How to Diagnose eCommerce Website Sales Performance Changes: Our Complete Guide
Anyone who has worked on an eCommerce business or supported one knows that sales fluctuate from time to time. When things are great, it can be easy to mask what is happening in the business, but when things dip or have a prolonged dip, panic tends to set in.
Below are some ideas on where we’d suggest digging in to better understand your business, in good times and bad. The work in this article can be done in Google Analytics or a different web analytics solution.
Our guide assumes that you have working knowledge of analytics including creating segments and refining data.
The guide is written from the angle of diagnosing when sales are down, but everything in here can also be used to identify what is performing better than in the prior period/year. That will allow you to lean into the things that are working better so that you can replicate them to scale your business.
Table of Contents Show
Identify when changes occurred at a macro level
We always suggest looking at performance on a year-over-year basis (using day-of-week to day-of-week comparison - meaning comparing a Friday to a Friday, as opposed to the 17th to the 17th if they fall on different days of the week).
If you are looking at a seasonal period such as Valentine’s day or Christmas, it’s also valuable to look at the number of days out from the holiday.
It’s also a best practice to look at week-over-week performance and compare it against a trend going back at least a few weeks. The challenge with comparing to a prior period is that it may not account for seasonal trends.
Note that sample size is critical so that you make a sound decision and don’t overreact to an anomaly. If your sales fluctuate significantly due to the price range of your offerings or if your transaction volume is low, you’ll want to ensure the data isn’t swayed by a few orders.
Control for external variables
You can usually spot the period of decline using the views above, but sometimes the exact timing isn’t obvious.
Consider looking at segments for just Desktop Direct Load or Branded Paid Search in order to control for other variables. This is an easy way to get a performance benchmark.
Helpful hint: If your eCommerce site gets a lot of blog traffic, you can also create a Google Analytics segment that eliminates all of the blog traffic. This can give you a more normalized view of conversion and other performance metrics. You’ll only want to do this if your blog traffic tends to be very informational and doesn’t typically lead to orders directly.
Look at Sale vs Non-Sale Periods in Aggregate
With so many retailers using promotions to drive sales, it can also be valuable to look at total performance during all sale and non-sale periods in a given year.
It’s possible that revenue is up during parts of the month when there’s a sale but down at other times. If you’re looking at overall data a month at a time, it can be harder to spot the timing for a decline without this segmentation.
Sale promotions can mask a lot of underlying issues, and when you look only at non-sale periods, it might be easier to find the actual issues impacting the site.
Was there a promotion or sale in one of the periods?
Check that first - if there’s a mismatch between sale dates or the number of days on sale, that will influence performance. Choose a different time period for your analysis.
Is there a major event impacting media or purchase behaviors?
Global, national and local events can impact performance, so ensure that this isn’t the core reason for the performance shift.
For example, if there’s a hurricane in one part of the country, sales from that state may drop off. If there’s an election, discretionary spending may drop off briefly, or your ads may be lost in the shuffle on social media.
As you validate whether your entire business or just parts of it are impacted, keep events like these in mind.
2. Check for Major Website Changes
Before digging into marketing channels and merchandise changes, you will want to speak with your developers or product managers to see if a site change was made around the time the decline in performance began.
Although it’s a best practice to A/B test website changes, the reality is that many businesses roll changes out without testing.
Even if it’s not the new feature or functionality that is impacting sales, it’s possible that the new code broke something else. Some developers don’t do full regression testing at the time of deployments, so this can be a culprit.
It’s also possible that changes may happen that impact your tracking or Google tag manager implementation. As you evaluate data, this is another thing that will impact all channels and merchandising categories, as opposed to just certain ones.
3. Understanding the Metrics You’ll Explore
Once you’ve identified when the change started, you will begin digging into other reports to figure out if the dip is related to one of the following items at a macro level:
Traffic - Are fewer qualified people visiting your site?
Orders - Are you seeing fewer total transactions?
Conversion rate (CVR) - Is the ratio between orders and traffic changing?
Average order value (AOV) - Are people still buying but spending less money when they do?
Items per transaction (IPT) - Are people buying fewer products each time they buy?
Average unit retail (AUR) -When people purchase, are they spending less on each item?
What’s the best high-level ecommerce metric?
In much of the analysis below, we will suggest that you use a blended metric called Dollars per Site Session ($/session) because it accounts for both AOV and Conversion Rate in one.
To calculate this, simply take revenue and divide it by sessions. Alternatively, you can use $/user, which would be total revenue divided by users.
Since revenue is a factor of orders * AOV, and conversion is orders / traffic, this metric provides a view into conversion and AOV in one by comparing revenue to traffic.
Sometimes a particular website change may lead to higher conversion but be driving AOV down in the process, so $/session can be a useful KPI to use for benchmarking.
Here’s a simple example - When you drop the price of an item from $29.99 to $19.99, the conversion rate will go up, but AOV will go down. If your goal is to make more money, you need to understand the combination of those two factors to know if you are making more or less.
Other Metrics to consider
While my strong preference is to focus on financial performance metrics, it may be useful to peek at bounce rate, time spent on site, and pages per visit. These may indicate an issue with how the site is loading; however, they are often a result of marketing mix changes as described below.
For example, Display traffic tends to have short time spent, and blog traffic tends to have low pageviews per session. We’ll discuss not getting lost in conversion rate pitfall - these metrics can also occasionally lead to that same type of swirl.
You can also check cart abandonment to see if people are delaying their purchase until the next promotion. Many users have shifted to using the cart instead of a wishlist or favorites - they simply store their desired products there and come back later.
4. Begin Digging into Traffic / Marketing-Mix
it’s important to recognize that not all traffic is created equally. The key here is to identify if you have fewer of the right types of visitors coming to the site in the problematic period versus the higher-performing period.
In order to work with the data most efficiently, the best bet is to export the information into a spreadsheet. You can hide columns and pivot the data to get a clear view of period-over-period performance.
You’ll want to look at the marketing mix to see what channels are driving sales this year, along with their associated traffic, order, AOV, IPT, AUR, and $/session data. Note that most of those metrics are not standard in Google Analytics reports, so you will need to add calculations manually.
Your goal with this exercise is to help narrow down whether performance issues seem to be one of three things:
Impacting all channels equally
Impacting certain channels disproportionately
Impacting only parts of certain marketing channels
Once you’ve identified a channel that is down (or up), you will want to peek at source/medium to see if only certain sources are down (or up) or if all are impacted.
Marketing Spend Changes
If traffic is down in only one or two channels, you’ll want to find out if marketing budgets have changed. Since there are generally some diminishing returns in paid channels, you should expect conversion and $/session to increase if traffic is lower. The inverse is true if budgets and traffic have increased.
Cluster Data for Your Marketing Channels
In addition to looking at channels on their own, I usually like to group channels into high $/session, mid $/session, and low $/session at the bottom of the exported report. This will allow you to see how those groups are changing overall in terms of absolutes and as a percentage of total traffic.
For example:
Branded paid search, email marketing, and affiliates may fall into high $/session traffic
Non-brand paid search, Google Shopping, and SEO may fall into mid $/session
Paid and Organic Social may fall into low $/session
By adding this extra layer of data, you’ll be able to see the forest through the trees. While it’s important to narrow down the specific channel(s) that are driving the drop in qualified revenue, sometimes it can be helpful to see a few channels in aggregate to see if you’re growing low $/session traffic and shrinking in high $/session traffic.
This is especially important when you’re dealing with low transaction volume on a single channel or a few small channels.
Watch for Empty Calories
Sometimes a channel can lead to increased traffic, but that traffic may be the equivalent of empty calories.
A common example can be from SEO. If a lot of blog posts are attracting visitors that have a lower chance of converting, it’s going to look like you’re growing the channel, even if revenue is down. Those blog posts may be very helpful in other ways, but you want to dig deeper if you find a disproportionate increase in traffic to educational/blog content.
Higher traffic with lower $ per session or user
If you spot a channel that is driving higher traffic volumes but lower $/session than the prior period, you should refine or cluster the data to understand traffic by landing page or campaign type.
For example:
Is traffic to category and product pages from SEO down while traffic to the blog section is up?
Is traffic from triggered emails down while traffic from batch & blast marketing emails is up?
Is traffic from branded paid search down while traffic from non-brand or shopping is up?
Is traffic from discount welcome emails up while traffic from other types of emails are down?
Did traffic from a specific channel shift to drastically different pages y/y?
Review Device Data
If you still haven’t spotted the issue, consider segmenting channel data by device (mobile, desktop, tablet) at the channel level. For example, if you are driving similar traffic volumes from paid search but lower $/session, perhaps your mix of traffic has shifted more to mobile.
Since mobile still tends to convert lower than desktop, this may help explain the issue. A shift between devices could lead you to find that even though traffic is up, revenue is down due to the mix. We’ve seen performance vary between the two as much as 50% or more, but it’s very store dependent.
Usual Suspects
Offline marketing changes
If direct load or branded traffic (hint: also look at Google trends) has decreased, think about any change in your offline marketing or upper funnel marketing investment.
For example, are you sending fewer catalogs or print pieces than in prior years? Are you spending less on upper-funnel marketing because the ROI wasn’t apparent on a direct basis? Did you turn off your PR? If you are using this article to understand why traffic is up in a certain part of your business, the inverse may also be true.
Increases in referral traffic
If the referral channel is up, check the sources. More often than not, these are items that need to be excluded - payment providers (Paypal, affirm, etc) are a frequent culprit.
If the credit goes to referral source and the $/session looks really high, you might have something swooping in to take credit at the tail end of the transaction process.
Spending down as a result of attribution issues
If referral-driven cannibalization (or affiliate / email signup or SMS promo-code cannibalization) has been happening for a while, it’s possible that you have inadvertently pulled back marketing spend on a great channel because it seems like ROI has dropped.
Note that you should look at both last-click data and first-click. Some channels, like affiliates, tend to be naturally more cannibalistic, and if you only look at last click, you may make poor investment decisions. Use the Model Comparison Tool and Top Conversion Paths in Google Analytics to identify if specific channels are seeing a larger dropoff in last-click transactions and revenue.
It is common to find that channels such as Non-Brand paid search see a more significant dropoff / shift to discount and coupon affiliate sites. If this has limited your investment in Non-Brand paid search, it could be squeezing the top of your marketing funnel.
Other Cuts
If you are still struggling to find the part of the business that’s impacted, you may want to consider reviewing data by store vs non-store markets (if you have physical locations or significant wholesale distribution) or USA vs international traffic. (Find this data in the Geo reports)
Ultimately, you might have a better shot of finding the opportunities if you start with the channel data to find the changes in revenue and $/visitor.
This will just set you up to work with a smaller segment of data as you work your way through the other reports and dig in. Just make sure it’s not such little data that you can’t get a good handle on what’s happening due to sample size.
5. Begin Digging into Conversion Related Issues
Our preferred metric, $/session is a blend of AOV and conversion rate, so we’ve tackled the topics separately. Your goal in this section is to identify whether your business is being impacted as a whole or if only certain products, categories or brands/manufacturers are struggling. You will also begin to find answers as to what might be causing the issue. As you narrow down your findings, you’ll have an easier time finding the reason for your performance changes.
Conversion
First, it’s important not to get lost in the rabbit hole of “conversion is down” - the real change in performance is often related to something other than simply “conversion.”
If your revenue looks good across channels, and all you’re seeing is a change in conversion rate, there may not be a real problem.
There are generally two culprits when conversion is down at a high level
Increased upper funnel marketing - If you are driving more traffic from traditionally upper funnel channels like social media or display, your conversion rate will be lower.
Increased mobile traffic - Things have settled down a bit in terms of device mix (mobile vs desktop traffic ratios); however, as more traffic shifts to mobile, conversion rates may also be lower. Increased mobile traffic can also be related to the first bullet above.
Merchandising Mix
You’ll want to look at sales by product, category, and brand. Similar to the marketing channel report, you’ll want to export this report in order to add your calculated metrics ($/session, AOV, IPT, AUR). Much of your analysis will be done in the Product Performance report.
Look for changes in sales by category, then by product or brand. Consider digging into subcategories if your transaction volume and/or revenue is substantial.
Your goal here is to narrow down whether the change in performance is impacting:
All categories or brands equally
Certain categories or brands equally
Only certain products
All categories or certain categories
As you research, keep in mind changes that may have happened to navigation (did you remove some links?), pricing, inventory, and policy changes. A good thing to have handy as you do this work is the prior and current period email calendars - sometimes, the issue is that a particular category was promoted in one period and not the other.
Only certain brands or manufacturers impacted
If you’re a reseller, the brands you carry could be promoting themselves more heavily directly - outranking you in SEO and paid search and possibly selling on marketplaces. Many consumers now prefer to buy directly from brands/manufacturers.
Conversely, your sales could also be down on certain brands or manufacturers because they have reduced their marketing spend. Imagine what happens to sales at Target when Kitchenaid runs a big ad campaign. Now imagine what happens when they stop doing so. Performance is going to change through no fault of your own.
We address navigation changes several times in this article, but if a specific brand is down, you may want to check to see if the product was removed from your navigation.
Only certain products are impacted
If only certain products are impacted, the issue is probably pricing or inventory related. You may also want to check ratings and reviews on the product to see if a quality issue has sprung up. (It’s also worth chatting with your customer care team).
Other things to check include:
Was the product “demoted” on core category pages (was it in one of the top rows and now it’s lower on the page or on page)
Are you spending less money on this product in Google Shopping (perhaps because it’s not converting as well anymore if the pricing isn’t competitive)
Was this product featured in an email last year or other ads, and isn’t this year?
We address more about inventory later, but if certain core colors or sizes sold are sold out, that can be as bad as being out of stock completely. (For example - size 8 women’s shoes)
If you are a brand or manufacturer, remember to take into account sales on all of your sales channels. If you’ve recently expanded your Amazon practice, it’s possible that sales are shifting there. We work with one brand that sells through major beauty retailers - a huge percentage of their sales come from there when they increase advertising.
If Average Unit Retail (AUR) is down
Are people shifting to buying less expensive products or buying from different categories? When AUR is down, it’s usually pricing related, but could also be attributed to the marketing mix.
One thing to look at is if google paid search is driving traffic to a different set of categories than in previous periods. If you are featuring products in mass media, email or display, that could also swing AUR from week to week or month to month.
If Items Per Transaction (IPT) are down
Are people purchasing fewer items when they buy? This could be related to free shipping thresholds or promotion changes. Consider creating bundles if this is an issue.
You should also check or expand recommendation zones. Merchants often like to manually override recommendations that come from tools that are driven by AI, thinking that they can outsmart the logic. This typically fails - check recommendation technology for an influx of rules.
Primary Factors Impacting Merchandise-Driven Performance Changes
True conversion issues or changes are typically driven by merchandising mix issues related to one of two things:
Pricing
Inventory availability
Pricing - In Your Direct Control
Pricing-related issues can be influenced by several potential changes or issues:
Direct pricing changes (increased prices on the same item year over year)
Did you increase prices on a specific item or line of products?
Ex: Roses vs mixed bouquets, single bouquets vs subscriptions
Changes in free shipping thresholds
Changing your thresholds will impact performance, especially around the holidays. However, decreasing revenue and increasing profitability might be the right move.
Make sure you check these things:
Did the messaging in your free shipping banner change even if you didn’t change the actual threshold (check both desktop and mobile)
Return policy changes may also impact performance, even if they aren’t directly related to pricing. People may just be less comfortable trying your product if they aren’t sure making a return will be easy
Coupon code usage
Consider looking at data for sale periods vs non-sale, and look at the percentage of your orders using a coupon or discount. A “welcome” offer when users sign up for email or text can be a big influence here, and if you don’t look at first versus last click performance, you may suddenly think your paid channels are failing.
Pricing - More Difficult for You to Directly Control
Brands or other retailers pricing below you
Spot-check this by searching for the exact product on Google and looking at the Shopping ads
Grey market items
Wholesalers selling goods at a discount where they typically aren’t allowed to sell, generally a marketplace like Amazon
Resellers breaking MAP (minimized advertised price)
If you are a brand and have minimized advertise pricing regulations, do some searches on google and amazon and see if other retailers are breaking the MAP policy
Inventory Related Issues
Inventory issues can impact the merchandising mix, leading to AOV and conversion issues:
Items that are not carried this year - Did you have a product, brand, or line of products that sold a lot last year and it’s not carried this year?
Roll up data to the category, brand, or product type (e.g., fast-pitch softball bats, women’s down jackets) level if you have a broad catalog
Consolidate variants such as sizes and colors if your Google Analytics setup has each of these all show up as individual line items
Items sold out y/y - Did you sell out of a best-selling item?
Consider variants, not just the SKU (stock-keeping unit) as a whole
For example, are you out of stock in size 8 shoes, white towels, or black t-shirts
Average Order Value (AOV)
AOV is actually a combo of Average Unit Revenue (AUR) and Items per Transaction (IPT). Below we tackle each separately
Exclude big spenders
Note that if your business caters to big spenders or occasionally has B2B shoppers purchase large orders, you may want to check individual orders to see if there are any outliers in your denominator that could be swaying performance. You can spot large orders in the Sales Performance report or in your core eCommerce platform reporting.
Ensure your AOV dip isn’t due to a change in your marketing mix
Certain channels may have a higher AOV than others, especially if you feature a certain SKU through them (for example, on social ads). If your overall AOV is down, but channel-level AOVs look ok, you may not have a real issue on your hands.
Other Merchandising-Related Areas
Sort Order
If none of the items above seem to be the cause for your conversion drop and it’s not related to marketing mix changes, take a look at the way your products are sorted on important category pages.
Are you featuring too many new products at the top and demoting your best sellers? Are you featuring expensive products at the top of the page?
While we often work on laptops and desktops with big monitors, keep in mind that a lot of traffic is mobile, and people may only see 2-4 products at a time.
You can help narrow down potential culprits by looking at landing page $/session and conversion rates period over period.
As a related note, if you notice that you’re sending a lot less traffic to high-performing landing pages, drill down to see which channel is driving the change. Perhaps your marketers are sending traffic to a new page that simply isn’t getting as much traction.
Site search usage
Site search users often convert higher than those that don’t use it (typically 150-200%+ better than users who do not leverage site search). Check to see if fewer people as a percentage are using search, and if they are, if the $/session for site search users is holding the same ratio relative to non-site search users
If $/session isn’t holding steady, check the actual site search terms - are people searching for something you don’t sell or are sold out of?
If usage is dropping, did your layout change on desktop or mobile? If it didn’t, does the design of your site allow you to make site search more prominent?
Are there more null queries?
Where to Look if Everything is Impacted, Instead of Just a Segment of the Business
Look for things that can cause macro conversion issues
Add to cart experience changes
Value proposition changes (shipping thresholds, other details you’ve highlighted previously)
Navigation changes (linking to different categories, featuring different brands)
Load time across the website
Filter/sort logic changes across the site
Sort logic for site search results
Changes to default products when people land on the product page
Changes to how products show up in the cart
Are you still stuck?
Review data outside of Google Analytics - Dig into channel-specific data platforms like Google Search Console, Google Ads and Merchant Center, your Email Service Provider (ESP) platform, and Facebook to see if you can uncover other performance clues. These platforms provide specific information not available in Google Analytics
Broken links - As your SEO team to run a crawl to see if there are some broken links on your website. Merchants or marketers sometimes rename category or product pages which can inadvertently break links that are found on your site.
Split data by Demographics - Are fewer of your core customers visiting your site in exchange for newer, unproven ones?
Operating system, browsers & resolutions - Is it possible the site is breaking on certain devices as a result of a design change? (For example, tablet resolution tends to be neglected) Do you QA in Chrome but not in Safari?
Usability testing - Watch over a person’s shoulder and get their feedback as they interact with your website. Ensure you are running tests on various devices (desktop, mobile, tablet) and with different demographics.
Feedback / Voice of Customer tool - On-site feedback tools like Opinionlab (now Verint) can help uncover issues. I think of it as having an army of remote QA staff!
Analytics audit - Have someone audit your Google Analytics implementation
Net Promoter Score (NPS) - Net promoter score and associated feedback can help you understand if you have a bigger problem on your hands that isn’t seen in standard business analytics
We know this is a lot of information. Please reach out if we can help evaluate your website performance and develop a digital strategy that helps you scale your eCommerce business.