Growth Engineering Case Studies

This post is the third in a series on Growth Engineering:

• Intro to Growth Engineering
• Growth Engineering Services
• Growth Engineering Case Studies (this post)
• Becoming a Growth Engineer

---

It's hard sometimes to visualize what Growth Engineering can do for your marketing team. Even if you understand the various Growth Engineering Services, it can still be hard to know what type of results you'd expect from a Growth Engineering project, or how to think about the strategy of approaching one.

To clear this up I've put together a master list of the various Growth Engineering projects I've been part of that were deemed successful. In each case study I've included the problem we were trying to solve, what strategy we used to solve it, the services this would relate to and most importantly, the results we achieved. This is anonymized to protect client information, but you should get a good idea of what worked and for what types of company.

Cohort Analysis for Mobile Banking App

• Problem: It took several weeks for all new users to finish bank account setup, but we needed quicker feedback to optimize £10,000+ spend per day.
• Strategy: Log performance by campaign over 28-day intervals for cohort analysis, letting us estimate the share of future conversions to predict CPA.
• Services: API connectors, Cohort analysis, Custom reporting 
• Stack: Facebook Ads API, Google Cloud Function, GSheets, Google Data Studio
• Results: We determined ~30% of conversions happened on day 1, so optimizing daily budgets we could estimate future CPA by dividing by 3.33.

ML Creative Tagging for a Fortune 500 Travel Company

• Problem: Difficulty in understanding what was working at scale, while testing 100s of ad creatives and spending €100k+ on LinkedIn ads per month.
• Strategy: Creative analysis using machine learning image classification, to be able to show performance segmented by what was in the images.
• Services: Image Classification, Custom reporting
• Stack: Google Colab, Google Vision, NLTK, Pandas, GSheets, Google Data Studio
• Results: Images with ‘furniture’ tags had a 66% lower CPL than ‘suitcase’ tags, meaning travel managers in offices outperformed business travelers in airports.

Automated Campaigns for a Global Events & Attractions Publisher

• Problem: Writers were responsible for building 100s of Facebook ad campaigns themselves as deals came in, often publishing late with no time to optimize. 
• Strategy: Connect to the product feed and automatically build campaigns based on pre-defined templates, turning on and off based on inventory.
• Services: API connectors, Campaign Builders, Ad Optimization
• Stack: Facebook Ads API, Google Cloud Function
• Results: 1000s of deals published on Facebook ads per month with no further involvement needed from the publishing team.

Web Scraping for a Seed-Stage Risk Modelling Startup

• Problem: Needed to reach PPC agencies that are active content marketers, with a cold email outreach campaign. 
• Strategy: Build a web scraper to download a list of 50,000 digital agencies globally with average customer size and focus areas.
• Services: Web scraping, Data cleaning
• Stack: Google Colab, Beautiful Soup
• Results: Extracted a list of 3,000 agencies that fit the profile for targeted outreach.

Test Results Meta-Analysis with a Mid-Size Marketing Agency

• Problem: The agency had run over 4,500 experiments on behalf of clients, but learnings were siloed and weren’t widely distributed across the team.
• Strategy: Conduct a meta analysis of experiment documentation to determine what we could learn about odds of success by type of test.
• Services: Data Cleaning, Exploratory Data Analysis, Natural Language Processing
• Stack: Airtable API, N-Gram, Pandas, Jupyter Notebook
• Results: Identified ~300 unique tactics and ~1,200 properly concluded tests, with a 44% success rate. Organic tactics (+7%), tests needing a developer (+20%) and seed stage / bootstrapped startups (+60%) had higher success rates.

Annual Budgeting for a Billion-Dollar Public AdTech Company

• Problem: Needed to build a compelling business case to support a new SMB-focused go to market strategy, and secure a sizeable budget allocation.
• Strategy: Research performance benchmarks across channels, using Fermi decompositions and Monte-carlo simulations to calculate various scenarios.
• Services: Monte-carlo simulations
• Stack: Excel
• Results: Modelled $1.2m in annual SMB marketing budget, broken down by channel, activity and vendor, with ‘what-if’ simulations for potential outcomes.

E-Commerce Tracking for a Fortune 1000 Fast Food Chain

• Problem: No e-commerce analytics tracking set up the night before a major product launch, no in-house expertise available at late notice.
• Strategy: Implement Google Analytics Enhanced E-Commerce tracking via Google Tag Manager in time for launch.
• Services: Tag Management
• Stack: Google Tag Manager, Google Analytics
• Results: Full tracking implementation done on development server, pushed to production just before midnight.

Dynamic Landing Pages for a Startup Trade Services Marketplace

• Problem: Needed to improve conversion rate of traffic sent to landing pages and communicate different value propositions per trade service.
• Strategy: Build a dynamic landing page script that modified the header and hero text based on the source of the traffic.
• Services: Conversion optimization
• Stack: Google Optimize
• Results: Decrease of 62% in Cost per Lead due to increased conversion rate.

GDPR-Compliant Tracking Plan for a New European Mobile Bank

• Problem: Launching a new banking app in Europe with incoming GDPR regulation, needed to create a tracking plan including in-depth technical privacy audit.
• Strategy: Write tracking plan for full Segment tag management setup, providing code snippets and extensive documentation on use cases and technical aspects.
• Services: Tag Management
• Stack: Segment, Mixpanel, AppsFlyer, Adobe Marketing Cloud
• Results: Joined up Appsflyer marketing attribution to Segment tag management. Limited or obfuscated personal identifiable information in Mixpanel, and pushed sensitive data to Adobe Marketing Cloud which had better privacy controls.

Custom Attribution Model for a Fashion Affiliate Startup (Acquired)

• Problem: Last click attribution was showing limited ROI from Facebook ads, but when turned off, organic revenue would also decline.
• Strategy: Build our own multi-touch attribution system using browser cookies to store marketing sources, then injecting into hidden form fields.
• Services: Tag Management, Exploratory Data Analysis, Attribution Modelling
• Stack: Google Tag Manager, Mailchimp Forms
• Results: Discovered 40% of conversions had at least one visit from Facebook ads, revolutionizing the companies strategy on the path to acquisition.

Word of Mouth Measurement for a Popular Music E-Commerce Site

• Problem: Word of mouth was a significant driver of the service, particularly during the onset of Coronavirus, but there was no way to measure or forecast it.
• Strategy: Explore historical data from Google Analytics to determine what correlates with direct (word of mouth) traffic, and can be used to predict or identify anomalies in this metric to better plan for goals and optimize performance.
• Services: Exploratory data analysis
• Stack: GSheets, Google Analytics
• Results: Found a significant correlation (R2=0.97) between returning users and direct traffic, to be used in modeling. Our work on this methodology was published in Reforge.

Hiring Freeze Database for HR Startup

• Problem: There was limited daily traffic and a low domain authority for SEO (8/100), so the quality content they were publishing wasn’t getting much reach.
• Strategy: Build a realtime user-generated database of companies freezing hiring due to Coronavirus. Seed the initial traffic by sharing within groups on social media.
• Services: Engineering as marketing, Technical SEO, Exploratory data analysis
• Stack: Airtable, Intercom, D3.js
• Results: Drove over 1.5 million pageviews and 8,000+ user submissions, as well as a significant increase in brand-related searches. In addition, over 2,200 backlinks were generated which increased domain rank to 60/100 enabling ranking on many key terms.

Viral Link Sharing for Content Blog

• Problem: Over 50% of traffic was direct, meaning we didn’t know the marketing source. We knew through interviews that people were sharing the URLs of blog posts, but had no way to track this activity when it was done via ‘dark social’ channels.
• Strategy: Add a unique ID to the end of every URL for every user, and track when a new user arrives with an existing user’s tracking link, to properly attribute it to link sharing.
• Services: Tag Management, ETL Pipelines, Data Warehousing, Attribution modelling, Custom Reporting, Exploratory data analysis
• Stack: Google Tag Manager, Google Cloud Functions, Google Firestore, Google BigQuery, Google Data Studio‍
• Results: Identified that more than 20% of traffic to the blog was due to people sharing URLs via ‘dark social’ channels, and that some posts had 3-4x the virality of others.

---

If you enjoyed this post, check out the rest of the series:

• Intro to Growth Engineering
• Growth Engineering Services
• Growth Engineering Case Studies (this post)
• Becoming a Growth Engineer