T-Mobile : Bespoke Bid-Management
November 10th, 2008 - Posted in UncategorizedClient: T-Mobile
Project: Bespoke Bid-Management for T-Mobile
Channel: Search
Challenge
To grow market share in an increasingly competitive sector by 30%.
The mobile sector has been declining in traffic size since the start of 2008 so our challenge was to buck this trend and deliver our best ever performance.
Strategy
Our strategy was to unlock the potential of high cost non-brand terms by fully understanding their value on T-Mobile’s campaigns earlier in the purchase cycle than just tracking the last click.
We worked with all the data at our disposal to create a modelling and forecasting tool which would then be able to match with daily, weekly and monthly consumer trends.
Solution
By combining the techniques of click-path analysis and time of day targeting we used search marketing to deliver results beyond expectations.
Standard tracking solutions where keywords are optimised manually or through a
bid-management tool only attribute sales to the final click. More sophisticated levels of tracking are also available which assign a weighting to each click in the process based on its position in the click-path.
We felt that neither of these methods of tracking provided the full picture of conversion so BLM Quantum used click-path analysis to attribute a value to each click based on the time-lag between those clicks. This dictated how important that click was in the path to the final sale. An example of this is that a click counted on a generic term 1 hour before the final click made was assigned a higher value than if the click had occurred a week ago.
Having discovered the true value of all keywords in our account we then applied time of day analysis to discover which times of the day were providing the greatest amount of value.
We could then pinpoint specific time slots during the week and analyse them individually. We noticed instances where an ad group may have delivered 1000 clicks at high cost with minimal sales. Further analysis revealed that keywords in this ad group appeared often in the click path that leads to other sales. This insight helped to maintain our sales volume through other keywords by ensuring that we made decisions based on the most detailed tracking possible.
All of the techniques above provided us with three overlapping conversion trends; hour of day, day of week and day of month. We then included results from other online and offline channels. This eliminated any impact that they may have had on search campaigns and ensuring that our model was as accurate as possible.
We produced a forecast by day for each month to predict our performance accurately and then assigned our budget and bids accordingly. A base line was set for conversion rate which would outline when we expected the campaign to perform effectively. For example, our forecasts showed we could expect the conversion rate to be below the baseline on a Friday in the middle of the month. In this case we reduced our CPC bidding in order to maintain efficiency. Conversely we found that Wednesdays at the end of the month were generally above the baseline so CPCs could be increased.
When adjusting bids based on our forecasting we performed optimal positional analysis taking into account clicks, sales, conversion rate and CPC for keywords in our keyword groups. Lowering keyword positions both reduced cost and had a positive effect on the conversion rate while delivering similar sales volume
Performance
Since implementing these techniques we have seen a dramatic improvement in order volume and efficiency.
Comparing campaign results from the same periods in 2007 and 2008:
Spend reduced by 12%
Pay Monthly Orders Increased by 42%
CPA Reduced by 38%
This was by far the best ever H1 performance to date for the search channel.
Campaign Creative
(click thumbnail for creative example)
