CXL Live 2018 Recap: Top 5 Lessons from Each Speaker

CXL Live is our flagship conference about growth and optimization. The 2018 edition of CXL Live brought together 400 practitioners from 22 countries and everyone was “locked in” at a beautiful resort for a full three days.

The event had returning attendees…

… as well as first-timers:

The event was again emcee’d by the only and only Michael Aagaard:

Here are 5 tips / thoughts / lessons (it was super difficult to narrow it down to just 5!) from each of the speakers.

Bryan Eisenberg – Saying Goodbye to the Buy Button

  • The interface that we’re going to be using to buy stuff in the future will not be the big screen.
  • Voice is growing super fast, time to invest in it is now. 57% of folks have ordered something via voice on their device. Amazon has hired more people for their Alexa division than Google has for the entire company – it’s not hunch, this is going to be the focus
  • The actual year of mobile was 10 years ago (release of the iPhone) and we still have horrid mobile experiences out there.
  • Amazon built its growth on top of these 4 pillars:
    • Customer Centricity
      • “The most important single thing  is to focus obsessively on the customer” – Jeff Bezos
      • They may believe “the customer is always right” but how do they act? [Lush example]
    • Culture of Innovation
      • Prime was a 6-page memo: inspired by Costco membership model – built as a test
    • Corporate Agility
      • Two pizza approach – organized for execution
      • How many emails do you get cc’d or bcc’d on? Transparency correlates to agility (Does your CEO know how many tests you ran last month?)
    • Continuous Optimization
      • Aligning customers & business objectives
      • Not just about lift – what kind of impact does this have on the customers?
  • Amazon is a big grizzly, we are the fish it’s trying to catch. In order to survive, we don’t need to beat the grizzly bear, just need to be faster than other fish. Fastest fish wins.

Hana Abaza – Thriving on Change, Driving Growth and Lessons Learned at Shopify

  • There are low-hanging fruits everywhere. Following that can be damaging to long-term goals.
  • You have to get more systematic about positioning. No more puking up unicorns and rainbows
    • Marketing can polish a turd. Positioning can turn turds into fertilizer.
    • Positioning isn’t something marketing makes up, it’s something marketing uncovers
    • Messaging and branding is an expression of positioning
    • If you’re in an established industry, positioning needs to focus on how you are better vs who you are.
  • Most marketers say they know that not everyone is in their target market, but they don’t always act like it.
  • Facts don’t change minds – true for both sales and customers.
  • You need clarity of purpose – otherwise everything else will fall apart and you’re not able to prioritize.

Ed Fry – Customer Data Operations: Unleashing your hidden growth engine

  • What you need to do hyper-personalization at each touch point:
    • Know everything about your customers.
    • Personalization is a data problem, not a marketing problem.
    • Data is siloed – sitting in different tools that don’t talk to each other. There are 5k marketing tools – completely overwhelming! Madness! It’s the “Frankenstack”. This causes a drag on all aspects of what we trying to do as marketers.
  • The solution is customer data platforms (CPD’s). Buying a CDP? Don’t be fooled – look for these must have capabilities/features…
    • Unified customer profile
    • Transform raw customer data into new traits
    • Query, build & update segments of customers
    • Map & sync profile data
  • The power is in how customer data is used. What experiences can you create from your tools, teams, and data?
  • 5 steps to freedom:
    1. Find your ideal customer profile. Find agreement on a common definition.
    2. Map entire customer journey. Track & trace all customer paths/journeys. Map out each lifecycle stage and variations, and name each. Assign content & channels for each experience
    3. Hire & fire – tools to do a job. Experiences happen in loops. Identify your channels. Use data enrichment.
    4. Integrate profile data between tools. Map all key profiles & data. Enrich – to understand if this is an ideal customer. Create segments, sync across data across your tools.
    5. Orchestrate 1:1 Personalization [at scale]. Ex. email: use data to send highly personalized message  by pulling data from across sources (e.g. previous conversations, facebook, intercom…). Explore the orchestration ideas together as a team. Work through 1 at a time.

Ezra Firestone – What Works in Email Marketing Right Now

  • Email & ads are the 2 big communication channels. On avg $1 spent on email returns $38.
  • Structure that works the best for emails
    • branded header and tagline
    • headline
    • sub-headline with CTA
    • remind people that you care about them
    • upleveled aesthetic
    • match the brand ethos, congruent with the website
    • updated footer
    • add links to relevant content
    • 88% check email actively on smart phones, needs to be mobile optimized
    • mobile is  ~50% of revenue, shorter form works better
    • 66% of unsubscribes takes place between 5 and 10pm
  • Target email openers and clickers, run ads to people who opened but didn’t click, clicked but didn’t buy.
  • Embed blog videos hosted on facebook so you can track consumption and advertise to them.
  • Messenger will become the biggest channel over the near years. Example: wechat in China
    • intimate communication is how we actually have our conversations, not email, or ads
    • more activity and engagement than any marketing channel, better than email, paid, physical mail
    • messenger sub worth 5 to 10x an email sub
    • 60-80% open rates and click through rates
    • instant delivery
    • buy right from Facebook news feed
    • it will never be as easy or affordable to build Messenger audience for your brand as it is right now.

Alexa Hubley – Master Customer Marketing by Watching Romantic Comedies

  • Customer Marketing – our job is to build relationships at scale so they fall in love with us. We want them to sing from the rooftops about how much they love us.
  • Two scariest words for SaaS customer – “pricing change’ – avoid shocking customers
  • Create intimacy
    • Hard to scale
    • Focus on planning for segmentations
    • At unbounce – works at a modified Google Sprints structure. Understand-decide-prototype-finalize-launch (data and research at beginning, decide on targets in middle, QA before launch). 14k customers at unbounce -> 14 groups/cohorts
  • Woo your customers. They used humor – created a Chief Discount Officer’ – CDO – ‘humor to sell a deal’ campaign that included multple emails and direct mail.
  • Build trust through recriprocity. When people are made to feel special, they respond favorably to offers.

Tara Robertson – How to 10x Growth by Optimizing Customer Marketing & Retention

  • Retention is the most important thing – if that’s poor, nothing else matters. Concentrate heavily on retaining your customers. Flip your funnel – only 5% of revenue comes from optimisation but 92% of revenue from retention.
  • 2 reasons for churn:
    • Customers don’t see value (perceived value)
    • Environmental (change in strategy, etc nothing to do with you)
  • Main focus – start testing your pricing model and the value proposition (what resonates with the clients)
  • Create win-back campaigns for churned customers
  • Use qualitative research: Send out surveys and use incentives (10 questions). Arrange interviews after the survey to get a deeper understanding. Only ask open ended questions. Ask for brutal and honest feedback (try to get negative/bad feedback) that shows what to change.

Bangaly Kaba – The Path to 1 Billion: Lessons learned from Growing Instagram

  • Focus on the marginal user (= non-power user) – matters because every click is expensive to that user. Growth exists to remove barriers to adoption.
  • Impact requires “understand work”. Sustained impact requires working from first principles – Understand, Identify, Execute vs Identify, Justify, Execute (the latter is the wrong way to do it).
  • Track ecosystem health daily. Start with retention. Understand growth accounting – new, resurrected vs churned = net growth (can also look at this at feature level, not only product). Work hard to define meaningful product metrics – enabler of team success.
  • Iterate to the vision. What is the end state you want to have? Have a vision.
  • Zoom out & solve the larger problem.
    • Understand: funnel analyis to identify opportunities – 75K people / day werent able to  login because of a bad login screen design
    • Identify: increase prominence of FB login entry point
    • Execute: things don’t always go as planned – the update did not work, so had to iterate

Candace Ohm – a Mathematician’s Guide to Growth Optimization

  • To understand acquisition you must understand lifetime value.
  • You have 2 types of users: occasional users and frequent users. Frequent users:  key to your growth.
    1. Increased engagement
    2. Virality: # of users that refer your product for you: exponential growth if your K factor is greater than 1
    3. Network effects: Uber effect. Each additional user leads to exponential growth
  • If you’re not focusing on the right things the slope of the curve will just shift up but stay flat. Spiceworks focused on increased users, but it ate them alive. In 2016, they had a huge layoff and had to shift strategy.
  • Dropbox: first achieved growth through virality. Friends referring friends. So they optimized referral process. Dropbox:  refer my friends and family – generates a nice personalized email. Easy process for the referred friend to follow, and it’s a win for both. Building trust. Everybody wins.
  • Optimize for getting people to become more frequent users.

Guy Yalif – Demystifying AI for Marketers

  • What is AI is good at? Managing a lot at once. Accelerating learning. Listening and reacting 24×7. Acting with precision (sample size can be big). AI helps to get more done, deliver a lot.
  • Before AI – we come up with an idea, do 50/50 of traffic and get a winner to show everyone. Is my audience all the same? No. AI can do 1:1 personalization at scale.
  • When not to use AI – when you’re a small startup – first figure out your market; if you don’t have internal or external people who come up with ideas.
  • AI can dramatically increase your testing velocity. You can use AI even when you haven’t mastered traditional testing.
  • AI is perfect for situations when your traffic sources change, and what works best changes.

Chad Sanderson – The Statistical Pitfalls of A/B Testing

  • Significance in the inverse of the p-value. P-value is the probability of an event happening under normal circumstances.
  • In an A/A test the distribution of p-values is random
  • If we are running AB tests and we are adding trials, then the more trials we add, the higher the chance we see an error<
  • Do not stop a test because you see significance.Set your sample size in advance. Run your test for weekly intervals or for other intervals that take into account customer behavioral changes through time.
  • The larger the number of visitors in your test the higher the likelihood to detect a significant difference.

Els Aerts – Without Research There is Nothing

  • Do the research for your situation. Do not copy other people’s templates or best practices. Do the research for your own models. You are gambling if you use other’s best practices.
  • Start with the user. Obsess with the user. Start with what the customer needs and start backward. It’s not touchy, it’s business. Customers drive growth.
  • In-person moderated user testing is great.  One on One. Allows you to have the user right in front of you. Do moderated user testing and not focus groups. The data is observed behavior and not customer opinions.
  • Focus Groups are great for ideation. A FOCUS GROUP IS NOT A USER TEST!
  • Three things for valid data
    1. Recruit the right users. Test your target audience (both existing and target)
    2. Write a good scenario. Do not ask questions, set tasks. Goal: observe behavior (not look for opinions)
    3. Be a kick-ass moderator. The moderator is the biggest strength and the biggest weakness in moderated user testing.

Matt Roach – How to Optimise Big Corporates (with a Lot of Legacy)

  • So easy to say you need to do this this and this but the devil is in the execution
  • 7 drivers of optimization success
    • You need to have a standardized process
    • Budget (tools & resources)
    • Organization structure. Typically the hardest thing to get right in CRO. Silos are not effective.
    • Skills. Value = (Knowledge + Process) x Skill x Attitude
    • Attitude (People). Enthusiasts “just get it”. Proofers “have to prove it”. Deniers “just don’t want to know” and need to leave.
    • Culture. Attitudes, opinions, and traditions you have within the company. Culture Need: Data-driven, permission to fail, focus on long term customer value.
    • CEO. With their buy-in, you can succeed, without it you’re fighting every step
  • Size exacerbates issues, but age is the real issue.
  • Not getting big AB testing wins doesn’t mean you are not optimizing.
  • You can optimize anything – even your worst nightmare.

Mats Einarsen – Lessons Learnt from Creating a Large Scale Experimentation Culture

  • Don’t assume people get it – Many of the terms we’re using are overloaded. You want to be sure you’re really explicit and make sure people are trained to understand the terms.
  • Train your people. Get everyone on the same line. Create a structured curriculum. Track who’s trained. Especially the new people. Do spot checks. Talk randomly about Type II errors with people from the team. Teach people how to convert.
  • Systematically recycle hypotheses. Mats uses spreadsheets for this. Where have I tested this? What other pages could I test this hypothesis on?
  • Decisions become rules. Bad testing habits tend to spread throughout organizations. Keep a strict decision-making process as the #1 in importance. Testing velocity comes second.
  • Everything must be tested. You need to test literally everything. Bugfixes, strategic initiatives, legal requirements, known wins, known losses.

Renee Thompson – How to Win at B2B Optimization

  • What makes B2B different?
    1. Offline sale – typically
    2. Long sales cycle – 18 months or more
    3. There’s a buying team (as opposed to individual)
    4. High risk – can’t send back if it doesn’t fit
  • There’s no single metric to track, there are many:
    • Leads
    • Quality of leads
    • Engagement
    • Are they in your target demographic, firmographics…
    • How do you tie this all back to business value?
  • How to do shorten a long sales process? Empathize. Activation:
    • Move them to the next stage
    • Email
    • In-experience (nudges etc)
  • Optimize for lead quality, not quantity. You’ll make more money.
  • Free email service (gmail ,yahoo etc) leads don’t become users: worth 2.7x less than business emails. Require corporate email addresses on your forms.

Yu Guo – Scaling experimentation at Airbnb: Platform, Process, and People

  • Airbnb growth is empathy driven, evidence fueled, through experimentation. Fueling growth means scaling experimentation effort.
  • All new recruits go through Data University. The goal is to empower EVERY employee at Airbnb to make data-informed decisions by providing data education that scales by role & team.
  • Everyone can see all experiments. Knowledge Repo – open sourced. Stores not just data, but the knowledge that comes with testing. Learnings, next steps, etc. Improves knowledge sharing.
  • There are guidelines for end-to-end lifecycle in data-driven decision making. This ensures decision consistency across teams.
  • There’s an Experiments Council: internal consulting to help with tricky design , setup and interpretation.

Vab Dwivedi – E-Commerce and Customer Experience Optimization Practices from Dell.com

  • Be customer obsessed, you’re really presenting the voice of the customer.
  • Deliver analytics in a consumable manner. Drive decision making through data, take out the emotion and opinions.
  • Serve as a channel of knowledge. Share test results via newsletters and open forums.
  • Be agile, move at the speed of business, don’t hold up product development.
  • Dell launched an entirely new Dell.com frontend and backend. Tested it on a smaller segment before pushing it live to everyone. They expected a quick win but it “really didn’t work” -33% revenue per visit. After more research, they introduced UI enhancements, 24 of them. Result: -45% revenue per visit. The reasons were all technical (bugs).

Merritt Aho – May the Best Ideas Win (They Usually Do)

  • We produce a nearly endless stream of bad ideas. We’re true maybe 30% of the time. It’s our imperative to find ideas that work. But how big was your bonus based on your learnings? No-one’s making a career based on learnings alone.
  • Ideas that are most easily accessed from memory seem to dominate our thinking. Out of a universe of possible decisions, we usually choose the first one we remember. It’s not just laziness. Brainstorming is not the right answer. Ideas die in team brainstorming sessions. They are not thought out deeply enough.
  • You have to believe that the best idea is not the most promising one that comes from your consciousness. The best ideas are the result of a process. Quantity over quality. 3-5 people. Subject matter experts + people with very different backgrounds (the spark). Remove distraction. Narrow your focus: talk about specific things.
  • During sessions go deep. Just get it all out. Kick it off with a couple of really dumb-sounding ideas. I don’t want you filtering the ideas out. You’ll never know when an idea sparks a tought in someone else. Just keep it at it until the pace slows down. Build on someone else’s idea.
  • Take care of decisions later – only when a decision really needs to be made. This gives you the ability to broaden your audience. No-one gets to critique, no-one gets to evaluate in the session.

Gary Angel – Extending User Experience Analytics into the Real (non-digital) World

  • Stores (physical) have not changed in 20 years, online has. But now everything is changing (Amazon Go stores – online to physical).
  • Optimizing physical stores has to happen, or it’s not really omni-channel. What gets currently measured in physical stores?
    • Door count (how many customers go in) + How much is sold
    • Experience inside is a mystery
    • Key questions that we can’t answer – foundational info we don’t have so we can’t optimize
      • Interested but didn’t buy?
      • Consider a new product category?
      • Did associate interactions improve conversion rate?
  • Tech helping us monitor behavior
    • Wifi. It’s everywhere and always geo-tracking you.
    • Passive sniffer. Similar to wifi, but a little more accurate.
    • Phone. Geolocation is in mobile apps as well (actually very accurate). Small snippet of code
    • Video camera. Security yes, but also doing double duty for measurement
    • Allows us to answer fundamental questions
      • What engages and when does that sell?
      • Are areas of the store underperforming?
      • Is the staffing level appropriate?
      • How much localization is necessary?
  • Measuring physical space
    • Floor Mapping (path analysis) – track route, where they stopped/paused, how long they spent at each stop
      • Parallel to what we see in analytics on URLs (same premise)
    • Functional analysis. Utilizing space to “push” customers into certain spaces, increase impulse purchase, etc.
    • Funnel abandonment. Where do people leave? What do they look at?
    • Content Attribution. What’s engaging people? What’s staff performance like?

Moe Kiss – The Pursuit of Customer Happiness: Why Customer Experience Across Devices Matters

  • There’s no such thing as an “app user”-  we use multiple devices across the whole user journey.
  • Embrace the randomness of customer behavior. Example: customer checking order status over 100 times in one day
    Trying to push into linear view of journey doesn’t work.
  • Talk to analysts to figure out what kind of user stitching you’re doing.
  • Challenge data silos (metrics split out by different devices) by reviewing your metrics.
  • Create an experiment about a frustrating user issue.

Rachel Sweeney – Building an Optimization Framework driven by the Cloud and AI

  • If you want to do Data Science, hire a Data Scientist. You need the broad knowledge, depth and skills of a data scientist.
  • If you want to embark on a machine learning project, decide on a project area. Find ONE specific thing that you can create a model that will learn what to do.
  • Stress test and build trust. Pilot test with LIVE Data.
  • Machine learning is new. Know if you can trust it. It only takes one team member to turn the team against it.
    Tip: Establish the benchmark that people are comfortable with (e.g. 95%, similar to statistical significance.)
  • Machine learning will be huge.


This was just 5 tips picked from 100 tips every speaker shared. The experience can’t be passed on via a blog post.

You need to come to CXL Live.

Related Posts

Join the conversation Add your comment

  1. This is great info, Very helpful. Thanks for sharing.

  2. Awesome recap blog post with some pretty amazing points. Thanks!

  3. This is a fantastic summary. Makes the conference even more appealing as well. Thanks for giving us access to the info.

  4. Great read! A recommended blog for any serious marketer.

Comments are closed.

Current article:

CXL Live 2018 Recap: Top 5 Lessons from Each Speaker