We believe active attendees of MeasureCamp love contributing to the greater digital community apart from their analytical peers. It is about the community. We have thus contacted 27 members of PPC, SEO, UX, and product management communities and asked them simply what they perceived as top challenges for data analysts. Even just by stating such challenges, as seen by other roles, we assumed some participants would like to take such as opportunities to shine by solving that or sharing their experience with similar cases. What follows is the compilation of answers from 15 respondents.
Putting data into practice
About half of our respondents, including Petra Větrovská, Adam Fendrych and Ondřej Ilinčev, are dissatisfied with the way how people are working with data in general. Petr Štedrý (2FRESH agency) mentioned the challenge of providing proper access to data to product managers so that they can better decide. And also how to educate to ask more powerful questions.
Respondents working with multiple clients observe some patterns in the way how to utilize data as a tool for investment decisions. Ondřej Ilinčev put it nicely into three steps:
- Make it clear what is currently collected. It is almost guaranteed that a client would not have all Google Analytics goals properly setup, something missing in enhanced e-commerce implementation, missing integrations with Optimize or Hotjar, etc.
- Present the results of the cooperation in a valuable manner, e.g. using Google Data Studio dashboard.
- Adam Fendrych (Rohlík.cz) would welcome tips on trainings, online courses, articles or books on the subject of extracting data insights.
- Vojtěch Mikula (Picards) asks data analysts to think about the actual results for a better actionability: when an SEO consultant creates a keyword analysis with 46 tables, the actual result should be a list of insights and recommendations that are specific for various roles. It is important to differentiate between data analysts who take care of data collection, and those who can dig a bit deeper to discover insights that are beneficial for the business.
- Educate the client on how to monitor the dashboards frequently and act upon the results. Jan Nachtigal mentioned the problem lies in user-friendliness of customer dashboards: they need to be easy to understand for users who don’t speak the same language as tech-savvy digital marketers.
Data collection challenges
Clean data of traffic sources is not always a standard. Pavel Ungr would welcome an overview and tips on an exact way to measure traffic from Google News, Google My Business, and Google Discovery Feed.
Jan Zdarsa (Google, United Arab Emirates) is worried about the impact of measurement via cookies throughout recent changes in the browser ecosystem. ITP/ETP now prevent certain tracking, however cross-device and cross-environment behavior (web x app) measurement has been limited for a long time. The longer purchase decision process, the more complex the situation becomes.
- Marketers thus tend to rely on low-funnel activities such as remarketing, as it brings the clearest results. Yet, as Jan points out, for proper marketing processes, one needs to model and predict conversions during the first session based on engagement quality, and also not to rely on cookie-based attribution models as these are similarly bad as the last click model.
- To correctly work with these predictions and hypotheses on better attribution models, marketers and analysts should educate themselves in cloud services, data science, statistics, etc. Similarly, Martin Michálek (VzhůruDolů) would appreciate basic measurement trainings and statistics 101.
- Marketers then need to follow-up with systematic hypothesis testing and verifications — incrementality testing has already been part of Google, Bing, and Facebook advertising platforms for several years + there are nice technical solutions like Causal Impact. Would you know how to design and run such marketing experiments?
- If you read annual reports of Criteo, you might get a feeling of how all current topics like ITP/ETP/GDPR/ePrivacy can impact audiences and targeting capabilities. To not cry over spilled milk, this creates a nice challenge to fully benefit from a good old contextual targeting that PPC managers have had good experience with.
Segmentation and audiences
Petr Štědrý described a challenge for UX designers who would like to integrate behavioral data used for target group definition with digital analytics tools. This type of segmentation should be more useful than traditional socio-demographic attributes that tend not to correlate with the practical results.
Matěj Slavík also mentioned that for audience segmentation (e.g. CRM segments), one needs to think ahead because many advertising platforms do not allow the retrospective definition of audiences for targeting.
Multichannel Customer Journey
Various respondents called out for a session on multichannel attribution, especially among online and offline channels.
For a more specific analysis of customer journey, Martin Zítek described a challenge of predicting the next touchpoint that would increase a probability a user would complete a conversion path successfully.
When the journey is not the simplest, some businesses focus more on lead management. Jakub Kašparů mentioned this presented a big challenge for CRM processes but also for digital marketing from proper measurement up to bidding decisions in PPC.
SEO and PPC
Digital marketing topics in PPC and SEO are pretty straightforward:
- Martin Zítek described a challenge of PPC and SEO cannibalization analysis with an actionable output of ideas when to advertise and which keywords should be left for organic traffic. Does anyone have experience with an automated solution with search query level data? Martin Šimko sees another cannibalization challenge in the incrementality of keywords with strong organic positions.
- Speaking of the collaboration among PPC and SEO, Martin Zítek also mentioned a prediction of sweet spots on a price elasticity curves for PPC bidding, where segmented data of SEO traffic should help identify what would maximize profit or revenues.
Working with keyword-level data presents various challenges:
- Identifying and predicting growth areas for organic traffic, as Jaroslav Hlavinka (Seznam.cz) called out.
- As a last tip from Martin Zítek, keywords or search queries could be automatically classified to See-Think-Do-Care framework according to their significance and relevance.
Automation and scalability are key topics for any PPC or SEO manager. Filip Podstavec (Marketing Miner) described a challenge of creating meta description values according to the analysis of page content and NLP techniques. Results from a simple NLP model that take as an input a category page, several sub-items, and outputs a 300-chars description can have greater performance than descriptions created either manually or in bulk using rudimentary tools.
Filip also recommended working with NLP on a pre-processing of content for analytical purposes. This way very general pronouns could be substituted by the specific terms, e.g. instead of saying “I wrote a homework and my dog ate it” it would output “I wrote a homework and my dog ate the homework”. Such output can provide more specific analysis using TF-IDF technique.
What a load of ideas! Many thanks to the respondents.
Does any of these topics ring the bell to you? We’re be delighted if you can share your experience and knowledge during MeasureCamp on September 7th in Prague.