Table of contents
Machine Learning – the Earlier You Start, the More You Gain!
Machine learning and its advantage over manual campaign management is incomparable if the campaign entails a large amount of data. Advanced algorithms used on optimization platforms can calculate data in real time, as opposed to Google Ads specialists who would have to spend hours working on it. However, not every single campaign works well under the optimization platform which adopts machine learning. Some platforms require a large amount of data with considerable budgets due to the fact that they need to collect a suitable sample in order to get statistical significance. It means that the more data the platform can collect, the more accurately it will be able to estimate the best bid. The best bid is the best ratio between rate and conversions or clicks within a given budget.
Machine learning algorithms, as a part of artificial intelligence, always should aim to improve their results. If the system is reliable but such state of affairs is non-existent, this means that the sample of data we supply is too small. With regard to low monthly budget of a few hundred zlotys, the platform may overstate daily rates substantially in order to gain good positions for adverts and consequently use up a monthly budget.
Selection of Optimization Platform Based on Machine Learning
There are many optimization platforms on the market, including Google’s – DoubleClick. In Verseo we use our own platform – Verseo Campaign Manager – which lets you optimize the campaign with the lowest monthly budget. Most optimization platforms operate on algorithms which have been developed for a long time and which are not available to the public. Due to their power, companies treat them as an intellectual property. Leaving aside algorithms which VCM platform is based on, let’s have a look at 2 examples which show the impact of optimization on large collections of data.
Campaign With a Monthly Budget of About USD 6000
As a result of plugging the campaign in the optimization platform, the number of conversions in the same week increased substantially while conversion costs dropped. This means that the budget was used effectively and the platform precisely selected rates for the auction in which adverts participated.
graph 1: number of conversions (blue) vs average conversion cost (orange)
graph 2: average CPC (blue) vs clicks (orange)
The platform also led to growth of the number of clicks and reduction of average CPC rate. You can clearly notice the impact of the campaign which lets you get more clicks and conversions at a lower price, which is the priority to anyone who conducts AdWords campaigns.
Google Ads allows spending a budget in one day up to 200% of the daily budget. Monthly AdWords must not exceed the monthly budget. On days when the click potential is higher, the daily budget is exceeded. But it will never exceed 200% of the daily budget. If you exceed daily rates at the beginning of the month, displaying is limited at the end of the month. This leads to lower potential of the campaign and number of possible clicks at the end of the month. Therefore, Google Ads manages the daily budget under specific rates. When the campaign is small and the number of ads is low, AdWords will boost rates to reach the 1st position so as to spend as much as possible.
The platform uses or increases daily budget rate in order to maximize the number of daily clicks. The difference is, however, that VCM platform manages rates within the specific budget and allows increasing the number of clicks within the budget.
Campaign With Monthly Budget of USD 600
Another example is the customer’s account attached to the VCM optimization platform. In this case you can see clearly that the manual management of this campaign would let you manage rates more precisely. In this case the customer has a minor monthly budget.
graph 3: number of conversions (blue) vs average conversion cost (orange)
The graph demonstrates that attachment to the platform fails to bring the desired effect, that is the platform failed to raise the number of conversions and reduce the conversion cost. The results are not stable, which means the platform is looking for suitable advert rates all the time. It is necessary to remember that you have to wait a few days before the machine “learns” your data and selects suitable rates.
graph 4: average CPC (blue) vs clicks (orange)
Also the second graph shows that the platform failed to boost the number of clicks and lower the average CPC. In this case it is more advisable to optimize rates manually. Not only the above graphs show if the campaign works well under the platform. It is also necessary to analyze other indicators, such as average position of adverts, or CTR.
Machine Learning for Google Ads – Conclusions
Based on the examples above, the conclusion is that machine learning for small campaigns and small amounts of data is not always more effective than manual campaign management. It is considered that the VCM platform effectively optimizes rates with a minimum number of conversions a month; you have to remember, however, that: the more data, the better. You need to remember that there is not any universal way of getting ahead in the online campaign. Agencies also assure well-selected methods of campaign optimization, adequately to its settings and budget. It is therefore favorable to have an optimization platform, and using it does not entail any additional fees. This tool supports both agency and advertiser.
