Last year, I published a piece on how marketers can use data better. I spoke to many different data experts and data-focused marketers across various industries about utilising the power of data in a smart way, which uncovered advice about methodologies, deeper analysis, and much more.
As the data-centric approach powers ahead, I wanted to look at common marketing misconceptions as highlighted by data and its practitioners. I spoke to 19 data experts, who each had their own views on where marketers were getting it wrong. A huge thanks to all the contributors!
1. Dr. Stylianos Kampakis
A common problem I’ve met in marketing and other domains, is the inability of practitioners to gauge the effectiveness of different interventions. Data can change that, but where marketing is lacking is incorporating the correct methodologies in order to do that. It looks like what many marketers are still doing is just looking at graphs and guessing, instead of using a concrete statistical methodology.
2. Roger Huang
A common marketing misconception is that distribution matters more than the underlying content. Our data has shown that wrong a hundred times over. Quality content always wins out over the number of people that see it.
3. Carla Gentry
data scientist at Talent Analytics.
I always see everyone jump on buzzwords like “Big Data” – YOU DON’T NEED BIG DATA TO GLEAN INSIGHTS ABOUT YOUR MARKETING EFFORTS OR YOUR COMPANY. Small data offers insight and is much easier to work with, besides only a small subset of businesses have BIG DATA. The rest just have a lot of data. Make sure when you collect data that fields are mandatory when possible AND binary (YES = 1, NO = 0) for correlations and other testing. When asking a customer to add their own remarks, you can use semantic analytics as well as sentiment analysis for additionally info, but even that is not BIG DATA.
4. Lissa Hyacinth
Data Scientist at Forward3D
It’s common for clients to have an additional step on conversion paths, such as branded paid search or affiliates & coupons, without a clear methodology in measuring their added impact on customer activity. For branded paid search, it’s not uncommon to test targeted reductions in spend, and see no significant decrease in revenue or conversion rate – without the branded advert, customers just clicked on the organic result.
5. Steve Jackson
CEO of Quru, a Helsinki-based data analytics and digital marketing agency
Data management platforms or DMPs are often discussed as if they are the “holy grail”. They will allow you to in the words of one of my favourite vendors of such tech give you a “360 degree view of the customer”. You might have to pay the best part of 200 grand to get one and then you’ll need expertise, training and a lot of tweaking so that if you’re lucky, a year later you might be doing what you planned. The problem with this is 70-80% of businesses don’t need a DMP. In many cases you could use a standard ad serving platform like Doubleclick for advertisers and GA360.
When most companies are buying DMPs to do the same thing as creating a lookalike audience from 1st party to 3rd party data sources it’s simply not required. Some companies might need a DMP but the vast majority just need to be able to use analytics tools to automate decision making when serving ads and building ad strategies. It’s quite often a misconception that you need the latest and greatest tools just because there is a buzz around them. DMPs have their place but in my humble opinion most companies don’t need the hassle.
6. Thanassis Spyrou
Big data expert at Citrix
In my opinion, actually, there is no misconception whatsoever, yet different mindsets. Marketers tend to be more strategic while data scientists seem to be problem solvers. Typical argumentation and defensive behavior typically lead to mistakes and unsuccessful projects.
My advice to marketers could be “don’t separate data scientist from the rest of marketing team and explain to them the marketing problems in depth”. This tactic can lead to huge payoffs. A data scientist can help you predict customer behavior, fine tune personalisation, make the messages more targeted to interested individuals. Always keep in mind that Data Science functions at a lower base compare to business because of problems/solutions complexity.
7. Sai Jayakumar
Lead data scientist, Leanplum
One of the things we have seen and called out in one of our earlier data science reports is the unbalanced emphasis placed on user acquisition vis-à-vis user retention. On average, apps spend around $2 acquiring each user. But by day 90, apps only retain an average of 1.7 percent of users.
Given the typical user retention rates we have seen across apps, in order to break even, apps would be well-advised to focus more of their budget on retaining users through targeted push campaigns. We have seen this approach create more successful apps.
8. Angeliki Romanou
Big data science engineer, Beat.
From the “what’s your gut telling you” to the era of data-driven decisions, there has been an enormous technology change for the marketers in the last decades. Along with this change, came also misconceptions about data usefulness, data-driven marketing and overall data science. In my opinion, the most common misconception that often marketers make is their extensive reliability and overuse of the new tools and systems that they have, neglecting the fundamentals and biases of information extraction.
A marketer’s goal is to understand customer behavior in order to know what to promote. The problem of mapping this behavior and ultimately influence future purchases is complicated due to the nature of consumer behavior itself and behavioral economics. If we take as an example the case of product recommendations, when we want to predict the acceptance/rating of a product from a user, we face multiple biases such as cold start, user volatility, user identity etc. Thus, the idea of “plug and play” systems that can be used with little or no error is often a misconception for both marketers and businessmen. Definitely, this fault lies also to misinformation and false promises of Data Scientists in order to promote their software. Exceptional marketers not only know to use various tools and methodologies in order to optimize their work, but also understand the pitfalls and assumptions of statistics and predictive modelling.
9. Giannis Zaoudis
Founder and CTO, Pollfish
Many marketers have falsely assumed about their user personas, which is something we are working extensively using advanced machine learning techniques. Along with survey data we provide those personas and the ability to analyse each user-segment separately to find the personas of power-users.
10. Kirsty Brice
Director EMEA marketing, 4C Insights
1. You already know which messages are going to resonate with your audience: our data has proved time and time again that brands have broader audiences than they think. Using our brand affinity tools, we have found that audiences no one could guess at can perform incredibly well. Mastercard and La La Land are a good combination; Star Wars: Rogue One fans enjoy Honest Tea; and that Playskool remained the most popular kids’ toy over the Christmas period – at least according to social media users engaging with Santa Claus. These findings confirm that there is a big audience outside of each brand’s traditional one, which is ripe for the taking for any firm that is willing to make the creative jump.
2. External triggers don’t affect the effectiveness of campaigns: consumers may access social media on connected devices at all times, but the weather provides a huge incentive to purchase for most of us. Rain, hail and snow should all be taken into account when planning a campaign, and quality real-time platforms provide ways to effectively leverage these external triggers to activate a message. This also applies to sports live broadcast: when a fan’s team scores a goal or a star player gets injured, supporters turn to social media to comment on the action. This provides a perfect opportunity for marketers to activate campaigns in real-time, using pre-set automation tools that push the right creative to the right audience when it is most emotional, while leveraging the reach of TV.
3. Once your social campaign is running, you have to wait to see the results: automated analytics tools for social media enable marketers to measure the return on investment of their spend on each creative and across all targeted platforms. Collating our Social Ads Report for Q4 2016, we found that Instagram was particularly popular for Home and Garden, while Consumer Packaged Goods have grown tremendously on Pinterest. This data can help advertisers better understand which channels is best suited to their product offering, target audience and when to drop a creative that is not reaching the right people.
11. Ben Corrigan
One of the main reasons we built Pouch is because we see a rather archaic and ‘scatter-gun’ approach to some marketers’ use of voucher codes in the performance marketing industry. Retailers will often create generic voucher codes (such as 20% off everything) which are available to almost any consumer and are thus un-targeted. Our data and experience shows that it is more profitable to tier voucher codes based on commercially important interests such as traffic source, gender, visit history, products (and product margins), time of the day, device, geo-location etc, and then targeted more refined and specific user groups to maximise impact.
12. Henrik Nordmark
Head of Data Science, Profusion
First of all, it would be unfair to lump all marketers into one basket. I have met marketers who are very data savvy and data-driven. They would do exactly what a data scientist does, they would let the data chips fall where they may and then base their decisions on the data without trying to introduce any preconceptions they may have had as to what to expect. What I have witnessed with less data-driven marketers is their unwillingness to trust the data if it does not match their own personal intuitions. It is tricky to introduce new ways of doing things very tactfully so that your non-data-driven marketers do not feel threatened.
13. Martin Adams
Co-founder and CEO, Codec
One of the key things marketers assume is that they have a rich enough understanding of their existing audience and that they can learn more and more about their audience from the audiences’ interactions with the brands’ branded content. That’s a mistake, because if you make content that is about your product, your category or your competitors, then you are really testing only a very, very micropart of that audiences’ interests. Moreover, you are only getting feedback on those areas. So if you are Oral B and you talk about toothbrushes, those people will give you some feedback based on the content about toothbrushes, but it is a very narrow picture of who they are.
However, if you invest in the right data and the right algorithms, you can understand audiences broadly from their interactions with millions and billions of bits of third party content, far, far outside of your existing product or category, and that is how you get a very deep or rich picture, and often a very surprising range of insights into what makes a particular audience tick. This puts marketers in a very powerful position to participate, contribute and lead in topics and conversations that matter to the audience in question that previously they would not have even had on their radar.
14. Joe Griston
Chief of Talent + HR, Roborace
“Metric X has gone up around the same time I did Y – that was me, right?” Sometimes, but usually not.
15. Mandy Menaker
Head of PR & Brand Development at Shapr
No matter if your budget is $10 or $10,000, try to measure success for every dollar you spend. Every dollar you spend should be considered an experiment. First, decide on your goal. Depending on your product, that may be sales, newsletter signups or virality of a campaign. Next, determine how you can measure that experiment, using google analytics, tracking codes, social shares, or promo codes to assess performance. Finally, review your campaigns regularly, to assess how effective each channel is for hitting your goal. These metrics will help you get smarter about generating a return on every dollar spent!
16. Andreas Voniatis
Full stack digital data scientist, Artios
You need big data:
Big data is technically anything that can’t fit or be processed onto a desktop computer which is about 5 terabytes. The marketers I’ve worked and spoken with always think that machine learning (ML) is some big data exercise and doesn’t apply to theirs or their client’s business. Not quite. Yes it is true you need enough data to form a decent sample to make ML useful, but nowhere near the imagined millions and billions of records to make predictions possible. In fact sometimes having too much data can ruin the whole predictive modelling exercise!
Amazon AWS or Azure ML and you’re done!
There’s a common perception that data science is just a matter of uploading your dataset into the cloud and letting either Amazon or Azure perform ML on your dataset. If only that were true, a professional data scientist will follow the preliminary steps of:
- Auditing data requirements and sources
- Statistical analysis of data
- Feature engineering
And that’s before we get to building predictive models and designing ways to display the data!
The data scientist knows it all:
No they don’t! The data scientist is not always the final answer. For marketing data science to be effective, the data scientist really needs to be a marketing scientist i.e. a data scientist with a decent amount of experience in marketing. I guarantee that if a marketing agency hired a data scientist, the first question a data scientist will ask is ‘where’s the data’?
At the graduate level this is acceptable, at any senior it isn’t! The experience a marketing data scientist will know what they’re looking at and thus know whether the model ‘looks right’. They will also be able to come up with new ideas for improving the models and forming new predictive ideas to engineer and test.
The more marketing experience you have as a data scientist, the more useful you are in a marketing context. This could be overcome by learning a few things like the process of copywriting, graphic design, web page design, SEO, PPC etc. knowledge of those areas means that the data scientist will have empathy to make those models more accurate and predictive.
In an ideal world, you’d have three data scientists, each with particular strengths in software engineering, statistical modelling and marketing, but they all must have knowledge in all three.
Search engines aren’t predictable:
Often I hear things like ‘search engines aren’t predictable’. It’s quite the opposite in fact. Yes, SEO and digital marketing have creative elements to them. But many things including search engines can be predicted. Especially search engines as they are systematic, and anything systematic which is consistent in its results, is predictable. It’s simply a matter of getting the data and doing the science to work out what’s working and not working.
17. Sophie Hudson
Head of Marketing, TalentPool
Not all users equal. We used to draw a line between our user acquisition analysis and our user value analysis. Our Marketing team’s objective was to obtain as many users as they could at the lowest CPA possible. Meanwhile our Product team’s objective was to optimise these users’ activity in order to increase our applications per role.
In our quarterly process review this summer, we discovered that our users who signed up with a desired start date set more than 3 months into the future rarely ended up actually using our platform – they only had a 31% chance of applying to a role. In contrast, our users who registered with a start date set within the first 3 months of them signing up had a 75% chance of becoming active and applying for a role through our platform.
With this knowledge, the Marketing team were able to adjust their acquisition targets to acquire the right, active users. Since then we have seen significant improvement in our user response and engagement rate, as well as a better ROI on our marketing spend.
18. Daniel Callis
Technical Search Consultant, StrategiQ
One of the most common misconceptions I’ve seen from data in my profession is comparisons of traffic and conversion performance of websites in Google Analytics over time.
So many industries have seasonal spikes and dips in search interest, caused by everything from the weather and the time of year or PR campaigns and news interest.
The best way around it is annotate the dates of as many external and internal factors in your Analytics if possible – that way you’ll be able to potentially account for any changes in traffic due to seasonality or trends much easier. Claudia Higgins touched on this earlier this year in a BrightonSEO talk.
If you want to be really smart about it, and you know the factors that potentially influence your business, you can use Google Tag Manager custom dimensions to automate the process somewhat and pull in relevant data with visits. We’ve just started using custom dimensions to pull weather data from each user’s location into Google Analytics for a travel client to see how it affects traffic and leads.
Another cool example of custom dimensions I’ve seen is pulling in national holidays for the country the user is based. In a former job we had a B2B client panic they’d lost a large chunk their Chinese-based traffic at the end of January and three of us spent an hour trying to figure out why until finally realising it was Chinese New Year and everyone was off work! Having that kind of data back then would have saved a lot of time and effort.
TL;DR – There’s no such thing as too many annotations in Google Analytics, and always consider external factors.
19. Pouya Yousefi
Data Science Developer, Machine Learning Engineer, and currently Senior Manager at IQVIA
There is a misconception that advanced analytics is scary or unattainable. Our experience is that good marketing starts with in-depth knowledge of who your customers are and what they want.
You need to ask the right questions. Are our best customers the ones who are purchasing the most online products? Perhaps not. Previous research shows that stores with a bricks and mortar presence might have shoppers who purchase many products online only to return them in the store.
Advanced analytics makes good marketing more possible than ever before. Consumers generate a plethora of data points click trails, transactions, social network interactions, and so on. Successful marketers will leverage all of this data to deliver successful campaigns.