MeasureCamp Stockholm 2023

M?ster Samuelsgatan 56. Ah! 56, just like the number of MeasureCamps I have attended. This time, it's MeasureCamp Stockholm. I have always loved being in Stockholm and Sweden in general, and I love it when the locals stick to Swedish when I address them in Swedish, considering that it's not quite there yet. But enough humblebragging. This was the third MeasureCamp in Stockholm, with more to come with MeasureCamp Malm? early next year. It's only a matter of time until Sweden becomes one of the few countries with three MeasureCamps. Sweden has IHM and Medieninstitutet, two of the most reputed schools teaching Digital Analytics, with campuses in Stockholm, Malm? and G?teborg (Gothenburg). Without further ado, here are my takeaways.

Adobe Analytics... where do I start, by Ying Liu at Adobe

What a breath of fresh air when someone does a session about Adobe Analytics! I have been on the Adobe stack since 2011 and only that stack since 2015. So, there was no way I would miss Ying's session. Ying started his session with a metaphor I found particularly useful and one I had never heard before: If Digital Analytics tools were a car, Google Analytics would be an automatic car, and Adobe Analytics would be a manual one. Adobe Analytics is not for companies engaging in Cargo Cult Digital Analytics or building the digital equivalents of Potemkin Villages. It's an expensive tool for companies that demand positive ROI on their Digital Analytics investment. Digital Analytics and A/B testing does drive change at such companies.

Ying is part of the Adobe Experience League and gave us an overview of their offer. Adobe Analytics is a complex product to implement and use compared with Google Analytics. Anybody using Adobe Analytics can benefit from these resources, many of which are free.

Tutorials

Many bite-size free Adobe Analytics courses are on the Adobe Experience League Website. Many take no more than a few minutes. There are also longer two-hour classes for people with the time to commit to deeper learning. These are free. Some of these tutorials are accessible from inside the tool itself. There are also instructor-led classes, but these are not free.

Experience League Communities

You can raise there questions, start discussions and submit product feature ideas.

Mentoring programme

Ahead of certification, you can join study groups. These study groups enrol students every six months. I will speak about the certifications a little more below.

Adobe Analytics User Groups

Adobe has YouTube channels offering replays of live events. Jen Lasser and Eric Matisoff , both product evangelists at Adobe, post on these channels. There is an EMEA channel.

Adobe Analytics annual challenge

Every year, Adobe organises a data analysis contest. It's too late for 2023, but they will enrol for next year's competition soon. They use real clients' data.

Adobe Certified Experts

There are three tiers of certification. As a former Adobe Analytics Implementation ACE myself, these expire after two years if memory serves. The first exams are not free, but renewals are free. By renewal, I do not mean retaking a failed exam. The retake should be cheaper than taking the exam for the first time. The renewal exam is shorter and consists of only a few questions.

Adobe Summits

Every year, Adobe organises a conference in Las Vegas and London. These conferences showcase all the latest Adobe products, unveil the company strategy, offer sneak peeks of the products in development, case studies, interviews with celebrities, and a concert on the first night's evening. It's great for networking. Attendance is only free if you are an Adobe client.

LinkedIn Learning

Eric Matisoff offers two courses there and lets the students use two sandbox accounts. That is huge. Getting test accounts for Adobe Analytics has always been challenging to access. Only Adobe Partners had them; getting the Adobe Partner status alone is no picnic. It comes with sales targets and several ACE-certified employees. There are three LinkedIn groups for Adobe Analytics, and you can also follow #adobeanalytics.

Other resources

The Measure Slack has several channels dedicated to Adobe Analytics, Adobe Launch, and Adobe Target. Eric Matisoff co-wrote a book with David Karlins called Adobe Analytics for Dummies. There's also the Digital Analytics Power Hour podcast. If you are in the US, I encourage you to join the Digital Analytics Association.

The CLV Revolution, by Valentin Radu , Founder at Omniconvert

Valentin is not the kind of person who will tell you: "I can't!" So he will do it when the challenge comes to condense 13 years of experience in Customer Lifetime Value (CLV) into 30 mins! And now, I will try to summarise that even further.

At first, he tried targeting potential customers with stock photos but localised messaging to sell car insurance. Since it was not as bad as taking pictures of penguins pretending to speak Romanian, Valentin managed to get a conversion rate uplift of 60%. From there, Valentin did more CRO, more experiments blending quantitative and qualitative data before reaching a plateau, which he called experimentation fatigue. It dawned on him that A/B testing is akin to serving a single message to all, but one size fits all can't work. You need to become more customer-centric and embrace personalisation. That's where the focus on CLV is crucial, according to Valentin.

Customer acquisition costs (CAC) are incurred only at the start of the process. But then comes customer retention costs (CRC). Valentin then experimented with Google Analytics, online surveys, data mining and recency, frequency and monetary (RFM) segmentation. RFM segmentation is about breaking up the monolithic prospect base he was trying to address with a single message into marketing audiences. For example, he identified Lovers, Ex-Lovers, Soulmates, Don Juans, and many more. Here, the spirit of John Wannamaker paid a short visit when Valentin mentioned how great it would be to spend his marketing budget more on Soulmates, who keep buying regularly and none on the Don Juans, who bought only once and never again.

By blending quantitative and qualitative data, optimising the product assortment, and improving acquisition with a focus on the ideal customer profile (ICP), Valentin achieved a 30% uplift on CLV in 6 months.?

Ultimately, nobody will buy back if your products are bad, no matter how aggressive your discounts are. Valentin has published a book called The CLV Revolution, which you can get there: https://www.amazon.com/dp/B0CJC7FKC5

Why you should get into data architecture design, by @Arianne Leijenaar, Founder at ProfitBread.io

Data Architecture and engineering are all about blending data sources to help analysts create a consistent stream of actionable insight. It will require processes such as Extract Transform Load (ETL) or, more often these days, Extract Load Transform (ELT).

First, you will need to get access to that data. You will need a seat at the C-suite table because the data you need is not just Digital Analytics data. Other business owners manage these other data sources. Depending on the business requirements, you will need access to that data quickly, sometimes even in real-time or near-real-time.

Arianne recognised that companies increasingly recognise that they will need data engineers if marketers continue to refuse to work closely with data. The same thing happened with tag management systems (TMS). The vendors planned to make TMSes user-friendly enough for the marketers to use. That didn't work. Now, the vendors have unearthed the skeleton of self-serve with the hope that the marketers will start using Digital Analytics tools and analyse data. They won't.

Arianne explained that companies need a blend of data architecture and data engineering. The former focuses on the overall design, the data sources, and how to join them. The latter focuses more on the day-to-day execution and exploitation of that data architecture.

Arianne took us through the six steps for the delivery of a data architecture:

1. Collecting the business requirements from the C-suite - remember that they know next to nothing about data and only heard of a few data-related concepts, but without understanding how they fit together.
2. You need to map these requirements to KPIs
3. Document the naming conventions and definitions for these KPIs
4. Choose the tools based on the business needs, especially which Cloud Computing platform and product the company needs. It's likely to be the Google Cloud Platform (GCP), Azure (Microsoft) or AWS (Amazon). However, each platform will offer various data storage options depending on the retrieval, storage speeds, and scalability your business requires.
5. Schema design, i.e. the data sources and how they join
6. Creating a data governance plan

As you present your plan to the senior management, expect some resistance. They will see something expensive and risky, especially if an existing data architecture is in place. What follows is from me: make the business case showing that inaction has a price and risks. Senior management may have the bias that the status quo costs nothing. Your proposed plan must be less risky and have an aggressive edge over the status quo. No sports team wins by having the best defence.

Share our novel use of ChatGPT, by Astrid Illum , Director & Tribe Leader of Customer Engagement at DFDS

Astrid proposed a fun session where we took turns offering unique ways to use ChatGPT. Here's the list:

• Content generation - this could include creating catchy titles (hint: if there's a colon in the title, there's a strong chance it came from ChatGPT; it has always included them in the titles I asked it to generate)
• Code check, explanation, generation, debugging - that would help my colleagues understand my code. I have been coding Javascript since 1998; my coding style is terse.
• Negative LLM alternative to the real thing - I don't quite remember what this one meant.
• Idea review/generation - I do this before writing some of my articles, using ChatGPT as a sounding board and asking it to find some angles I might have overlooked.
• Semantic analysis of customer feedback - take many reviews and extract the most common complaints.
• Anxiety management tips - ask ChatGPT what you can do during a medical procedure where you are not anaesthetised to stay relaxed.
• Client coaching - tips on how to get client buy-in
• Topic exploration - I use this to get ChatGPT to identify influential thinkers and philosophers on the topic and see how they group into schools of thought.
• A/B test analysis - seems only possible with ChatGPT 4.0
• Summarise across languages
• Scholar AI plugin to summarise and find the academic consensus about a topic: Valentin did this with his research on CLV.
• Tweak CSS - CSS can be a pain to write and debug, but not for ChatGPT.
• Prove me wrong - intriguing one; I think I'll try it soon.
• Join dataset - if you don't know SQL, that could be useful.
• Next-level Google - many are now turning to ChatGPT rather than a search engine. Stack Overflow is feeling neglected, too.
• Cooking recipes
• Plot exploration for popular TV series
• Perfect prompt extension for ChatGPT
• Developing interactive prompt skills
• Task management - that one got a lot of people thinking after Astrid explained how she dumped a to-do list on ChatGPT. ChatGPT then produced a task list with estimated times and congratulations when she ticked an item off the list. It even recommended secretly playing?bullshit bingo?in meetings to stay focused.
• Voice-to-text summary for meetings - I am actually looking into the reverse of that, i.e. training a model from my voice to prepare course material audio. It would save me time recording audio.

CDPs - Beyond the bullshit, by Patrick Oliver Mohr & Marcus Stade , co-founders at MohrStade

The CDP's promise of blending all the data and sending it to marketing platforms like Meta or Google doesn't work. First, we need to manage the consent permissions. The CRM then manages the personally identifiable information (PII). The CDP then creates audiences from the data collected. We can then share the data with the marketing platforms, which will push the targeted messages to online visitors.?

Marcus uses WebHooks. He sends just the external id, nothing more, because there's no reason for the marketing platforms to know anything more than this external id.

Patrick said that although you can choose a vendor CDP, nothing prevents you from selecting a Cloud Computing platform to do the same thing. Patrick added that the CDP should have access to the website HTML code and push the external data id in a hidden form field.

Marcus showed us how to push the CDP external data id using the Facebook graph API.

This was probably the most technical session of the day for me, and the one with which I have the least immediate experience. I might have transcribed something wrong, so keep your eyes peeled for Patrick's and Marcus' posts.

It's time to use data for good, by Astrid Illum and Valentin Radu

Echoing MeasureCamp London last week, Valentin was also co-presenting a similar session with Alice Jennifer Moore and Jono Alderson , but his time with Astrid instead, but also Camille Chaudet 's session at MeasureCamp Paris this year. It was very interesting to see the Swedish perspective.

Astrid highlighted the importance of what she called interoperable data, i.e. data sharing a similar structure and conducive to analysis.

A participant was doubtful that companies would comply with data protection regulations. GDPR had a tremendous impact, and although companies faced massive fines, especially Meta, many companies still operate under the radar, counting on the relevant data protection authorities' lack of resources to catch every breaching company. The regulators require a customer complaint before opening an investigation and considering the lack of resources, they will prioritise investigating companies that are the object of the most complaints. Although the participant did not raise these concerns, my overall impression was that he might be from Eastern Europe, where governments do not consider data privacy concerns as a high priority as in Western Europe and Sweden in our case. Some Eastern European countries are still grappling with corruption, and companies in these countries may get away with questionable practices regarding data use.

Another participant brought up the topic of AI and ChatGPT. I added that, for every 20 to 50 ChatGPT requests, the servers require 500ml of water to cool down the servers. Training the ChatGPT model required enough water to fill the cooling tower of an entire nuclear power station. I mentioned the topic of the Onion Model, a model I described in an article I published earlier this year. What technology affords us is vast, but we must comply with data protection regulators, reducing our options significantly. But this doesn't stop here. Considering the ecological footprint of data centres reduces our possibilities further still. There is what is technologically possible, legally enforced, and ecologically defensible.

A few participants touched upon gamification, i.e. rewarding people for doing good with data at an individual level. The idea has roots in positive reinforcement, a well-researched concept in psychology, pioneered by B. F. Skinner and benefitting a revival of sorts under the concept of?nudging?introduced by Cass Sunstein and Richard Thaler in their book?Nudge. Another participant suggested rewarding people financially if a company is making profits with the data that individuals have voluntarily shared with that company. Another participant mentioned the Brave browser earlier during the talk. Brave is based on the Chrome browser but rewards users for sharing data with third parties through a proprietary cryptocurrency called BAT token. Brave has been my primary browser on my personal laptop and mobile for many years. It's a bit controversial in our field, considering that it blocks all Digital Analytics tracking requests by default, and we would no longer have jobs if everybody were like me.

Astrid asked whether we could use data for good in politics, social justice and healthcare. Valentin shared with us the details of a discussion he had with Jono Alderson in London. They talked about how there should be a public database showing what the politicians have stood for in the past and their public record so that people can evaluate the promises these politicians are making today could contrast with their record.

I related the story of two ladies, Maggie Petrova and another participant, attending the same session at MeasureCamp London last week. Maggie is Bulgarian and during the pandemic, she helped with analysing electoral data and also children health data. She felt like it gave her a great sense of purpose. The other London session participant also mentioned DataKind , a UK charity where data analysts can put their skills to use for good causes. There must be similar charities in Sweden.

Another participant brought up an interesting angle about AI: AIs increasingly strive to help us succeed but at the expense of letting us make mistakes and learning from these mistakes. We do not want AI nannies.

I shared with the rest of the audience how this desire to do good with the data echoes beyond Sweden. At MeasureCamp London a week ago, there was a similar session, and at MeasureCamp Paris this year, and also shared with Siobhan S. in Greece and Stephane Hamel ???? in Canada. There is a desire to distance ourselves from the data abuse revealed by the Cambridge Analytics scandal. At MeasureCamp London, Nicholas Redding proposed the equivalent of a Hippocratic Oath where, as data practitioners, we refuse to collect and process data in unethical ways. Nicolas called this a Code of Conduct, and we will hear more soon.

Following comments from Camille Chaudet and Stephane Hamel after MeasureCamp Paris this year, there is an intense desire for Ethical Digital Analytics among digital analysts, but much less so from our potential clients and employers. Chris Wylie, the Cambridge Analytics scandal whistleblower, revealed how he received a deluge of job offers from companies that wanted him to do the same thing he did at Cambridge Analytics within the constraints of what the regulators tolerate, very much in the grey zone.

Stort tack till Chris Beardsley , Celine Derkert , Ashit Kumar , Philip Bromley , Anna Forsén , Josefin Kjellbris , Lisa Lindh Risberg , och alla studenter, troligtvis fr?n IHM Business School eller MEDIEINSTITUTET I SVERIGE AB eller b?da f?r organisation. Tack till Conversionista! f?r att ha oss alla, och ochs? alla andra sponsorer.

And now the hardest part of this article: Ruben Vezzoli , Marco Tognon , Claudio Ferrara , Moritz Bauer , Matt Gershoff , Ezequiel Boehler , Nathaniel Weiss , Simo Ahava , Mari Ahava , Piotr Korzeniowski , Emi Olausson Fourounjieva , Ingela West , Corie Tilly , Caroline Vidal , David Vallejo , Johan Olsson , Daniel Ford , Daniel Eckert , Victoria Smith , Luiza de Lange , Weiwei Liu-Schr?der , Ida Kjelldahl , Karoliina "Liina" K. , Lotta Holm , Peter Meyer , Kim Dahlroth , Negar Rayegani , Alexandre N. Koletsis , Johan Strand , Mae Tadena

For more articles like these, follow me on LinkedIn or X: @albangerome

#MeasureCamp #MeasureCampSTHLM #DigitalAnalytics #WAWCPH #CBUSWAW