Automated bid optimization algorithms can have tremendous impact on the ROI of PPC campaigns, but require a substantial amount of performance data in order to make the right bid changes at a keyword level. Locally targeted campaigns are at a disadvantage as budgets are often too small to automatically react to changes in search behavior in an agreeable time. Can nature help us respond to trends quicker and increase the ROI of local campaigns?
The Scarce Data Problem
The major problem for effectively optimizing local search PPC campaigns for leads and conversions is the lack of available data points. A typical local SMB search campaign may spend $500 per month, using 50 to 100 keywords across 5 to 10 ad groups. The campaign then receives about 100 to 200 clicks and – depending on the lead type that is being measured, normally generates something around 3 to 6 leads per month.
Example of local ppc campaign data object count
Optimization algorithms start with initial values for biddable objects and then modify those over time via rules or heuristic learning based on data collected up to the time of decision making. The data will tell the algorithm how a given biddable object is performing relative to other biddable objects of the campaign. Performance is often measured by the cost-per-lead, which can be calculated by dividing the price of the click by the percentage of clicks that result in a lead (Lead-thru-rate). The lead-thru-rate is initially unknown and differs between biddable objects.
Due to the fact that algorithms cannot understand the semantic meaning of biddable objects they must rely on computations or rules that instruct the algorithm when to draw a conclusion from the collected data and hence make a change to the bid. Market wisdom tells us that we have to wait for about 30 leads (conversions) before a campaign can be effectively optimized. That is, of course, a very rough guidance. The actual waiting time depends on factors such as the number of ad groups and keywords used in the campaign and the degree of confidence that the decision maker wants to have before a decision is made to push (increase bids) or not push (decrease bids) for an adgroup or keyword.
90% Confidence Intervals for two biddable objects based on a binomial distribution assumption for lead-thru-rates
The confidence level of the calculated lead-thru-rate of the biddable objects increases with the number of realized clicks and the 90% confidence interval boundaries converge towards the “true” lead-thru-rate of the biddable object. In the example above, the algorithm would have to wait to reach 200 clicks for each of the two biddable objects to be able to make a statistically sound decision that the red object has a higher lead-thru-rate than the blue object. Given the low number of clicks in local marketing campaigns the resulting waiting times are often longer than the lifetime of the SMB’s campaign.
In nature some animals face a similar challenge. Think of bees that fly out to gather nectar. Each bee must decide in which direction to fly to find a field in blossom. If each bee had to find that direction by itself the queen and brood of the hive would most likely starve. However bees communicate through dances to share information about the best foraging ground. Together they achieve amazing cooperative results! Ants use a similar approach, when they have found food they exchange information by laying down pheromones as a trace on the way back to the nest. An ant leaving the nest uses the pheromones from other ants and is thereby more efficient in finding food. Each individual bee or ant is too limited in its abilities to act or to gather information on its own. Together they are very strong performers. They share important assets such as knowledge, memory, creativity, strengths and sensors. This makes them fitter for survival, increasing their abilities in foraging, predator evasion, reproduction, motivation, energy savings and social motivation.
Waggle dancing honey bees
This collaboration among the group members is referred to as collective, group or swarm intelligence. In human society collective intelligence can be understood as the enhanced capacity that is created when people work together, often with the help of technology, to mobilize a wider range of information, ideas and insights. Collective intelligence emerges when these contributions are combined to become more than the sum of their parts for purposes such as learning, innovation, and decision-making. Google, Wikipedia or other results from the open source movement have demonstrated the potential of digital tools and distributed networks of individuals to generate forms of collective intelligence.
Collective Intelligence Brings Big Data Insights to Local Marketing
Collective intelligence can be used for local marketing too. We can aggregate information from different client’s campaigns provided that these campaigns exhibit a certain level of similarity. The use of marketing templates can ensure that campaigns from similar businesses are standardized and comparable. If, for example, every campaign built for a new dentist client is created from scratch then all campaigns will differ enough that you cannot compare apples to apples between two dentist campaigns. If, on the other hand, the campaign manager uses pre-built templates for the new client, which he then adapts to the specific situation of the dentist, then the overall structure and content of all dentists’ campaigns will be fairly similar. The more elements ranging from keywords and ad-text to the used landing pages that are similar, the better for collective intelligence. The shared information from each individual client is more meaningful for the group to be used to take action on all the group’s campaigns.
Collective Intelligence in Action
The results can be astonishing: We have used a templated campaign across 500 locations from one vertical and observed the bidding behavior. Over a seven month period we have seen 200 bid changes to improve lead performance for ad group level bids that were made with a confidence level of 90% based only on the information gathered for the individual location. As expected, a truly small number. However, Adplorer’s Collective Intelligence Algorithms that additionally combined the information of similar ad groups across the 500 locations were able to make 200,000 bid changes with a confidence level of 90%. A 1,000 fold increase in optimization effectiveness!
The benefits are manifold for agencies utilizing this technology:
- When using templates, campaigns can be easily pre-prepared and then only need to be slightly adjusted for the specific client. This is a well known advantage and is often capitalized on.
- With a Collective Intelligence algorithm, systematic bid and budget optimizations can be made much earlier. The client will thus see effects a lot faster than without such algorithms.
- The overall results for the clients will be better as fewer unproductive clicks are being bought in the learning phase of the algorithm. Learning does not only occur at the beginning of a campaign but also when there are changes in search behavior, taste and seasonality adjustments.
- New clients benefit from existing clients’ data. For example a new plumber that joins an agency which already has many plumbers’ campaigns under management will immediately benefit from the settings of all other campaigns.
- This will lead to network effects for agencies. Agencies that build such taxonomies and use collective intelligence will have a strategic advantage. The more clients from the same industry they have, the better their performance gets. Hence, new clients from this industry are better off joining that agency than another agency without this knowledge.
Since ancient times scholars have been fascinated by the swarm intelligence of animals, sometimes even attributing their actions to the will of gods. Today scholars and practitioners have applied the basic principles of collective intelligence to modern problem solving in many areas. The time has come for local marketing to also tap into the power of the many. Join the crowd!