Hearth & Home
- Rajwol Khadka
- 3 hours ago
- 8 min read
Where We're Losing Customers and What We Did About It
Background & Objective
Hearth & Home is an e-commerce retailer selling a small catalog of home goods including blankets, throws, and gift sets through a website that tracks every visitor's session, from landing to our home page all the way to a completed order. Over the three years captured in this dataset, the business grew substantially: website sessions climbed from ~62K in 2012 to over 233K in 2014, with orders and revenue concurrently growing.
In order to ensure growth continues to follow the upward trend we've been seeing, stakeholders have tasked me with:
Addressing certain pain points in the website experience that may be limiting order conversions.
Evaluating initiatives we've taken to understand their impact and use that process as a model for how future changes should be assessed.
Understanding whether ad spend is being allocated towards search channels that deliver the strongest return.
This analysis includes a full funnel analysis mapping every stage of the customer journey, an A/B test measuring the impact of a new billing page deployed by the engineering and UX team, and an analysis comparing the performance of paid, organic, and direct website traffic. By pairing this rigorous dataset with advanced SQL queries and Tableau reports, I aim to give stakeholders and board members key insights and recommendations to ensure millions around the world have a piece of Hearth & Home.
Executive Summary
The full analysis of the customer journey (first website visit to completed order) saw two significant leak points in the funnel:
Product Page → Cart (45% conversion)
Billing → Order (62% conversion)
Evaluation of the new billing page redesign saw a 17-percentage point lift in conversion (45% → 62%), sustained for the past two years and drove an estimated $514K in additional revenue.
A separate analysis comparing website traffic sources found that organic search and direct traffic convert at least as well as paid searches, at nearly a zero acquisition cost.
Together, these findings point toward two definitive next steps
A focused review of the site's Product Page
A closer monitoring of ad spend efficiency relative to free channels
Methodology & Data Validation
This analysis draws from six core tables that track the full customer journey on a variety of levels. This includes website sessions, page views, orders, order line items, refunds, and products spanning from March 2012 to March 2015.
Before diving into any analysis, it's critical to ensure that these relationships are validated to ensure downstream findings could be trusted:
Primary Keys & Summary Checks: This ensures each table's primary key is unique, in addition to giving a review of baseline stats (row counts, date ranges) for every table to give a reliable starting point. View SQL Queries
Referential Integrity: This confirms that every foreign key correctly resolves to a real and existing record (e.g., every order references a valid session, every refund references a valid order item). All 7 relationships checked cleanly, with no broken links found. View SQL Queries
Cross-Field Consistency & Value/Logic Checks: This confirms that related fields agree with each other (e.g., an order's stored total matches the sum of its line items) and that values fall within logical bounds (e.g., no negative prices, no refund exceeding the item's original price, no timestamp occurring out of sequence). View SQL Queries
Note: To ensure proper query optimization, several of the analyses below (funnel, billing A/B test, and product page abandonment) share a validated temporary table that chronologically tracks each site visitor's journey through the website at the session grain, built once and reused across each analysis. This process proves to be more efficient in comparison to deriving similar logic independently for each analysis. View SQL Query
Finding #1 - Funnel Analysis
Where are customers dropping off?
Stakeholders and board members are curious about where Hearth & Home is losing business in the purchasing process. With access to data on each page view a customer encounters, this analysis identifies at which stage of the customer journey do we see the biggest drop off, and what can we do to mitigate it moving forward.
I was able to create and validate a chronological roadmap that all customers work through:
Website → Catalog → Product Page → Cart → Shipping → Billing → Order Confirmation
Note: Each stage grouping was individually validated against the raw data. This includes confirming that no session ever viewed more than one product page, or saw both versions of the billing page, and that 100% of sessions progress through stages in strict order with no reversions to an earlier stage.
This visual is fairly general, but it provides stakeholders with a pulse-check of the business before drilling into specific pain points. Key insights:
Retention: After first landing on the website, just over half (55%) of visitors continue on to the next step in the journey (Reached Catalog).
End-To-End: We see that ~7% of all visitors complete the entire journey and place an order.
While this chart shows how many visitors make it to each stage overall, it doesn't highlight where the biggest losses actually occur. By comparing each transition to the page viewed immediately before it, rather than back to the very first visit, we're able to isolate the true pain points.
This view makes it clear exactly which pages are worth investigating further to help close the gap on that ~7% end-to-end conversion rate. Key insights:
Product Page Drop-Off: Only 45% of sessions that reach a product page continue on to Cart, indicating it as the weakest transition in the funnel.
Our flagship product, the Original Weighted Blanket, was seen to have a notably higher abandonment rate (57%) than the rest of the catalog (35%-54%). Given its great share of total product traffic, even a modest improvement here could have an outsized impact, making it a strong candidate for a focused page review.
Order Drop-Off: While Catalog and Cart see larger drop-offs numerically, this stage is arguably the most concerning: a customer reaching Billing has already shown strong intent to purchase, unlike someone who exits after a single homepage visit. Only 62% of customers who reach Billing go on to complete their order, indicating a meaningful loss at the highest-intent step of the funnel.
Note:This figure reflects performance across the full three-year dataset. The next section examines this stage more closely and finds that performance here has already shifted meaningfully over time.
Finding #2 - Billing Page A/B Test
Did the billing page redesign actually work?
Following the launch of the Hearth & Home website, internal members received complaints centered around our billing page. Finding #1 pointed to Billing → Order as a weak transition worth addressing. After meeting with members on the engineering and UX team, we settled on a redesign of the original billing page:
/billing - live from March 2012 to January 2013
/billing-2 - live from September 2012 onward, fully replacing the original by early 2013.
Rather than immediately replacing the original page, we decided to run both simultaneously for roughly four months (September 10, 2012 - January 5th, 2013) in order to confidently measure their performance under identical conditions (same website, same traffic mix, and same time period). We additionally confirmed that no single session ever saw both versions, ensuring a clean and non-overlapping comparison.
Results:
Billing Page Version | Total Sessions | Conversions (Orders) | Conversion % |
Original | 1,663 | 750 | 45.10% |
New | 1,657 | 1,029 | 62.10% |
The redesign page converted 17 percentage points higher than the original, even with both groups drawing nearly identical traffic volume, ruling out sample size as an explanation.
The chart above confirms this wasn't a temporary snapshot, as both pages held consistent conversion rates across their entire respective lifespans and not just within the tested window. The original page's ~45% and the redesigned page's ~62% each held fairly constant throughout.
Business Impact: By applying the original page's ~45% baseline rate to post-rollout traffic, I estimate the redesign resulted in approximately 8,565 incremental orders accounting for roughly $514,000 in additional revenue over the 27-month period.
Finding #3 - Channel Performance
In addition to fixing specific leaks in the funnel, stakeholders also want to know whether ad spend is being properly allocated. For this analysis, I grouped sessions into four channels based on how each visitor arrived at the site:
Paid Search: Ads shown on Google and Bing search results, (paid per click)
Organic Search: Website listings that appear below paid ads when a user searches (unpaid)
Direct: User navigates to HearthandHome.com directly, with no search engine involved (unpaid)
Channel | Sessions | Orders | Conversion Rate |
Google (Paid) | 316,035 | 21,333 | 6.75% |
Bing (Paid) | 62,823 | 4,519 | 7.19% |
Organic Search | 43,411 | 3,262 | 7.51% |
Direct | 39,917 | 2,856 | 7.15% |
Note: average gross profit per order is nearly identical across all four channels (~$37–38), ruling out product mix as a confounding factor.
Paid search drives the majority of total traffic and orders, but interestingly does not convert visitors more efficiently than organic or direct traffic. Notably, all four channels grew substantially and improved in step with one another over the three-year period, suggesting site-wide gains lifted every channel roughly equally rather than any one channel outperforming on its own.
Since organic and direct visitors arrive at essentially zero marginal acquisition cost (no ad spend is paid per click), orders from these channels are likely the most profitable on a per-order basis. Every order made through a paid channel carries an acquisition cost that organic and direct traffic simply don't.
Recommendation: I recommend not reducing paid search investment based on these findings alone, as paid searches still drive the majority of total order volume. Furthermore, I recommend two follow-up actions to further flesh out this analysis
We pair this analysis with actual cost-per-click and margin data in order to quantify true channel profitability.
We investigate what's driving organic and direct traffic (brand recognition, customer loyalty, referral sources) as a potentially low-cost avenue for future growth.
Conclusion & Recommendations
This analysis set out to answer a valuable question for stakeholders and board members: where is Hearth & Home losing customers, and what can be done about it? Through various SQL queries and Tableau reports, I identified opportunities to enhance the customer experience and drive the ultimate goal of Hearth & Home becoming an established home goods brand:
Fix check-out stage friction. The funnel identified both Billing → Order and Product Page → Cart as the two weakest points in the customer journey. The billing page redesign has already been validated as the driver of a 17-point conversion lift for over two years. This test also demonstrates a repeatable template for evaluating future site changes by identifying a leak, establishing a natural comparison, and validating it rather than simply relying on before/after assumptions alone. I recommend a similar investment and methodology for future page improvements.
Give the flagship product's page a focused review. My analysis saw The Original Weighted Blanket abandons at a meaningfully higher rate (57%) than the rest of the catalog (35%-54%), indicating even a modest improvement here would have a great impact on its conversion rate. I recommend a content and UX review of this page, following a test modeled similarly to the billing page redesign.
Further investigation (and data) needed to further evaluate ad spend allocation. While paid search has and continues to drive the majority of website traffic, organic and direct channels convert to orders just as well with nearly zero acquisition cost. While Hearth & Home continues edge competitors in its market, we don't recommend reallocating ad spend based on this analysis alone.
Comments