Summary

• Improved basket sizes consistantly month-on-month from £67.88 to £80.34 over a 9 month period, exceeding £75 target.
• Product improvements contributed to the business improving margin by +8 percentage points in 9 months.
• Ran shopper surveys and interviews to determin priorities for the whole business to support increasing basket sizes.
• Ran a suite of A/B feature tests to determin drivers of basket sizes that did not negatively impact key conversion metrics.
• Introduced personalised product recommendations, live search, bulk-buying, cross-selling, and 'Before you go' features that added a combined 18% revenue per customer.

Responsibilities: Led Product, Engineering, and Marketing teams. Identified key areas for improvements that informed the growth strategy for the whole business. Devised a test plan to quickly introduce, evaluate, and interate on features. Hands-on design of prototypes through to finished features. Scoped and broke down work into 'bets' for engineering team to tackle. Measured and analysed results of split tests to determin which features to keep and which to drop.

Problem

Good Club launched in Q4 2019 on a mission to make sustainable staple products more accessible. After an intial launch, the business was tasked by investors with increasing margin and improving unit economics of an order to attract Series A investment. With high fixed delivery costs, a key driver of success was the businesses ability to concentrate customer orders into large, single monthly order of baskets above £75.

Initial research

The problem posed was primarily a business one - maximise revenue to increase margin - but at its core, there is a user need to tap into: help customers find more of the products they need as they shop.

I started by conducting shopper surveys, supported by 10 deeper interviews, to understand barriers and pain points that were stopping shoppers from buying more from us.

The results of this research formed the basis of our OKRs across the business throughout H1 2020, prioritising for solving:

• Adding more products to the Good Club range that shoppers wanted
• Ensure shoppers could find products easily and quickly
• Reduce out of stock levels to <5% (solved by moving warehouse in a separate project of work)
• Offer size variants with discounts to encourage additional purchases of existing SKUs

Solutions

'Good Club don't sell all the products I want' was the number one response from shoppers. Further digging into what exactly they wanted uncovered that the problem was in fact two fold: 

1. We didn't stock lots of the types of products shoppers wanted

2. We did stock the products, but shoppers couldn't find them

UGC ranging priorities

To help our Buying team prioritise which products to bring into the range, within a week we integrated a product voting system (using third party voting platform, Upvoty) into the site that allowed shoppers to suggest products and upvote those that they would like to see. This resulted in hundreds of suggested products, with thousands of votes and comments, giving our buyers a very clear steer on what products to introduce to the range to have the biggest impact on baskets.

Live search

Looking into site search analytics (see takeaways below), and watching recorded user sessions back, it was clear that shoppers were often struggling to find the products they were looking for, even when the products were in the catalogue.

🍕Key takeaways:

• Only 28% of all shoppers used search (compared to a 43% ecomm benchmark).
• When a shopper did search, they converted at 3x higher rate.
• Search was further underused on mobile with only 12.9% shoppers searching.
• 49.83% of searches require refinement, indicating that almost half of site searches were not successful.
• Common search terms (e.g. 'tea bags') returned zero results even though we stocked the product.
• Only certain product data was included in the search index (e.g. 'size' is not included) meaning more specific searches (e.g. '5 litre refill') returned zero results.

To rectify these issues, we integrated search product Algolia, which gave us a host of advanced search features such as live search, product recommendations, keyword synonyms, and detailed facited search functionality out of the box.

🏆 Feature results

• Search utilised by 39% of shoppers - 40% improvement
• Reduced 'no result' rate to 7.83%
• Reduced search refinements to 9.88% - 80% improvement

‍

Before You Go

An interstitial page after basket but before checkout to up sell and cross sell products before the shopper completes their basket is a classic grocery feature that basically every supermarket we reviewed employed. This proved solid validation for the value that we might also get from such a feature.

Timeboxed to two weeks development, we built and split tested the full feature with both new and returning shoppers over a one month period in order to gather enough data to see a statistically significant result.

🏆 Feature results

• New shoppers saw a +£2.93 uplift to AOV with a 92% statistical significant confidence
• Returning shoppers saw a +£3.85 uplift to AOV with 88% statistically significant confidence verses shoppers who were not exposed to the feature.
• Ideally we would have collected more data to reach a higher statistical confidence but as time was of the essence and we saw a strong early positive result without any drop-off in conversion, I made the call to launch the feature.
• After an initial launch, we iterated on the products served to provide an additional +£0.59 uplift to AOV across both new and returning shopper cohorts.

‍

Personalised product recommendations

New shopper baskets were consistantly smaller then returning shoppers, but their baskets grew from £58 on shop 1 to £74 by shop 4. The hypothesis was that if we could get more of the products that shoppers wanted in front of them in their first journey, we could increase basket sizes of new customer cohorts.

Personalisation seemed like an interesting avenue to explore to put more relevant products in front of shoppers. As we did not have any data for new users, I devised a simple onboarding flow that asked shoppers a few questions about their shopping; how many people they were shopping for, what values were important to them, and what product categories they needed at that time. The flow would then show a page of product carousels suited to their needs, along with an offer a free gift (already validated as the strongest aquisition reward) to encourage them to complete the flow, and some top-level metrics to illustrate the benefits of shopping with Good Club (saving plastic, CO2, and money).

Building a live data prototype

After running some initial user tests on clickable wireframes, it was clear that to really understand if the feature could deliver value, we'd need a live data prototype that would show shoppers real, tailored results.

With a one month timebox, the Shop Squad set about building a lightweight prototype to begin testing with a small cohort of new shoppers.

Scrapping the feature

After a month of development, it was clear that to get meaningful, personalised results would be much more difficult that first imagined. We would need more data-points and a much more intelligent recommendations algorithm.

While it felt like there was potentially value to be gained, getting to it would be costly in development resource and so I took the decision to stop development rather than investing in it further.