Integrating client data for advanced analysis. Stock levels, search data and revenue. 

The Problem

Our PPC teams work with fast moving retail clients and need to be able to react to both changes in customer demand and rapidly changing stock levels. At the same time, our ability to spot opportunities within search and then communicate them to client remains crucial. 


The Solution 


The solution has been to import additional client data to our internal Datasets system. This allows our teams to factor this new data into their analysis and move beyond simple cost and revenue calculations. They are now able to analyse the impact of stock levels at different points, the product categories available, and even the sizes and colours.


A by-product of this solution is that we can take the same approach and use it to analyse historic performance in a new way, that better informs future strategy both in terms of paid search and merchandising.


Datasets gives our analysts access to our databases where we collect and store all of our data relating to their clients. They are able to join data from multiple sources using Hive Query Language, then apply custom scheduling to automatically run analysis and alerts.


In these examples our analysts had to find the common link between revenue data, search engine data and product information. Our account structure made this simple as the campaigns were split by product line, meaning that they could link campaigns to specific products and tie revenue to campaigns as they normally would for optimisation.


A snippet of the query to produce this analysis

A snippet of the query to produce this analysis


Taking into account stock thresholds, demand and ROI, we can make intelligent decisions as to how to best spend a client’s budget, and continually monitor for potential trends.




The first run of analysis was to use Datasets to join revenue, stock levels and search engine data and then split by product area (mirroring our account structure). We quickly realised that the first areas to address were non-linear relationships as shown in the following basic examples:

We may find for example, that there is high search volume (demand) but a low conversion rate, which immediately suggests that there is an issue with the product or page itself. Is the pricing competitive? Are we missing certain common sizes? Are competitors offering something more suitable? The actions may also be to test adjustments on the client’s site such as pricing or shipping options. If the price of a product is resultantly dropped, we can then continue to monitor the conversion rate at particular prices to discover what the optimal price for that particular product appears to be, based on customer actions.


Once we were able to prove the value in performing this analysis and that we were getting valuable actions, we looked at more complex examples using more granular data. Additional metrics such as new vs. returning customer data, as well as assisted conversions and revenue allow us to make even more informed decisions on how to better target new customers. We can gauge what the full impact of these campaigns would be on other channels by looking beyond simple last-click attribution.


There is also the related ability to monitor impressions and clicks of different product groups side by side, which gives some insight into customer demand for individual products or ranges. This can be fed back to the client to assist them with their future buying strategy.


We have one tool that provides us with data on:

  • Size and Scale of Product Range  
  • Stock Levels
  • Customer Demand
  • Advertising Spend
  • Revenue Generated
  • Direct and Assisted conversions
  • Cost per New Customer
  • Overall ROI



With this level of automated insight immediately available, it has freed up a huge amount of our analysts time. This is time that can now be spent optimising the campaigns and improving the numbers off the back of this data insight, rather than simply measuring the metrics required to gauge success.


Taking into account stock thresholds, demand and ROI, we can make intelligent decisions as to how to best spend a client’s budget and continually monitor for potential trends. Because we store this data on a daily basis, over the course of our relationship with our clients we build up significant amounts of historic data and can perform this analysis for any time period. We could use this to analyse the changes in demand and performance that we saw over the previous year for example, and  then use this insight to inform the buying and marketing strategy for the coming year.


For clients, this analysis has turned PPC from simply an advertising channel into a highly quantifiable, measurable and scalable look into the online behaviour of their customers. Looking at these reports, buying teams are also able to base wider decisions on the insights gained from PPC.


The Next Steps


The above examples were all born through analysts performing custom analyses for their clients. Once certain patterns and repeatable actions are identified, these can be scheduled as regular pieces of optimisation or reporting. This automation frees up the analysts to start looking for the next opportunity for the client, meaning we are constantly improving performance and innovating for our clients.


The examples contained here are also all retail specific. A strength of our Datasets system and Stage as a whole is the ability to perform this same analysis using completely different data, as long as there is that common thread between the data sources. For example a hotel client could replace the product stock levels for room occupancy, an airline could use ticket availability, etc.