Has data science killed the survey?

For many years the survey was the first thing marketers would reach for in their toolkit. What better way is there to gauge where you are on key purchase funnel metrics such as awareness, consideration, intention and action?

Yes, the survey has evolved over the years from face to face to telephone to online to restech platforms but asking people have they heard of a product, are they likely to buy it and what did they think about it afterwards is as old as market research itself.

That's changing now. Business insight can be derived by applying data science techniques to publicly available datasets. But is it any good?

Awareness and engagement with the brand, product or service

Social media sites such as Twitter and Instagram are really useful tools in measuring awareness about your brand and the depth of engagement with it. At a basic level, comparing the number of followers you have with the competitive set will tell you where you are in the market. Comparing your number of followers over time will indicate the success of your acquisition strategy.

For more in-depth insight you can map your network by plotting who your followers on Twitter and Instagram also follow. This map, known as network analysis, shows a brand who else is in their consideration set and the hierarchy of importance to consumers of the different brands.

You can also gauge awareness and interaction with your brand, service or product by using Google Trends. This tool is very useful because it enables you to see the relative volume of searches over a set time period. You can search for general terms about your brand and drill down to specific actions to see the strength of engagement consumers have with you.

To illustrate, we show trends in Google searches relating to the launch in the UK of TalkTV on 25th April and more specifically Piers Morgan's new interview show (Piers Morgan Uncensored) that was the centrepiece of the launch.

In the graphic, the blue line shows the volume of searches for 'TalkTV' between 22nd and 29th of April. Google records the volume of searches within a given time period as an index. The point with the highest number of searches is indexed to 100 with other values set on a scale of 1 to 100 in relation to this. So if there were half the number of searches on day 2 compared with day 1 then the score would be 50. We can see from the graphic that searches increased from 24th April reaching a peak just before the show was aired on launch day.

The red line and yellow line show volumes for two terms measuring interaction with the TV station and the programme. Searches for the term 'How to watch TalkTV' are shown in red and those for "how to watch Piers Morgan' in yellow.  Volumes for both of these lines are shown relative to the blue line. At 7pm on 25th April, searches for TalkTV were at their highest volume during the week shown. Searches for 'How to watch Piers Morgan' were at 18% of the TalkTV volume at this time and searches for 'How to watch TalkTV' were 4% of the total.

From this graph we can see that awareness and interest in the station and the programme increased immediately before it launched and that up to about one fifth were sufficiently engaged to take action to watch it - or at least intended to do so.

We can also see that although searches declined in subsequent days, a series of mini peaks occurred with people searching just before the next daily broadcast of Piers Morgan's show. Interest and engagement with TalkTV peaked with the launch and is now running at a much lower level.

In fact the sharp fall off in searches reflects what has happened to viewing figures. Industry data (BARB) estimates that 317,000 people watched the first show but this had fallen to 62,000 a few days later. It seems, therefore, that there is a good relationship between intention to watch as measured by Google Trends and actually watching as measured by BARB.

Customer experience

Views on the customer experience can come from ratings and review sites. Survey data will only tell you so much about your brand, product or service. Respondent recollection of their experience can often be hazy and lead to inaccurate results. CRM data will tell you how many customers you have, how often they buy and how much they spend. But it won't tell you the key drivers of their purchase decisions, the strengths and weaknesses of your product and how it compares with rivals. You need this insight to grow your market share.

This is where consumer views on sites such as Amazon, Trust Pilot and Trip Advisor are so important. Here, you can hear the real voice of the consumer and get vital feedback from people who have used your product or visited your venue.

At SPARK, our text analytics capability combines machine learning with human created business rules to undertake topic and sentiment analysis to identify the factors that differentiate products and those that are key drivers of purchases. Applying this capability to ratings and review sites enables us to identify the metrics  which consumers rate as important and then measure how a brand performs on these.

A recent example of how we use text analytics to assess the performance of three meal kit providers can be see here.

We also analyse followers of brands on Twitter and Instagram to map the most influential players in the market and to provide detailed insight into the likes, aspirations and motivations of followers that can be used for segmentation. We also use ratings and review sites to analyse the flows between brands - which brands people move from and who a brand loses customers to.

All of this is done without a survey.

So is the survey dead?

So, is the survey dead? No is the short answer. Survey data can offer greater detail especially around awareness, can provide a link between consideration and action and can fill the gaps in existing data.

We use survey data for four key areas: 

1. Awareness - we use both spontaneous and prompted awareness to assess depth of knowledge. 
2. Attitudes and perceptions - what do people know about you and in what use cases do you come to mind?
3. Engagement - we measure how people engage with you, how frequently and why they engage and the touch points of engagement. We also measure their future intentions. 
4. Customer experience - we measure satisfaction and advocacy.

We also use survey data to model the key drivers of engagement, assess customer satisfaction and undertake trade off analysis to give insight on the combination of features and pricing that will resonate best with your target audience. 

Our view is that there is a lot of hugely valuable insight already out there and for many uses is every bit as good as doing bespoke primary research. It's also cheaper and quicker. But there are weaknesses around detail, availability for an individual brand and linkages and this is where survey data should be used in addition.

Markerters should utilise all sources available to them and not just restrict themselves to one tool. Expanding the toolkit to combine survey research with analysis of existing data will give brands complete business intelligence.