Things I learned at Marketing Analytics Summit 2019 in Las Vegas [Day 2]
This is a continuation of my recap of the Marketing Analytics Summit. To learn about the talks from day 1 and upcoming Marketing Analytics Summits around the world, check out the previous article.
Agenda for day 2:
“The executable data strategy” by John Lovett
Day 2 started with a keynote presentation from John Lovett, Senior Director of Data Strategy at Search Discovery. His talk was about the executable data strategy – a plan for using data with purpose that includes understanding your data, architecting your data to make it useful, and activating your data for business impact.
Organizations need a data strategy to guide them, since harnessing data in today’s business environment can be complex and confusing. There are three modern data problems: lack of data literacy, data silos, and lack of trust in data.
Lack of data literacy
The problem here is that C-levels and heads of departments often don’t understand data. According to “Gartner Trend Insight Report: Fostering Data Literacy and Information by 2020,” 80% of organizations plan to initiate deliberate competency development in the field of data literacy, acknowledging their extreme deficiency.
You can split data literacy into three buckets:
- Metrics literacy – When people know what data means, use the same terms, and understand the values behind them. This is an essential level that allows everyone on the team to speak the same language.
- Tool literacy – When people are able to self-service data needs as appropriate and are aware of the potential and tricks and limitations of the tools. This point really correlates with June’s talk on self-service analytics at Twitch.
- Conceptual literacy – When people approach and apply data with clarity and sophistication and can explain themselves using data.
To be data literate, you need to understand your data. Answering the following questions should help:
- What types of data do we have and what types do we need to answer our questions? Use an interactive data inventory, data dictionaries, and catalogs to understand where data comes from and what it consists of.
- Can users self-serve data or do they need someone to serve it to them?
- Who has access to our data and how can the data be used?
- Do users have an accurate understanding of our data, or are we speaking Greek?
- Is there a high degree of comfort in using data, or is it a struggle every time a new report needs to be built?
It’s not a surprise that most companies struggle with scattered data because their data sources don’t talk to each other:
To break these data silos, you need to define where your data comes from and how it can be connected. Then you can prioritize what should be implemented first. Here are some questions that should help you with that:
- What data connections and APIs exist? Most sources that you want to connect can be integrated via ready-made APIs or third party tools, so don’t waste time and other resources building your own integrations.
- What transformations are necessary for analysis? Usually, your data comes in different formats and at different speeds, so in order to pull it into a single report you need to process it beforehand. Think about this ahead of time and plan accordingly.
- Are there integrated processes across data sets? Think about what processes will be influenced by having this data in place.
- When is data refreshed and shared? Check how often you need to update data and who needs access to it. This should help in selecting the most suitable tools for sharing reports with the team.
Once you’ve answered those questions, you can decide what approach you want to use for architecting your data: data lakes, data fabric, or a data hub. According to recent research by McKinsey, only 8% of data lake implementations have moved from proof of concept to production.
Lack of trust in data
According to KPMG, only 35% of surveyed organizations have a high level of trust in their organization’s use of data analytics. To fix this problem, you need to build trust and activate your data. Start by asking yourself the following questions:
- Where data can trigger automated actions?
- How do we use data for targeting and personalization?
- How do we build trust in data?
To activate your data, you need to make sure that
- Data is reliable and people trust the numbers they see. This can be done through different testing and monitoring approaches.
- Data is available for collaboration. Go and talk to stakeholders in the organization to understand what data they need to act upon.
- You have a data governance council who is responsible for overseeing all data-related processes.
John shared a couple of real-life examples, including the story of an American non-profit broadcaster and the approach the Data Discovery team used to build an executable data strategy. The stages they went through were defining the goals, aligning them with business objectives, and setting up measurements:
Once all these business objectives were written down, the Data Discovery team prioritized them based on the business impact and level of effort required, and defined their implementation roadmap moving from the most crucial and valuable tasks to the less important:
Don’t let the low success rate of data lake implementations scare you off. The potential impact of a successful implementation is way higher than the necessary investment. To get started and avoid other companies’ mistakes, use these 10 secrets of scaling analytics. You can always get inspired by looking at case studies of other companies. Here are a few examples:
- Ile de Beaute used end-to-end analytics to measure the ROPO effect
- Comfy built end-to-end analytics in nine simple steps
- BUTIK. improved customer LTV by combining data
The next session was called “Birds of a Feather,” where everyone could name a topic or pain point they wanted to discuss and move to a table to meet like-minded colleagues for an in-depth discussion. I personally really liked this session because it gave me a couple of ideas on a problem I didn’t know how to tackle.
“What you need to know about you – psychology for success” by Michele Kiss
Michele is a charismatic speaker who knows how to grab people’s attention, and I really liked how she showed the tricks of our brain — limited working memory, confirmation bias, conformity to the norm, etc. — that have been proven by classic studies. It’s important for analysts to be aware of these contradictions, because analytics and optimization are all about using quantitative methods to understand people and why they do what they do. Here are some of the takeaways:
- When presenting information, you need to work within the limits of perceptual and memory systems.
According to George A. Miller, a famous psychologist, the amount of information we can retain in short-term memory is seven plus or minus two items. An item is a chunk of information, such as a word, number, data point, etc. That’s why expecting users to draw connections between two data points six pages apart or thirty slides ago is a recipe for failure.
- Even if your data seems self-evident, if you come in with “breaking news” that goes against what the business has known, thought, or believed for some time, you may need more data to support your contrary viewpoint.
You may also want to allow plenty of time for discussion rather than simply sending out your findings, as those discussions are critical to getting buy-in for this new viewpoint. “A man with a conviction is a hard man to change. Show him facts or figures and he questions your sources. Appeal to logic and he fails to see your point,” said Leon Festinger.
This is also called the theory of cognitive dissonance. It tells us that people don’t like the feeling of inconsistency (for example, between our beliefs and our actions.) We seek to reduce this uncomfortable feeling by justifying our beliefs and avoiding information that might conflict.
- When working with data, don’t stop at your first hypothesis. It can be influenced by your cognitive bias.
Our brain is built in a way that it seeks to affirm existing beliefs. And that can lead you into a mistake, since there could be many reasons why the data shows what it does. I actually find understanding and recognizing this bias to be really handy when interviewing people.
You usually get your first impression within 15 seconds, and after that your brain is seeking something that will support that first impression. It’s super important to notice this moment and start looking for facts that support a contrary position. It will help you escape from your cognitive bias.
- If you have analytical findings that challenge the status quo, you should discuss them slowly and one-on-one.
Avoid putting people on the spot to agree or disagree within a group setting because it’s less likely that people in a group will stand up for your new findings against the norm of the group. Similarly, this argues against jumping straight to a group brainstorming session. Once in a group, 76% of people will agree with the group (even if they’re wrong!), so it’s easier to get more varied ideas and minimize groupthink by allowing for individual, uninhibited brainstorming and collecting all ideas first.
This effect is called conformity to the norm, and was proven by Ash in his experiments in 1951.
- When bringing findings, make a recommendation on how to use them and be as specific as possible.
Define whose responsibility it is to act if something changes and make sure you’ve been heard so you don’t get trapped by the bystander effect – the more bystanders are present, the less likely it is that an individual will step in and help. A great example of this is in a medical emergency when everyone is just standing and waiting for someone else to take action.
There are way more examples and biases that you can find in Michele’s presentation. If you’re interested in learning about biases, check out this article on practical psychology for data scientists by Conor Dewey.
“Identifying leadership personality” by Valerie Kroll
Over lunch, I listened to a really inspirational talk about identifying leadership personality by Valerie Kroll, Optimization Director at Search Discovery and President of the Digital Analytics Association Board of Directors.
In her talk, Valerie shared her 12-year journey from a quant intern to where she is now and her way to becoming a leader. Here are several lessons she has learned:
- You don’t have to manage people to be a leader. A leader motivates and inspires others to willingly take action; they carry the culture and constructively challenge the status quo. I can also add that to be a leader, you need to have strong empathy and a willingness to take responsibility.
- Find leadership opportunities and don’t get overtaken by imposter syndrome:
To combat imposter syndrome, use your analytical skills — try looking at facts and assessing your progress and current level based on it, write down your strengths and achievements, and find ways to apply your skills. Also, you can give a try to The Anxiety and Phobia Workbook by Edmund J. Bourne.
- If you face biases or discrimination, talk about it, ideally with the wrongdoer. If not, find a colleague, mentor, or someone in HR. If your company doesn’t have processes in place to handle an important issue, bring it up. Don’t be afraid or ashamed of who you are – your age, gender, nationality, etc. Maybe it’s something that makes you successful and helps you bring a different perspective.
- Build your personal brand on social networks and encourage the T-shaped model, evolving not only in breadth but also in depth.
- Leverage your network by making real connections and have a ready-to-go elevator pitch. Find connections at meetups, ask people to introduce you to other interesting people, and develop those connections by providing value. Be ready to share what you’re working on that’s creating the most value and is the most innovative in less than a minute.
- Look for mentorship. It can be mutual, because the best way to master a skill is to teach someone. When looking for a mentor, consider soft skills as well as hard skills, focusing on your career goals.
By the way, DAA has a Women in Analytics mentoring program, so if you’re interested, check it out.
“Analyzing Analysts to Improve Process Efficiency” by Mariia Bocheva
Then it was already time to share my story on Analyzing Analysts to Improve Process Efficiency. Growing a team requires a lot of time and effort and the proper management tools. The pain points are common: inefficient distribution of tasks, no time to teach and coach new employees, not enough time for seasoned analysts to do R&D and improve their skills, having no idea how much time a given employee spent on which tasks, and the list goes on. We’ve used data lakes to improve task estimation, ensure that painfully learned lessons are shared with everybody, and balance project priorities.
As a result we
- realized that the workload of an analyst is far from what we expected and that average values can hide our growth zones
- proved that most of our analysts (~85%) answered emails on time
- mapped the typical tasks that we run into, how long it usually takes to accomplish them, and how the time for each particular task can vary
- found weaknesses and strengths for each analyst to customize their personal development plan
found areas for automation.
You can find my slides on Slideshare and an article on ConversionXL that plots out our experience.
Sadly, I missed Garry Angel’s, Tim Wilson’s, Matt Gershoff’s, and Moe Kiss’s talks, but I heard them all speak at Superweek and know they’re brilliant. Even though I missed their talks, it was great to see them again at #MASConf!
The very last session at the Marketing Analytics Summit was a live recording of the Analytics Power Hour podcast. Each episode of this podcast has a closed topic and an open forum.
The audience enjoyed listening to Michael Helbling, Analytics Practice Lead at Search Discovery, Moe Kiss, Analytics Consultant, and Tim Wilson, Senior Director of Analytics at Search Discovery share their thoughts and experiences. You can find our previous exclusive interview with Tim Wilson on the state of modern analytics here. Also, you can find the episode from Marketing Analytics Summit here. I highly encourage you to listen to other episodes too.
- Don’t take a title at face value; cut through the hype and focus on the impact you deliver.
- To understand if self-service analytics works in your company, ask your team if they have the data support they need to do their job effectively and if they’re able to get data in a timeframe that meets their business needs.
- Use a data role matrix to make sure everyone understands what can be expected from each position.
- Be proactive when it comes to your data quality, not reactive.
- Only 11% of marketers feel “very confident” in the accuracy of their attribution model. But that doesn’t mean you shouldn’t try.
- To assess your data maturity level, you need to evaluate your data across three different categories: unified insight (performance measurement and attribution), unified data (siloed, co-located, or unified data), and predefined data foundation (tracking precision, touchpoint coverage, taxonomy standardization).
- Businesses that link marketing metrics to desired business outcomes are twice as likely to significantly exceed those outcomes. And businesses that link marketing metrics to revenue targets are three times more likely to significantly exceed those targets.
- Both marketing specialists and analysts want one thing: the right tools and processes to know what to start, stop, and continue doing in order to reach their marketing goals.
- Use a success plan to align marketing activities with your company’s business objectives and demonstrate the value of marketing and analytics investments to senior leadership.
- Focus on revenue-driving processes first, as they bring more impact.
- To make your processes work, map them, analyze them, and communicate them to the team.
- Physical interaction is easier to understand and seems more persuasive than digital – physical attention leads to 70% higher brand recall. So to be remembered by your audience, you need to engage more senses, reduce distractions, and encourage interaction.
- Control your brand with on-SERP SEO, as the search results page is your new landing page. You need to leverage rich snippets and claim your panels.
- Double down on branded demand creation.
- Leverage social algorithms by alternating non-promotional and promotional posts.
- Use storytelling to create content that’s connected to your brand, earns high engagement, and creates emotional resonance.
- There are three main problems with data: lack of data literacy, data silos, and lack of trust in data. All of these problems can be tackled with an executable data strategy – a plan for using data purposefully that includes understanding your data, architecting your data to make it useful, and activating your data for business impact.
- Phycology and analytics are connected — be aware of mental shortcuts and use them to leverage your analytics and marketing powers.
- To find your leadership model, use your differences, develop your network and personal brand, look for a mentor and mentee, combat imposter syndrome, and inspire others.
- The sooner, the better. Collecting, merging, and preparing data is about 75% of your efforts. Make sure you trust the quality of the data you’re collecting.
- Start with an MVP dashboard. Focus on no more than 10 critical KPIs.
- Define what you’re going to do if a metric changes dramatically at 5 p.m. on Friday. You should have a plan for what to do if a metric rises or falls unexpectedly. If you have no idea why you should have such a plan for a certain metric, consider whether you need to track it at all.
P.S. It’s been a great experience. A separate thank you goes to Jim Sterne for inviting me and to Roxanne Glavina, Speaker Management at Rising Media, Inc. for the smooth organization.