When PPC is as important as it is for Argos, understanding how external factors impact on conversion and spend is vital.
Argos’ paid search agency Summit won a Masters of Marketing award in 2015 for doing just that, optimising bidding for the retailer in reaction to competitor TV ads, weather and seasonality.
Let’s look at the what made this a winning campaign…
Summit’s performance marketing platform, Forecaster, reviews PPC search data, Argos ROI targets and transactional data, as well as external purchase triggers.
It then uses predictive analytics to forecast daily revenue of marketing activity and adjust spend and creative accordingly for every possible bid representing Argos’ 50,000+ products.
This allows for better planning of PPC budget and increased margins.
External purchase triggers
The challenge was to model these external factors with sufficient accuracy to be able to make daily budget and optimisation changes in real-time.
An initial review of historical data across Argos’ entire product range aimed to discover seasonal patterns of activity around 9m keywords and product categories in Argos’ search accounts.
Forecaster mapped the annual seasonality of each of Argos’ products and grouped together any with a similar seasonal pattern.
This led to the creation of 72 different seasonal curves that represent the seasonality for the entire Argos product range, predicting changes in impressions, click through rate, cost-per-click and conversion rate.
A seasonal curve showing impressions for a number of Argos keywords with similar seasonal patterns. Source: Summit.
During this mapping of different seasonal patterns, Summit identified 13 different seasonal curves for the Christmas period, alone.
As impression share and clickthrough rate change was mapped throughout the year, campaigns could be optimised more accurately.
Each seasonal curve allows for the distribution of budget, changing of bids and amendment of ad copy.
Over time, these seasonal curves are updated to ensure they reflect an accurate picture of buyer behaviour.
Examples of four (of 13) distinctly different Christmas seasonal curves. Source: Summit.
To further develop this seasonal insight, Summit looked at the impact of specific weather changes on key product ranges, discovering how seemingly insignificant changes in temperature and precipitation can influence buying behaviour.
10 years of historical data were used to identify the items in Argos’ portfolio that are affected by changes in weather.
Mapping the effect of temperature difference (against the seasonal norm) on product sales allowed the identification of trends, e.g.:
- A 5°C change in temperature against the seasonal norm gives a 100% uplift in electric blanket sales conversion.
- Conversion rate for sledges doubles when it snows and the temp drops below -3°C. On snow days, search volume for sledges increases 300%.
This analysis led to the creation of a library of weather templates for every product, identifying the specific effects of temperature difference on each product.
Seasonal pattern and extremes (sales) related to changes in weather. Source: Summit.
These weather templates identify the seasonal and weather triggers that are then used to automate PPC campaign adjustments, proposing bid changes based on the season and weather.
Forecaster maps a 10-day weather forecast, broken down into 26 UK regions, to Argos’ 750 stores.
Using this information, geo-targeted bids can be tweaked for weather changes in specific areas.
Lastly, Summit analysed how relevant television ads can prompt an increase in online shopping activity.
People in the UK watch 455bn hours of TV a year, with ‘second screening’ having a major impact on online activity, which significantly increases in the first few minutes after a TV ad airs.
A key challenge for retailers is knowing when relevant ads will be shown (both Argos and competitors), and how to quickly make appropriate changes to online marketing such as PPC.
Summit mapped online uplift against ad length, time of day and program content.
This allowed Argos to understand the impact a TV ad campaign has on sales (see below).
Forecaster used this analysis to identify how to respond to campaign exposure and duration of the exposure, predicting the likely uplift from a TV ad.
This predictive model means that bidding can be managed in almost real-time in response to TV ads.
Summit triggers ads against over 2,000 TV ads per month across 38 major television channels, 80% of which are competitors’ ads. New and modified search ads are live before each ad has finished airing.
Here’s a generic illustration of how the Forecaster platform influences PPC activity, based on all three external factors.
For more on PPC and external factors, see: