Under the Hood: How We Built the Unbounce Landing Page Analyzer

7 Benefits of Hiring Digital Marketing Consultancy For Your Small Business in India
Spread the love
  • 2

Grace: Where did the idea to build a landing page analyzer come from?

Brian: During a brainstorm session a while back, an analyzer was one of the things that we thought would be a good bet. Our then-CMO, Jeremy, had been advocating for one for a while, and I had built something similar when I came to Unbounce. It was actually the first thing I built: a Landing Page Checklist Chrome Extension which got some traction (2000+ users at its highest point), so I already had an idea of what was possible. HubSpot’s website grader was also a big inspiration.

Fun fact: Building a tool that you offer for free can be an effective way of getting more leads and revenue without spending money on advertising. “When you advertise, the money you’re spending is what drives how much attention you get. Want more clicks? Spend more money… With inbound marketing, even after you stop producing new content, the old content can still drive ongoing visitors and leads.” — Dharmesh Shah, HubSpot Founder

Grace: Nice. So when you got to the MVP, what did you know that you had to build out first? (MVP stands for “minimum viable product,” or the bare bones of a product that you can build quickly but still have enough features to make your early adopters happy.)

Brian: First, I looked into all the areas where a landing page could be analyzed. This included copy, meta tags, page speed, image recognition, layout, design, colours, etc. You name it, I looked at it! Two things helped me out during this phase:

  • Our thought-leadership on landing pages and existing knowledge we had put into the world. Oli Gardner and others on the team have contributed a lot to the internet’s library of landing page design over the years. I scoured this content for how we could measure the performance of a landing page.
  • The fact that we had launched The Conversion Benchmark Report some time before helped with this a lot — we have millions of data points(!) on landing pages that we could use.
RELATED:  Why does your marketing team need a growth-hacking workshop?

Grace: That’s a lot of data.

Brian: Yeah, I wasn’t sure if everything that we wanted to analyze was possible from a technical standpoint, so I just started building out the tool locally. It didn’t look pretty at all (see below), but it helped prove what we could and couldn’t analyze on a landing page.

An early iteration of a landing page report (see the full pdf here).

Grace: It’s really come a long way. And for the technical aspects…

Brian: Like coding?

Grace: Yeah. For our more technically inclined readers, could you elaborate a bit on that?

Brian: Yup, the tool is a Flask app built in Python. Information on a landing page (like the URL, if a form is present or not, etc.) is posted to the application and then it returns the report.

It’s “serverless” using AWS Lambda to handle requests and spin up as needed — this makes it very easy to scale as it automatically adjusts based on traffic. For example, if we get a large influx of traffic from Product Hunt, it can scale to handle it.

Here are two other services we used to enhance the Landing Page Analyzer:

  • Google PageSpeed Insights API lets us measure page speed and suggest improvements. It powers the page speed section.
  • Urlbox.io is a screenshot rendering service that’s used for grabbing screenshots of each landing page.

Grace: Wow. It’s expected and all, but still kind of crazy to actually see how many working parts are in these sections. So for the report itself, how do the suggestions in there get generated?

Brian: It takes a number of parameters as inputs, like the factors we talked about before and also stuff like traffic sources and keywords. It then uses these inputs to analyze the page.

An example of one of the sections is Industry Performance, which takes the inputted conversion rate and industry, and compares it against our proprietary Conversion Benchmark Report data (a group of data scientists and conversion optimization experts at Unbounce used AI technology to analyze the actions of 74,551,421 visitors on 64,284 lead-gen landing pages created in the Unbounce platform).

RELATED:  What to Expect When Your Startup Hits a Growth Spurt

Another is SEO, which performs a basic SEO analysis. It checks for search engine accessibility, page title, meta description, and H1 tags.

Grace: Okay, let’s get down to the tough parts of building the analyzer. What were some of your main challenges?

Brian: Subjectiveness was definitely a big one. The Landing Page Analyzer is great at measuring quantifiable things like if certain tags are present, or counting words on a page, or checking for a mobile view — but not at measuring subjective things. Subjective things can include if the design fits the purpose, if the language matches the goal, and other things requiring a human eye.

Grace: Until AI takes over sometime in the future, at least.

Brian: That’s the thing. Tech is getting smarter all the time, but sometimes the human touch just can’t be beat. It’s important to strike a balance between the two, but the LPA certainly biases toward the former.

Categorizing the findings by industry was another challenge. We’ve got nine industries based on the Conversion Benchmark Report, but not everyone’s businesses fit neatly into those nine kinds of businesses. Your business might technically fit in an industry but be a small subset, or it might overlap across two or even three industries.

Grace: Awesome. Thanks for your time, Brian!

If you want to run a few of your landing pages through the Landing Page Analyzer for yourself and get a personalized report with tips to improve those pages further, head over to unbounce.com/landing-page-analyzer to get started!

Original article published here

Spread the love
  • 2
0 I like it
0 I don't like it

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.