Christopher Gutknecht: ‘Be lazy with manual labour task – use your energy to automate the process’

Date: 5th February 2019

Christopher Gutknecht: ‘Be lazy with manual labour task – use your energy to automate the process’

Christopher GutknechtWho is Christopher Gutknecht?
A happy dad and husband living in Munich who’s been climbing for over 20 years, doing PPC for over 10 years and been writing code for 5 years. I am currently Head of Online Marketing at Norisk Group, including three e-commerce agencies with overall 80 people. The performance marketing team has 11 people. Of all acquisition channels, I still love paid search most: there is no other channel with so many angles of direct interaction and potential to leverage technology to see immediate results.

When did you start with the automation of SEA(rch)?
I started around 2010 with Excel Macros, writing ugly VBA code to handle recurring tasks like rule-based bid management or checking cross-account query overlaps.

After teaching myself Javascript as a side-project throughout 2012/13, I really discovered my passion with Ads Scripts in 2014. Since then, I spend most of my PPC time on automation. Our motto is: Be lazy with manual labour task – use your energy to automate the process.

Throughout the last year, we kept seeing the limitations of Ad Scripts regarding more complex tasks so we started moving components over to Python via Google Cloud Functions.

Why should everyone start automating Search?
Many tasks can be described in a rule-based fashion and should be completed regularly. Not handing these tasks to an automated process means wasting time – and time is the most valuable asset which is dearly needed for strategic thinking and research.

On the other hand, automation is a never-ending process of maturity that takes years to master. It is important to teach PPC youngsters the automation mindset very early in their careers to not waste time. Always be lazy with manual labour tasks.

Where do you find inspiration for automation and innovation?
Most of the use cases come from our client projects and their challenges – it is essential to listen, analyse and identify what is needed. Then seeing parallels across multiple projects means there is potential to save time and improve results.

The other 30% of inspiration comes from blog posts, conferences and discussions with other practitioners in the field. We have a constantly growing backlog and keep coming up with more ideas than we can execute. The last idea was to pull a Sistrix SERP competitor list via API to automate exploratory campaign buildouts for competitor bidding.

What is your biggest success with automation? (And if any 😉 biggest fail / pitfalls?)
Our main successes at Norisk were building our tools for shopping structure sync, feedbased campaign building, auto-adding converting search queries and – more recently – automating the process of mapping retailer promotions to ad extensions.

Some fails include blindly relying on these automated processes and assume they’ll always work – I like to compare many of these processes to a one-year toddler that stumbles at the first little irregularity. It takes many iterations of improvement.

How do you prioritize your automation challenges?
We try to take a similar and systematic approach as in CRO projects: potential, impact and ease.

But very often it’s less systematic:
For new tools, it’s mostly the ‘log-thrown-back-in-the-lake” analogy that drives our decisions. If a topic arrives at your shore, throw it back into lake and make a note. If it keeps coming back to your shore, in other projects, you know it’s there to stay. 🙂
For tools close to completion, it’s client urgency that drives the last 20% to initial release.

How do you decide when to build your own Ad(word)s scripts or use software from a vendor?
If we can fully describe the target behaviour of a system, then the idea of building a tool is worth exploring. If we’re missing critical parts of knowledge, like for bid management or anomaly detection, we search for vendors. To be honest: for most automation challenges in PPC there are no vendors, as many tasks are specific to an industry or even an organisation.

You work at an agency, does your company have a different team/organisation structure than other agencies regarding innovation in Search? What advise could you give an in-house team on investing in innovation and automation?
We have a core team of PPC managers working very closely with a marketing technologist (& me) who writes code, be it prototyping or feature releases. If a PPC wants to get started with own little prototypes, everybody is welcome. It is absolutely critical to build a culture of ideas, prototyping, feedback and continuous development on existing tools. Ideally, there are no or little external dependencies to IT or DEV: self-empowerment is the key.

And don’t say: We don’t have the for time this. Make time for it. That time is typically wasted later on when things have to be done manually – having a army of interns to do manual labour is also a waste of resources because more and more of these youngsters know how to code.

Do you expect Google (and Bing) to ultimately automate their complete (advertising) platform?
Best-in-class PPC will never be entirely automated. Full automation will be the equivalent of a having a robot serve a deep-frozen pizza: It’s zero effort, it will serve the purpose, but could have done so much more. Excellent PPC bring many dimensions together, it’s like a brilliant, creative chef rethinking her/his approach every week.

Google and Bing will only offer tools that work for many advertisers or key industries. The real potential lies in automating specific business processes that drive user value – this usually requires creativity and company-specific data.

One key challenge for performance marketers will be to generate better results with own datasets and custom processes than Google with its generic – though increasingly smarter – deep-learning based systems.

What are the trends in search that we can expect to influence us in the near future?
There will be more services to explore the relationships between input data and outputs, thus guiding PPC managers decisions: the way smart bidding handles audience and device modifiers is one example for this.

The key here will be connecting data sources in a way that ML-based systems can learn these relationships – this is where cloud platforms can help.

A big theme will be rise of the technical marketer, who is less and less dependent on IT to build the needed toolset. We need cross-functional teams for this, but I don’t think we need full-time developers.

What can people expect from your session at friends of search?
I will show attendants how to actually get started with entry-level machine learning tools in a PPC context. There are a few tools that like Google Colab, Cloud Functions and ML APIs that don’t require real ML knowledge to get first results. You learn by playing around with the use cases that matter to you and follow up with the theory later. Cloud tools are becoming more and more beginner-friendly, now is a great time to get started!

Why should marketers visit Friends of Search?
In my opinion, conferences are the best source of inspiration. Inspiration is what makes our jobs more fun and drives us to accomplish the next bigger and more complex goal. My first conference is what gave a long-term drive to aim to rise above mediocrity and make a mark in PPC.

Christopher will speak at Friends of Search on February 6. Get your tickets here.