The 2016 holiday season is upon us, and many clients are planning to run promotions for Black Friday, Cyber Monday, etc. In anticipation of these promotions, clients often ask for accurate conversion forecasts. However, it is difficult to determine the boost in conversions due to promotional offers, especially if your client has never run promotions on the products in question. Thus, relying solely on historical performance will result in an insufficient forecast as one lacks data for how the promotion prices affect conversions
In order to overcome this lack of price change data, we need to revisit your economics 101 class. In short, you can use the price elasticity of demand (PED) equation to estimate the promotion’s effect on conversions. In economics, PED refers to the slope of the demand curve.
The law of demand states that as price decreases, quantity demanded will increase (and vice versa). Thus if we know the PED, we can calculate the quantity demanded given a new promotional price. However, what if your client has never changed the price of their products? Without two points of data, we cannot calculate PED.
In order to overcome this, you can use another economic concept, substitute goods, to serve as a base assumption. Substitute goods refers to two goods that can be used for the same purpose and if the price of one good goes up, the demand for the substitute is likely to rise (this is known as a positive cross elasticity of demand). Think of different types of speakers, baseball bats, cooking pans, or hiking shoes. Almost all consumer goods have a substitute, and the closer the relationship the better for this forecast methodology. As long as the good and its substitute are priced differently, we can use it to calculate PED.
*DISCLAIMER* The following methodology will produce a forecast estimate; do not expect 98% accuracy. Search engine marketing is organic; advertisers cannot control what consumers search for. Although the math is based on sound assumptions, there are many other factors that are not accounted for. Enjoy!
PED = Change in Quantity / Change in Price = ((Q₂-Q₁)/Q₁)/((P₂-P₁)/P₁
Q₂ = new conversion quantity
Q₁ = original conversion quantity
P₂ = new price
P₁ = original price
Let us assume our client sells kitchen supplies and is offering a “$30 off” promo for three types of knife sets from Black Friday to December 31st. Using historical data from the same time period the previous year (the below is fictitious data for the purpose of this blog post), we can compare each knife set’s conversion volume from paid search and calculate their mutual PED.
The PED results are quite interesting because they fall into three categories: elastic (E<-1), inelastic (1<E<0), and positive (for an explanation of the differences, here is the Wikipedia article). Because positive price elasticities break the law of demand, let’s discount this calculation and average the two knife set 1 PEDs, which equaled -0.75505
Having solved for PED, we can then use the PED equation to solve for the estimated promo transaction volume of each product.
Q₂=(Q₁((PED*P₂) – (PED*P₁) + P₁))/ P₁
We can now use the conversion lift % numbers to make our forecast given the December promotional budget. There are a number of ways to use this for your forecast, each with its own pros and cons. For the sake of simplicity, we will calculate the forecasted conversion volume for each knife set using historical data and proportional calculations relative to the current year’s budget.
Previous Year Data
- Total Transactions (all products) = 7054
- Total PPC Cost = $470,542
Current Year Forecast
- Approved PPC Budget = $640,000
- Estimated Transactions Relative to 2016 Budget = $640,000*(7054/$470,542) = 9594
Transactions without promo = Total 2016 transactions * % of all trans 2015
Transactions with promo = 2016 transactions w/o promo * (1 + Promo lift %)
Note that this method assumes all other things being equal and there may be other factors you may want to control for in your calculations (see Appendix below). Regardless, the promo lift % we calculated earlier can be used universally across different forecasting methods. Although an approximation, this forecasting method is based on sound logic and minimal assumptions. This methodology can serve as a model for teams facing similar forecasting challenges, allowing them to refine holiday execution strategies and tactics. From SEM to demand economics, reaching across fields ultimately displays 3Q’s innovative approach to providing business solutions for our clients.
Appendix: Other Considerations
Here are three factors you may want to consider:
1. Changes in the competitive landscape
New and old competitors come and go, and this most directly affect CPCs. Try comparing the auction insights report YoY to get greater insight into your competitive landscape. If your CPCs are drastically different from the previous year’s because of new/different competitors, you may want to use a different forecasting method that works backwards from CPCs and CVR.
- Calculate the previous year non holiday to holiday change in CPC
- ∆CPC = (CPC₂ – CPC₁)/ CPC₁
- Where CPC₁ is average pre-holiday season and CPC₂ is average holiday season
- Estimate the current holiday season CPC
- Use the current average CPC non holiday season (45 days should be good) = CPC₃
- Estimated current holiday season CPC₄ = CPC₃ (1 +∆CPC)
- Use CPC₄ and the budget for the period/campaigns in question to calculate click volume (assuming you will spend all of the budget J )
- Clicks = Budget * CPC₄
- Calculate conversion volume without promo using previous holiday’s CVR
- Conversions₁ = Clicks *CVR
- Calculate conversion volume with promo
- Conversionsᴾ = Conversions₁ (1 + Promo lift %)
2. Changes in account performance YoY
For better or worse, a lot can change in a year. Changes in ad copy, campaign/ad copy settings, and keyword sets can have a huge impact on performance. If this is the case with your campaigns, you may not want to use historical data to calculate your forecast. Instead, rely on your current performance. The CPC/CVR method above will work well, but instead use current average CVR in step 4. You can also revise the method to use CPM and or CTR instead of CPC if that makes more sense for your campaigns.
3. Changes in the overall economy
If the past decade has proven anything, it’s that the global economy is volatile. Components as diverse as oil prices, subprime mortgages, and political climates can have tremendous economic implications. This in turn affects consumer’s purchasing power and their willingness to buy things from search engine ads. Macroeconomic changes are much harder to factor into PPC forecasting and should be considered with the grainiest of salts. However, it is always good to keep your finger on the pulse of a number of macroeconomic measures as these can provide general indications of overall economic health. If the economy is performing poorly, you may want to scale back PPC budgets and anticipate lower CVRs – vice versa if the economy is doing well. Here is a brief list of economic measures to keep in mind:
- Gross Domestic Product (GDP)
- GDP per capita
- Gross National Income (GNI) per capita
- Inflation/Real Income (check out the Consumer Price Index and the Big Mac Index)
- Changes in employment/unemployment statistics (Bureau of Labor Statistics)
Good luck! I’m happy to answer any questions, so please leave a comment.