Derrick Kwa

BEAUTONOMY CASE STUDY In my latest role, I worked with Beautonomy, an e-commerce site that sold customisable beauty palettes. During that time, I managed most of the paid marketing channels, as well as the email automation channels and assisted with sourcing and managing partnerships with influencers and other publishers. FACEBOOK ADS We started with Facebook ads initially. The creatives were provided by other team members, I managed the targeting, reporting and calibration, together with my boss (the head of marketing). We tested multiple audiences (I can't remember the exact number, but probably about 10 different audiences. For example, one audience would be a list of beauty bloggers as interests, and another would be a list of beauty magazines. I used the audience narrowing to find people interested in multiple of these; the idea being that someone who follows 3 or 4 different beauty bloggers is more likely to be very into beauty, instead of someone who just follows 1. I'd also intersect these with an interest in online shopping, to focus on people likely to purchase online. In addition, we also tested broad interests groups such as the intersection of people interested in "beauty" and "shopping". As we got more visitors to the website, and more customers, I also built retargeting and lookalike audiences based on our visitors, newsletter subscribers, and customers. These Facebook ads were an ongoing calibration. Every week, I would look at which ad sets were performing better/worse. I'd then: - increase the budgets of the best performing ones decrease budgets of ones that weren't doing so well pause ad sets that were performing very badly add variations of high-performing ad sets (for example, if the beauty bloggers ad set was performing well, we'd create another ad set with a different group of beauty bloggers to test. After gathering data for a few weeks, I also ran a simple machine learning analysis on our Facebook ad data (tested both decision tree models and regression learning models) to determine what aspect of the ads were making the most impact (copy vs day of week vs device vs color of image, etc). This helped us narrow down which aspect to focus on (though I can't remember what the exact results were). As the creative team came up with more creative assets/visuals for ads, we would test these too, running all the new creatives for a week, and then turning on/off some ads based on performance. GOOGLE ADS As we went on, we also tested Google Ads. I worked with the creative team on the copy, and managed the targeting/bidding myself. For Google ads, we used a tiered bidding strategy, with enhanced bids. For each ad group (each ad group basically contained one main key phrase and some close variations on it), I would have a broad match, a modified broad match, phrase match, and exact match - with increasing bids in that order (because the exact matches are more valuable to us). We also ran ads towards remarketing audiences and similar audiences built on that remarketing list. We started with about 4 ad groups, and calibrated as we went. Each week, I would: - - look through the ad groups to see how they were performing add bid adjustments for devices, time/day, demographics based on performance go through the list of actual search queries, and add negative keywords to the groups for queries that were performing badly look through the search queries to find keywords that were performing well but that we weren't targeting specifically, and add them either in existing ad groups, or as a new ad group (depending on how closely related they were to the existing ad group). pause ad groups if they were performing very poorly brainstorm new keywords that were similar to the high-performing ones, and create new ad groups to test adjusted keyword bids as necessary To further enhance results, I also created personalisations of the homepage using VWO, where the headlines and CTAs were varied based on the search keyword/ad they had clicked on to arrive on the page; the personalised pages had more than double the conversion rate of the default home page. EMAIL FUNNELS I used Drip to build our email automation funnels. We had a few different automated sequences with different triggers, and the users were tagged in Drip based on actions they performed on the site. The emails were written and designed by the creative team; I set up the funnels, segmentation and tagging, and all the technical aspects of the automation. NEWSLETTER SIGN UPS 5 days after they sign up / create an account, we would send a promotional email with a coupon code, to encourage them to purchase WHEN THEY ABANDONED CHECKOUT For users who added items to the cart but didn't complete checkout, we would send an email reminder the day after, to check if they had any problems, and offer to help. We'd then send another email 5 days later (if they still hadn't purchased), to offer a promotional code. POST PURCHASE After purchase, we wanted to keep the customers engaged beyond just the standard order confirmation. A week after their order, we'd send an email asking what they thought. Customers would then be segmented based on their response to this - and get slightly different follow up emails. The ones who had a bad review, we would reach out to and try and see what was wrong and if we could help them. For ones who left a good review, we wanted to keep engaging them. So a week later, we'd send another email, asking them to share a photo of their new look (since we were selling makeup palettes) and to tag us on IG. A month later we would send another email asking them to refer us to a friend if they were liking their palette, with a link to a referral program. INFLUENCER ENGAGEMENT When I was at Beautonomy, we only started the early stages of engaging with influencers. My boss shortlisted a few influencer search platforms. I wrote the brief, shortlisted and selected influencers to reach out to (based on their reach, how much they would charge, etc) and worked with them as they made their posts. REPORTING Every day, we would generate an in-depth report. Using Supermetrics and Google Sheets, I had created an in depth dashboard, that pulled together data from Google Analytics, Facebook Ads, and Google Ads, as well as our own direct sales data from WooCommerce. I used various (and sometimes complex) Google Sheets formulas to combine these data into a digestible report. The report was customisable by date (there were start and end date cells that updated the rest of the document when changed). Based on this data, I would create a weekly list of actionable insights. This included ad performance and test I wanted to perform the next week, but also suggestions for the development team. For example, if I noticed a big drop in engagement on a certain page, I'd dive into the stats to try and figure out why. If it was a load time issue, or if I had other suggestions (copy, design, etc) I would suggest it to the team for them to update. END RESULTS Overall, through these efforts, we increased our conversion rate on the website over the 6 months, and took the company to >15,000GBP of monthly revenue.