Note: This post has moved from Leapthree.com to Ayima as part of the 2018 acquisition.
That was the question I was asked last week regarding the performance of an online retailer. There were three days of poor sales and my client wanted to know why. I realised while investigating this that there is a standard process I follow.
The first step is to confirm the drop in sales and which days were affected. This is done by comparing revenue for the suspect days against the same days in the previous week, giving you the % drop in sales. Not only do you identify which days were affected but the calculation provides a useful benchmark for the rest of the investigation.
Before looking at any data, I would ask a few questions internally. Were any changes made to the website or marketing efforts? Did the website go down at any stage? This information could immediately provide answers or context for your investigations.
A drop in sales MUST be the result of at least one of three changes in visitor behaviour. All are linked to a web analytics metric so the question of which were affected can easily be answered. Simply compare the days where sales dropped against benchmark days for these three questions with their related metrics:
- Did less people use the website – visits?
- Did less people place an order – conversion rate?
- Did people spend less money on each order – average order value?
At least one of these must display a drop in performance. It may be all three although life becomes a lot easier if it was only one. Whichever of these metrics has suffered a significant % drop dictates what should be investigated next.
If sales have declined due to less traffic, you now need to understand what caused the drop in traffic. These questions are designed to narrow down the list of potential culprits. For each, compare the same two sets of comparison days to see if the decline in visits is consistent across all subsets or isolated in a particular one. If consistent, move to the next question. If you have isolated a factor, then you are one step closer to the answer.
- Was there an issue with one of your marketing channels – traffic sources?
- Was it a particular group of visitors – visitors by country?
- Did the website go down – visits by hour?
- Was it a particular part of the website that was less appealing OR have you lost code on part of the website – landing pages (grouped by site section if possible)?
In investigating a drop in visits, do not rely totally on web analytics data and neglect external factors. Warm weather generally means more time spent outdoors and less time on the internet. School and public holidays both mean people are off on holiday, again resulting in less time spent on the internet.
There are various stages during the process a visitor to a retail website must follow before they can make a purchase. So if the conversion rate has dropped, see if there was one particular stage of the process which has declined:
- Visits => Get to Product
- Get to Product => Create Cart
- Create Cart => Commence Checkout
- Commence Checkout => Place Order
If it is not a single stage that has been the cause of the drop in sales, it is possible the website has not actually performed any worse. Instead the mix of visitors from different sources has changed with a higher proportion of lower quality visitors. This may be due to:
- Higher proportion of visitors from a lower converting traffic source
- Higher proportion of visitors from a lower converting country
If this is the case, it means that while total visit numbers are unchanged, there has been a drop in traffic from high quality sources. Return the investigation to the visits section but applying the questions only to the traffic segment you have identified as suffering a drop in numbers.
Average Order Value
The number of visits and orders may be “normal” but with less revenue generated through each order. The initial questions to answer are:
- Did customers buy fewer items each – items per order?
- Was the value of the items purchased less – average selling price?
The answer here gives the context for the next point of investigation, into how the mix of products that were purchased has changed. This can be reviewed at category level first, descending to subcategory, brand and product levels as required. Potential answers that are being looked for include:
- popular items being were on sale (likely to cause a drop in average selling price)
- a popular range running out of stock (likely to cause a drop in items per order)
- a change to the policy on the cost of shipping (likely to cause a drop in items per order)
The next set of questions
So the field of suspects has been narrowed down and you may even have isolated the factor that led to a drop in sales. For each factor, there is another set of questions that could be asked and another below that. This list starts getting more specific to the business given the nature of the business and the answers to previous questions. The web analyst needs to be creative in devising these questions as the decision tree we started with distorts into more of a chaos theory shape.
Identifying the factor that has changed, at a granular a level as possible, is only part of the job. This should help in understanding the cause of the drop in sales and whether it is a temporary or permanent change.
But key now is to decide what actions should be taken as a result of the intelligence you have gathered. It could be a change to the marketing strategy. Or a change in business strategy as you can now calculate the impact of poor availability on business profitability. As when doing any analysis, remember that the role of the web analyst is to transform data into intelligence, with this intelligence used to inform decisions that improve business performance.