Technology – Why we use machine-learning technology to increase PPC performance

While humans are great managing campaigns; machines can process huge amounts of data and make calculated changes in real time. We have the best humans and the best technology to manage PPC campaigns.

Our objective is always to get the most clicks and conversions for the minimum cost. While humans are very good at this, to maximise this, near constant optimisation work is needed to respond to times of day, competitor performance and many other factors – something which is impossible to do on a real-time basis for a human.

Humans are only capable of doing so much (We’re great a planning strategies, writing compelling ads and thinking logically for example) but, for some tasks, software can automate various aspects and react to data much quicker than a human ever could.

For this reason, DPOM has heavily invested in specialist software to allow us to make near real-time changes to your advertising account. Your campaign is still managed by a human, but our software continually uses algorithms to spot trends in your data and then adjust your bids and budget to ensure your campaigns perform brilliantly.

How does it work?

Traditionally, PPC managers may make adjustments to your account once a day as a maximum. At best, these adjustments are based on the previous period’s performance, and the adjustment is a reaction to this data with the intention of increasing performance. Our machine-learning algorithms make constant calculations resulting in near real-time bid and budget optimisation and as the name suggests: Constantly learns what works well and what doesn’t.

Example: Our software detects when your competitor’s ads are running our of budget on a busier than usual day. To lower your cost per acquisition, our software paces your ads slower than normal and then accelerates serving them when the competitor’s budgets have run out – resulting in a lower cost per click and CPA as there are fewer ads to compete against.

This is quite a basic example when we start considering other factors such as time of day, device, the day of the week and other factors, but it does provide a basic overview of the capabilities of our machine learning algorithms.

Example data:

technology

 

The above graph (actual client) demonstrates the number of conversions is increasing, whilst the cost per conversion is decreasing. The red box demonstrates the point our algorithm (together with the work from our PPC managers) dramatically improves the campaign to the point of achieving it’s most profitable level month on month. Had we not been taken on board, the campaign would likely not have seen these improvements.

Summary

Our software helps us provide a competitive advantage to our clients. It does not replace a human but given that pay per click platforms use complex algorithms to serve ads, it makes sense to use complex algorithms to get the best possible performance.

If you’d like to learn more about our technology, please contact us.