The document discusses how analysts can avoid being "report monkeys" and instead focus on more strategic work. It recommends that analysts:
1. Get organizational buy-in by creating a measurement plan, telling stories with data, and doing branding for their team.
2. Address implementation pains by using tag managers and respecting IT processes.
3. Prevent weird data issues by understanding how analytics works, documenting changes, and using debugging tools.
4. Avoid time-consuming recurring reports by automating data collection and reporting through tools like the Google Analytics spreadsheet add-on and dashboards.
12. #GAUCBE
The biggest frustrations of an analyst…
#1: You are not heard
#4: Time consuming recurring reports
#3: Weird data
#2: Implementation pains
13. #GAUCBE
Tired of being a Data Monkey?
Join the monkey revolution!
Register for our
“Important steps to become
a better Analyst”
mailing
(www.humix.be/GAUC)
Image copyright Banksy
15. #GAUCBE
#1: You are not heard
You spend your days
generating report after
report
but you are not heard…
You need Organizational
buy in
16. #GAUCBE
#1: You are not heard: Organizational Buy In
How to get
organizational buy in.
1. Create a
Measurement Plan
2. Tell Stories with data
3. Do your branding
19. #GAUCBE
#1: You are not heard: Measurement
Plan
The businessgoals we would like to reach:
Webshop is the store with the largest revenue
Which questions take our closer to our goals:
Are we convincing visitors to buy online?
What do we need to measure:
# visits, # visits with product views, # visitis with transactions
What metrics & dimensions support our KPI’s:
Traffic channels, device type, user demographics
Example of Hubo
What metrics & dimensions support our
KPI’s
20. #GAUCBE
#1: You are not heard: Measurement
Plan
“You have a Measurement Plan. Now, what?”
21. #GAUCBE
#1: You are not heard: Measurement Plan
Share your Measurement Plan
within the organization
• Ask people feedback & approval
• Hold them accountable for the KPI’s
relevant for their department
Use it as a development roadmap
for data that’s not yet available
22. #GAUCBE
#1: You are not heard: Measurement Plan
Remind yourself that it’s
not written in stone. It’s a
continuous effort
• A measurement plan
should change along
with the organization
• Learn, improve, evolve
23. #GAUCBE
#1: You are not heard: Measurement Plan
How to handle measure
requests?
• Create a form for requests
(make the submitter think
before a request)
• Ask them to score on added
value for the organization
• Combine each request with
the goal-weight of the
impacted objective to
prioritize your work
24. #GAUCBE
#1: You are not heard: Tell Stories with
data
the numbers: last 7 days
25. #GAUCBE
the numbers: last 7 days
• Is this good?
• Should it be better?
• What are sessions?
• …
#1: You are not heard: Tell Stories with data
26. #GAUCBE
What are sessions
> human voice added
“Last 7 days we had 199 visits on the
site.”
#1: You are not heard: Tell Stories with data
27. #GAUCBE
Is this good or bad?
> benchmark added
“Last 7 days we had 199 visits on the
site.
This is -9.55% from our target.”
#1: You are not heard: Tell Stories with data
28. #GAUCBE
Are we improving?
> benchmarked vs previous
period
“Last 7 days we had 199 visits on the
site.
This is -9.55% from our target.
Compared to previous period of 7
days, this is a -8.3% difference.”
#1: You are not heard: Tell Stories with data
29. #GAUCBE
I’m more like a visual
kind of person….?
> Add big numbers
#1: You are not heard: Tell Stories with data
30. #GAUCBE
Is this last week or
last 7 days?
> Add period
#1: You are not heard: Tell Stories with data
31. #GAUCBE
I’m like a really visual
person?
> Add Gauge
#1: You are not heard: Tell Stories with data
32. #GAUCBE
Did we have good or bad
days?
> Add Sparkline
#1: You are not heard: Tell Stories with data
34. #GAUCBE
Context is crucial for story telling:
“My marketing agency is doing a good job”
#1: You are not heard: Tell Stories with data
35. #GAUCBE
Context is crucial for story telling:
…combined with “is it getting colder?”
#1: You are not heard: Tell Stories with data
36. #GAUCBE
#1: You are not heard: Do your branding
Promote
• yourself
• your work
• your team
37. #GAUCBE
Create a brand
Make your team known within
the organization.
Come up with a name & logo
(even if it’s ugly).
#1: You are not heard: Do your branding
38. #GAUCBE
#1: You are not heard: Do your branding
Big screen dashboards
make your work visible
39. #GAUCBE
#1: You are not heard: Do your branding
Promote your work
Create an internal newsletter
to share insights &
experiment outcomes
Helps in establishing a testing
culture within the organization
40. #GAUCBE
Create a central
knowledge base:
a place where al insights,
whitepapers, … are easily
findable for your “clients”
#1: You are not heard: Do your branding
41. #GAUCBE
#1: You are not heard: Do your branding
Take a supportive attitude
to your “clients”:
how can I make the others
shine through my data
results?
Deliver bad news in a “shit
sandwich”.
<Good – Less Good – Good>
45. #GAUCBE
#2 Implementation pains: Tag Manager
Two words:
TAG MANAGER
IT people love Tag Manager
• They see you less often
• You can help yourself most of the
time
• Versioning
IT people hate Tag Manager
• Wait what? You can insert code
on ANY page of the whole
website?
46. #GAUCBE
#2 Implementation pains: Tag Manager
Our secret recipe: gtm.blacklist
Define clearly what type of tags can
and can not be published without IT
involvement.
And off course…
• Respect their sprints/ticketing system
• Educate them on Analytics
48. #GAUCBE
“Analysts spend as much
as 40 percent of their time
validating data”
- Forrester research
#3: Weird Data
49. #GAUCBE
#3: Weird Data: High Traffic, weird source
A city that is a
bigger provider
than Telenet or
Proximus??
50. #GAUCBE
#3: Weird Data: Bounce rate gone, auto event firing
Somebody made an event + custom
metric…that auto-fired on every
page.
Bye, bye bouncerate.
(Kudos for documenting it)
51. #GAUCBE
#3: Weird Data: A lot of Visits, cross domain errors
Cross domain tracking not set up
correct:
New session triggered
www.mysite.com
blog.mysite.com/home
blog.mysite.com/page2
www.mysite.com/
blog.mysite.com/
52. #GAUCBE
#3: Weird Data: not understanding hit vs session vs user level dimensions
More Users than sessions. That can’t
be correct … but it is...
53. #GAUCBE
#3: Weird Data: how to prevent?
Implementation issues and how to solve them
1. Understand how Google Analytics works
2. You Everybody needs to document changes
3. Data debugging. Know your tools
54. #GAUCBE
#3: Weird Data: Debugging tools
Wasp
Chrome extension
Tag
Assistant
Chrome extension
Observe
Point
Chrome extension
GA Debug
Chrome extension
Screaming
frog
crawler
Charles
web debug proxy
56. #GAUCBE
Must know!!
Impressive!
Regular Expressions
Measurement protocol
Debugging tools
Analytics Interface
Tag Management
Excel/Sheets wizard
Common Sense
Analytics API’s + scraping
Dashboarding
Javascript, Xpath,… Good job!!
#3: Weird Data: Knowledge levels of an analyst
57. #GAUCBE
The biggest frustrations of an analyst…
#1: You are not heard
#4: Time consuming recurring reports
#3: Weird data
#2: Implementation pains
58. #GAUCBE
When you still are that
monkey who:
1. Goes into Google
Analytics/Facebook
Insights/<insert tool here> …
2. ...and exports a bunch of data
3. ...to cut/paste/replace last’s
month data
4. ...and email it to your collegues
#4: Recurring reports
60. #GAUCBE
Try this.
Just stop sending those reports. Kill them all.
You’ll be amazed how much complaints you’ll get.
Too often: zero to none.
#4: Recurring reports: Break free!
61. #GAUCBE
#4: Recurring reports: Automate
Do not export – import: Automate
GA Spreadsheet Add-on
Google Sheets add on
SuperMetrics
Google Sheets add on
62. #GAUCBE
#4: Recurring reports: Automate
Supermetrics uses 32
different sources for
your data
(paying version)
You only need to adapt start
& end date + refresh the
queries (and that can also
be automated)
63. #GAUCBE
#4: Recurring reports: Automate
Get data that is online,
but not even on your
own site (weather, …)
Quick example:
GAUC attendees list
64. #GAUCBE
#4: Recurring reports: Automate
Quick example:
HTML container :
<div class="description-
inner">
Google Sheets Formula
=IMPORTXML("http://www.gauc.
be/attendees","//div[@class='des
cription-inner']")
65. #GAUCBE
#4: Recurring reports: Automate
Learn the API:
Use the query explorer
Craft your own data sets
(example: the Google
Analytics Demo Account)
66. #GAUCBE
#4: Recurring reports: Automate
Dashboards
Telling stories with data
works better in custom
environments
(We like Klipfolio & Tableau)
67. #GAUCBE
#4: Recurring reports: Automate
Dashboards
Because the internet doesn’t
stop at Google data
(klipfolio datasources)
69. #GAUCBE
“we like it in this format
because we are used to it”
Do not comply:
• Try to tell a story with data
• Mix plain text with graphics
where needed
#4: Recurring reports: Excel
Dino’s
72. #GAUCBE
Make your organization data driven
#1: Make yourself & your work known within the organization
#2: Define clear roles & responsibilities together with IT
73. #GAUCBE
Make your organization data driven
#1: Make yourself & your work known within the organization
#2: Define clear roles & responsibilities together with IT
#3: Understand the data collection
74. #GAUCBE
Make your organization data driven
#4: Automate reporting
#1: Make yourself & your work known within the organization
#2: Define clear roles & responsibilities together with IT
#3: Understand the data collection
JENTE: you can not build a data driven organization with a bunch of data monkeys.
Not talk about the organizational processes needed for data driven decision makingFocus on initiatives that you, within your role as analyst, can take to contribute to the digital maturity.
JENTE:But first, let me introducing ourselves. As Age goes before Beauty, I’ll start with my colleague Yves.
Yves is a Digital Strategist and has been working with Google Analytics for… well very long.One of the dinosaurs of the Belgian Internet industry.
YVES: This is Jente, a typical millinial kid: those young folks with high expectations of their job and a lack of respect of my age....
YVES: We both work at Humix.
Humix is a knowledge center of different domain experts. We focus on website optimization.
YVES: As Humix experts , we always work on site with our clients. So we experience the digital transformation challenges of these clients in or day to day jobs.
It’s this experience that we would like to share with you today.
JENTE: 2012. Harvard Business Review: the sexiest job of the 21st century.Extremely motivated.
Harvard’s definition of “sexy” was different than mine. You are hired to help transform the way they are working. Understand Need and have ambition
JENTE: Wrong!
Reality of day-to-day tasks feels like monkey work…
After 4 months, you realize that nothing has changed since your start. So: no Groupies and No Authority. A harsh reality check. Ended up as a bored-out report monkey.
YVES: That sounds very familiar
I bet that many people in the audience have been confronted with that scenario.To often, we end up contributing to the “business as usual” within an organisation instead of driving change.
Events like the GAUC are often good networking opportunities to meet other webanalysts…And when you combine meeting peers with drinks, you end up nagging about things that frustrate you. In all those conversations, the four same frustrations come up…
YVES: You are not heard.
You create reports and analyses but nobody is paying attention.
YVES:
You are trying to get that additional analytics tracking script on the website since the beginning of the year and still no result…
YVES: Your colleagues don’t trust the data. It doesn’t make sense.
YVES:
Deep dive?
A/B testing experiments?
Nobody got time for that when your generating reports all day.
YVES: When you encounter these frustrations yourself, you have two options:
accept that your job is depressing and continue working on auto-pilot and be a victim:
it’s not your fault is it?
Or you try to change your organisation from within your role as an analyst
We have chosen the second option and will share our methods during this presentation.
As our time is limited today, we will set-up a mailing to go more in depth on some of the topics discussed here today.
It would be awesome if you would all subscribe for it!
You can send us extra topics if you like!
YVES: Let’s start with what is probably the biggest frustration of all.
YVES: What does the average day of a data monkey look like?open up Google Analytics and start exporting the numbers of last week.
organise the data in Excel and
create graphs that you can use in a PowerPoint.
You notice that organic traffic is down (again) and write down some good action points to turn this downward trend around.
You send your reports to the marketing manager.
You’re sure that this will wake them up: it’s clear that they must act on this!
However, next week: same story all over. Just as the week after. And after that.
YVES: What you need is: Organizational buy in.
You need to deliver reports that influence people’s job.
It should be clear who will be hold accountable for the KPI’s that you are communicating.
In order to achieve this, you need to involve all stakeholders from the start. Create a measurement plan together with them. Communicata your data in a form that’s understandable to them and …
make sure that they all know that you exist and what you do.
JENTE:
New project starts with a “Goals Exercise”.Rather simple: stakeholders together, Post-Its, how to define succes?, no management speak.
Cluster Ideas into groups
JENTE
Ask your stakeholders to assign a weight to each goal: spark the discussion!
Political tensions within the organization.
End with a simple Goal Overview that’s aligned and prioritized.
JENTE:
Starting point for your Measurement Plan.
Multiple approaches, our approach, example of HUBO.
Translate each business goal in questions.
help us in answering those questions?
Which dimensions define actionable recommendations?
JENTE:
how will this measurement plan contribute to organizational buy-in?
JENTE:
Based on the input of stakeholders. Ask their feedback.
Communicate: clear for everyone what is important. Tak into account when evaluating employees.
KPI’s that are not available yet. Work out a roadmap in the long run.
JENTE:
Not written in stone. The market is changing. Grow along with the organization.
Don’t be afraid to admit that certain KPI’s aren’t so ‘key. Evaluate and improve!
JENTE:Start requesting additional things. Can be overwhelming.
Ask people to submit their request through a form. Forces them to think.
By combining these with the goal-weight you can prioritize.
YVES:
Having everyone alligned on the KPI’s, does not yet make people actually read your reports.
You need to package your reports within
a story
to make it attractive and understandable for your report consumers.
Rethink your existing report tables:
how can you package the same information in an easy-digestable format?
YVES: You all have seen the numbers, could you tell me if our site was doing well in the last 7 days?
No? Well, let’s try something different…
YVES: Be aware that you GA lingo is not understood by “normal people”.
YVES: When is a metric good? What did we expect that it would be?
Benchmarks facilitate the evaluation.
YVES: People don’t remember by head the numbers that you send them last week. Provide them with those information!
YVES: A bunch of text is hard to interpret in the blink of an eye.
Visualisations help interpretation and point out what’s important.
YVES: Avoid confusing around your numbers.
You can’t predict how people will interpret “last 7 days”. Make sure that you add definitions.
YVES: You can even visualise more text: use a gauge meter when comparing to a target.
YVES: Trends can easily be visualised with sparklines. This puts your numbers within their context.
YVES: Is your audience familiar with website metrics? If not, don’t be afraid to add a dummy proof explanation.There are no stupid questions, remember…
YVES: Context is crucial when you’re trying to tell a story. Numbers that lack context can be misleading.
A good example is this website visits report of an energy supplier. They had offline campaigns running and were certain that the advertising agency was doing a good job.
However, I noticed that the commercial boosts were not always followed by traffic peaks…
YVES: So, I decided to add some context to the numbers.
By plotting the minimum temperature against the site visits, I discovered a negative correlation :
temperature goes down, site visits goes up.
So, cold temperatures made people go online and search information on energy prices.
Not so much the radio ads.
JENTE:
Make people curious. Make yourself, your team, and your work noticed.
Your analyses should not be limited to boardrooms and C-level mailboxes.
JENTE:
Promote your team as you would with a brand.
We have the D-Devils, short for dashboard devils. Reference to epic nineties music bands.
JENTE:
Design dashboards for big tv screens. People can’t resist.
HUBO example.
JENTE:
Newsletter: additional information.
See what works best: existing newsletter or a specific mailing.
Great for communicating A/B testing results. People like reading these kind of things.
JENTE:
No time to read your reports/learnings when you publish them.
A central knowledge base, published on a closed URL, is great for these kind of things.
Frontify used at VRT News. But often also Confluence etc.
JENTE:
Promote yourself, let people experience what’s in it for them. Don’t be the bogeyman!
Try working the other way. Find out how you can help make them shine.
“you are the expert on this, can you help me explain this data?”
“I can look this up for you…”
Don’t be the bad-news guy.
YVES:
Our second frustration is the struggle that we often have to get additional code on the website. Most of the web analysts have a marketing background and they have no idea on how IT departments work.
Yves:
They need to keep things running, not measuring
Often they lack experience in more advanced set-ups.Google Analytics Developer pages are their manual (and that can be confusing sometimes)
That’s not in their budget
Which departement are you from?
YVES: Two words: TAG MANAGER.
YVES: Tag Manager is both loved and hated by IT.
It lowers their work load, but it also makes the site vulnerable to human errors. And IT guys just don’t trust marketing dudes with code.
YVES: We found a way to set the IT guys at ease AND to enable the marketing people to get tags on the website themselves.
You can blacklist certain type of tags (for instance all custom scripts) within the GTM DataLayer. These kind of tags will never be fired by GTM.
Off course, mutual understandment of how the other is doing his job helps also in getting along.
JENTE:
Another frustration can be the mistrust in the available data.
Therefore: weird data bumps can be a real threat to your trustworthiness.
Jente:
Forrester research has estimated that …
That is an insane amount! I bet you did not become a web analysts with the idea of debugging all day?
I’ll walk you through some time-consuming examples of weird-data.
JENTE:
A Belgian City. They were experiencing high bounce rates, short sessions etc.
the internal network was the biggest provider.
The City Hall has an open WIFI network, that is being used by lots of people on the town square.
Started on the homepage of the website.
JENTE:
bounce rate dropped to 0 overnight.
caused by someone that experimented with an event, that auto fired on every page.
It took them a month to notice something was wrong.
JENTE:
An unexpected peak in the amount of sessions is always something to be suspicious about.
huge amount of new traffic, but a change in the URL structure caused cross domain tracking issues. The blog moves to a subdomain, without involving the Analytics team.
JENTE:
Keep in mind the levels on which Analytics captures data: hit, session and user. Don’t mix these!
Weird results are caused by yourself!
Subscribe to our mailing. These examples, and others, will be discussed in more detail.
JENTE:
Understanding how Google Analytics works is already a big step in preventing data issues.
But when things go wrong, make sure that you know how to debug.
When multiple people are working within the same environment, documenting changes is crucial.
JENTE:
This is an overview of the debugging tools that we use for Analytics implementations.
JENTE:
When I say: understand how GA works, I mean: know what’s happening beyond the interface.
Jente:
Be aware, that these are the things that people expect from you as an analyst.
Don’t state that you’re a senior Google Analytics expert, if you have worked in the interface for 5 years. No, you should know your way around API’s and dashboarding before you can call yourself that.
YVES: Our last frustration focuses on those skills.
YVES:
Stop being the report monkey.
YVES: Break free!
YVES: Kill your reports and see what happens.
Discover who is reading your reports and work together with these people.
YVES:
Instead of creating monthly reports manually, try to automate them.
Every report that needs to be delivered on a regular basis can be automated.
Don’t be afraid: it’s not rocket science! Thanks to some handy plugins.
YVES:
we’re a fan of Supermetrics.
YVES:
Let me show you a quick example of what you can do with scraping.
YVES:
All you are within my Google Sheet.
Yves:
Lean the API if you want the data combined in your formatThis is BTW the Google Demo account you can all access. See , for example, wat enhanced ecom looks like, even if you don’t have access to a “real” site
YVES:
Stay away from manual work within dashboard. Make sure everything is automated and refreshes. They help you communicate your stories in a automated way!
+( and they look so good! ) Mhhh…Dataporn!
We are big fans of Klipfolio & Tableau. But are also eager to get our hands on Google’s new Data Studio.
YVES: an overview of the data sources of Klipfolio.
YVES:
So you can’t automate reports, because they don’t flag weird data?
They can…but its more complex?
Why don’t use Intelligence events for that?
Yves:
Never accept a request that states: can you rebuild this Excel in your Dashboard…because we are used to the data in this way.
Don’t try to recreate bad reports with good data.
JENTE
That’s it: these we’re our frustrations and how we tackled them in our projects.
Let’s go quickly over our key-take-a-ways to summarize.
JENTE
Make yourself & your work known within the organization.
JENTE
Define clear roles & responsibilities together with IT
JENTE
Understand how the data is being collected and know how to spot issues.
JENTE
Don’t be a report monkey, automate as much as possible and make time for the interesting stuff.
YVES:
Hope that these tips were usefull to you.Let us know if you have different experiences with trying to make an organization more data driven.We’d love to hear them during the drinks afterwards!