Rule Based vs. Portfolio Bidding Models

Bid management is a discipline I find truly fascinating – and not because I find increasing or decreasing Max CPCs’ something to write home about. More so, it’s the ability to significantly improve the performance of your SEM campaigns through advanced statistical modeling – through the mapping of historical performance trends in your account that correlate spend to conversion. We’re all familiar with the brand names: Marin, Kenshoo, Adobe, IgnitionOne, [Insert Company X] … you get the point. But how many of us have solid, functional understandings of the science behind these automation tools? The technology is advanced, but the concepts behind bid management are actually quite simple. Below is a breakdown of rule based vs. portfolio bidding models in 600 characters.

Smaller automation tool sets typically come with “out of the box” solutions for bid management – insert rule based bidding. In rule based bidding, a system will evaluate ROI on a keyword by keyword basis – which sounds great, until you have more than 10 keywords in your account. In large SEM programs, rule based bidding fails to be effective because (typically) volume in the majority of campaigns tends to be skewed towards a select set of keywords. Without a sufficient amount of historical data, rule based systems will be unable to competently manage the majority of keywords in your account – usually resulting in long tail keywords receiving very low, static bids. Supplementing complications, rule based systems are also unable to make trade-offs between keywords – meaning, if you’re asking for an overall $15 CPA keywords in your account will never breach that ceiling, even if it may benefit your overall ROI.

Unless your KPI is something very generic (increase overall clicks,) I would steer clear of rule based bidding systems – Google can do this kind of thing for you for free, anyways.

Insert portfolio based bidding models. Portfolio modeling, as opposed to rule based systems, is simply more advanced  – portfolio systems utilize algorithms that compile a vast amount of historical data to model keyword performance. By modeling keyword performance, portfolio systems can gain unique insight into a keyword’s impact (think bid change) on the account’s performance as a whole. Portfolio modeling is based on three tiers: dynamic clustering, marginal-based optimization, and performance forecasting.

Dynamic Clustering -unlike rule based bidding, data scarcity is not an issue in portfolio bidding. Portfolio systems use a tactic called “data clustering” in which keywords with similar traits (campaign structure, destination URL, meta data) are grouped together. Data clustering allows portfolio systems to build models for keywords with little to no data (by bidding them similarly to their counterparts.) Dynamic clustering is a pivotal tactic used to maximize the performance of your account.

Dynamic ClusteringMarginal-Based Optimization – Portfolio systems perform thousands (if not millions) of calculations everyday to understand the opportunity cost of each additional dollar spent on a keyword. By focusing on marginal return, portfolio systems will bid on your account holistically to maximize performance. Remember our $15 CPA example from above? In portfolio bidding, keywords are bid above and below CPA thresholds in order to maximize return while maintaining your overall CPA requirements.

Performance Forecasting – all components of portfolio bidding culminate in the ability to forecast future performance. A main component of statistical modeling, portfolio systems utilize historical weighting  – meaning, each individual keyword is given a certain level of significance in the portfolio based upon the amount of historical data the system has on that keyword. Through historical weighting, portfolio models forecast future performance based upon a keyword’s historic data, events, and trends.

Performance Forecast

Understanding the basic concepts behind rule based and portfolio bidding models are first steps in being able to choose which tool sets are right for your campaigns. Apart from pure performance uplift, time savings, automation, performance insights, and efficiency gains are paramount – and of course, cost. The bid management software market is fierce and will continue to be so in 2014. My advice: research multiple tools, partake in multiple demos, and let the vendors compete against each other for your business.

Iestyn Mullins About the author
  • Eric D.

    Great article, Dustin! A concise summary outlining the pros and cons. Thanks!

    January 28, 2015 at 11:40 am

Leave a Comment: