(Or the optimization bonus, or the budget multiplier…these blog post titles require a lot of thought…)

Imagine this scene: the CMO goes to see the CFO. “This economy is tough…I need more budget to hit my KPI target.”

The conversation does not go well.  Not only does the CFO not want to approve more budget, they want to CUT budget, to “protect the bottom line”.

If this, or something like this, has happened to you (or a friend) then chances are you are not using a technology that I would like to talk about today…predictive optimization.

Predictive optimization combines two analytical tools; predictive modeling and customized optimization algorithms.

The idea is to capture the dynamics of your business with an accurate predictive model, and then use that model to drive simulations in which we vary spend allocations (by channel, time and market, typically, although other factors can be simulated as well).  One of those allocations is identified by the algorithm as optimal; that is, the highest KPI effect for the money spent over the time period being planned. (we will come back to that point, below).

The key to making it work: model accuracy.

Aim to have your model explain a very high percentage of variation in past behaviour.  Most of our implementations use models that are over 90% accurate when we compare predicted to actual on recent time periods (yes, including the post-COVID period; see our last blog post here)

Integrated marcom campaigns behave in complex ways; we want our models to capture these effects.  But the price we pay for that accuracy is to have a complex model.  Think about the dynamics it must account for:

• Diminishing return effects overall and by channel
  • One of the few “laws” of marcom planning is the diminishing return effect.  As you increase spend, there comes a point where each additional dollar has less and less an effect.  In our experience, this applies not just to the overall spend in the campaign, but to individual channels as well.  A planner needs to know where each channel is on its curve for the spend proposed.
• Channel multipliers
  • We ideally want the channels we pick to reinforce each other’s impact in the market.  Generally, the more channels we have, the better the effect, but to a limit.  What we absolutely must avoid is cannibalization; the effect of spend in one channel diminishing the contribution of another channel.
• Time
  • Marcom has an effect over a period of time, as we all know.  What is less well known is how that period varies by channel and by creative within channel.  Brand ads tend to have an impact over a longer period of time; promotion/call to action shorter.  Most digital forms have very short effect periods while other channels impact over a longer period
• Local geographic effects
  • Retailers and other location based marketers must take into account the impact distance to store has
• Creative
  • What is the right balance of brand vs promotion?  Budget allocation across different creative executions?

With all this (and more) going on, I just don’t see how planners can put together a campaign that takes full advantage of these effects without tools like predictive optimization. 

Further, to give guidance to decision makers, it is important to be able to produce credible, accurate forecasts of KPIs that will result from a given plan.  That is a big advantage of using a technology like predictive optimization.  It is one of the aspects of the technology that is most important in bringing the CFO onside.  Once they see the accuracy of the model, it is easier to trust it to help with decisions like these.

In our experience, this approach can lift KPIs by double digit amounts, while keeping budget constant.  Or, it can offset some or all of the effect of a budget cut. Or, it can make a persuasive case for a budget increase.

More impact on the same budget?  That’s the free money effect.  We think predictive optimization has a key role to play in helping brands recover in this tough COVID economy.