“I remember getting my walls plastered, and when I asked the plasterer why they weren’t straight, he said I hadn’t specified that I wanted them straight – true story. I imagine it’s the same for marketers – why wasn’t their project built with performance in mind? It wasn’t part of the conversation.”
Andy Nicol, MD of Sputnik Digital, a digital transformation agency focusing on design and build, is an advocate of prioritising web performance amongst marketing teams.
As Sputnik has just published a guide to high performing infrastructure (gated form), we asked Nicol for some thoughts on web performance, the potential impact of AI on web development, and what his average day looks like…
Talk us through your average day…
AN: We’re fully remote, so our entire team work from home.
My day starts around 8am, clearing out my inbox, doing our bookkeeping, clearing my desk and updating my to-do list.
I join our all hands standup at 9am which takes around 10 minutes. It’s been a vital part of our culture since we left the office.
My simplistic view of running a successful agency involves three things – recruitment (finding passionate, curious and conscientious people); business development (finding clients who are aligned with our proposition and values), and production (the bit in the middle).
Much of my time is spent on the areas of ‘great people’ and ‘great clients’.
We’re fortunate to have nurtured a culture around best practice – our CTO, product owners, designers and developers are all motivated by building beautiful, high performing tools and applications. By allowing them to focus on innovation and quality, the rest of operations takes care of itself.
Things like HR and finance need hardly any attention if you can get the team, clients and production functioning smoothly.
I try and run every day, at lunchtime if my schedule allows. It breaks up my days, and gives me a chance to get some sunlight now that we don’t commute. 10k and a shower pretty much takes an hour, so lunch is normally something simple, and eaten at my desk.
What common issues crop up within organisations seeking to improve performance?
AN: I think the biggest issue has been lack of awareness of the importance of web performance.
Most of the focus for marketers is at the sharp end of paid media, UX, SEO and conversion.
I remember getting my walls plastered, and when I asked the plasterer why they weren’t straight, he said I hadn’t specified that I wanted them straight (true story). I imagine it’s the same for marketers – why wasn’t their project built with performance in mind? It wasn’t part of the conversation.
I think the biggest issue has been lack of awareness of the importance of web performance
Getting the engineering right helps meet all of the objectives – it reduces cost per acquisition (through an improved Quality Score), improves user experience (and therefore conversion), and lifts search rankings (through Lighthouse scores).
We have case studies where conversion has improved 19%, and entire portfolios of keywords have jumped 15 places.
In cases of legacy tech, it’s almost always possible to significantly improve performance. We’d start by looking at the front end and working back, through CDNs, static site generators, refactoring, etc.
Obligatory AI question: ChatGPT seems to be able to code – do you think performance diagnosis will become more of an automated process in the future?
AN: Elements of coding such as application scaffolding, monitoring and diagnosis can already be largely automated with frameworks and testing tools but ChatGPT can certainly be a useful tool in the toolbox and one that we’re employing to speed up some existing coding workflows such as creation of lambda functions to run integration tests and generating configuration files for an AWS infrastructure.
The concern with ChatGPT is that it’s generative. In other words, it makes stuff up. That’s great for poetry and song lyrics, but less good when you’re looking for the optimal answer to something.
It all depends on the training data, of course. But the average of the world’s code is…mediocre. The examples I’ve seen so far might pass for an MVP, but they aren’t optimised, production-ready applications.
Two reasons to be optimistic for AI to eventually fix performance, not just diagnose it:
Firstly, AI development is happening at a phenomenal pace. It was only a month ago critics were laughing at generated images having six fingers, but that’s no longer an issue. I think there will be game changing improvements in AI’s capabilities within the next 6-12 months; many of todays mistakes will be ironed out.
Secondly, much of the hype for generative AI is around image generation and large language models (LLM), the latter of which happens to be able to code.
The future will see more and more specialist AI products, trained with specific / expert models, and with their own UI.
The AI space is moving fast, and this answer will be out of date within a week. Yesterday, ChatGPT was unable to browse the web, so it wouldn’t be able to carry out a performance audit on a live site, for example. Today, ChatGPT announced their plugins, including a web browser which uses Bing to search, then clicks through and crawls the website. It’s only crawling text today, but now that it’s been unleashed, assessing code and performance of live websites is now possible.
What digital products/experiences do you particularly admire?
AN: There are so many great products out there, and I know it’s been said many times before, but I’m going to go for Monzo.
The user experience is a dream. Screens are clear and make great use of available real estate.
Your lasting impression is formed by the worst experience, and details like using magic links for logging in remove those bang-your-head-against-a-brick-wall moments.
And their technology blog provides an insight into the innovation and adherence to engineering best practices.
It feels as if, if Sputnik built a bank, it would very much resemble Monzo.
What’s next for Sputnik?
AN: We’re always scanning the horizon for disruptive technologies, and while we will continue to adapt to an evolving digital world, our focus will remain on combining intuitive user experience with “enterprise engineering”.
Once Apple and Meta release their next gen AR / VR products, UX / UI won’t rely on Figma and HTML pattern libraries.
Our values and approach will likely remain the same as they have for the last two decades, some of which has been distilled into our guides to high performing infrastructure and avoiding conflict when outsourcing complex digital projects.