then use a machine learning model to analyze the content of those responses to gauge how likely the ad is at driving interest in the product.

4) Claims Testing: Online reviews can be a great source of information for designing better product claims. There are algorithms that can be trained to learn what language consumers use when discussing certain product benefits in online reviews. After training the algorithm, you can test new product claims to see how consistent they are with the language that consumers use when discussing your category.

These are just a few examples - and this is only the beginning of the AI revolution.

The future of the consumer insights industry is very exciting. While it will require a huge shift in how companies think about understanding their consumers, there are a ton of opportunities for folks that are eager to learn about this new world.

If you are interested in learning more about the best-in-class ways to make the most of these approaches, contact us at Jumpspark.