The Mathematical Truth about VDPs, Price to Market and Sales. Noah John Co-founder Autoscores Orlando, FL (330)

Transcription

1 The Mathematical Truth about VDPs, Price to Market and Sales Noah John Co-founder Autoscores Orlando, FL (330)

2 The views and opinions presented in this educational program and any accompanying handout material are those of the speakers, and do not necessarily represent the views or opinions of NADA. The speakers are not NADA representatives, and their presence on the program is not a NADA endorsement or sponsorship of the speaker or the speaker s company, product, or services. Nothing that is presented during this educational program is intended as legal advice, and this program may not address all federal, state, or local regulatory or other legal issues raised by the subject matter it addresses. The purpose of the program is to help dealers improve the effectiveness of their business practices. The information presented is also not intended to urge or suggest that dealers adopt any specific practices or policies for their dealerships, nor is it intended to encourage concerted action among competitors or any other action on the part of dealers that would in any manner fix or stabilize the price or any element of the price of any good or service. 2

3 Table of Contents Learning Objectives...4 Computing sales correlations using Excel...5 VDP view case study results...6 Influences on VDP views sales correlation...7 How to modify Google Analytics to start capturing all your online customers clicks on your VDPs...8 A basic approach to predicting your inventory car sales step A basic approach to predicting your inventory car sales step A basic approach to predicting your inventory car sales step How our dealership benefited

4 Learning Objectives Upon completion of this workshop, you will be able to: 1) Compare sales predictors, like VDP views and Price to Market, by calculating their statistical correlation to inventory sales. 2) Understand how current sales predictors, like VDP views and Price to Market, correlate with sales across different North American and European dealerships, according to a recent case study. 3) Construct a predictive model in Excel that produces a sales predictor that typically correlates 2 to 4 times more to inventory sales than VDP views. 4

5 Computing sales correlations using Excel Excel has a function called CORREL that computes the correlation of two variables. To use it, you must select two columns (or rows). In the following screenshot, we ll compute the sales correlation of VDP views per day: There are 5 columns in this table. The first column contains all the vehicles in your inventory. The second column has VDP views/day. The third column contains the date for the given row. The fourth column tells us if the vehicle sold on the row s date. The final, fifth column converts the fourth column No s and Yes s to 0 and 1 respectively. In the above table, when we look at the first row, we can see that our 2014 Acura TL was receiving 1 VDP view/day on 9/3/17 and it did not sell. The date column is important because you should have at least a month s worth of data before computing the sales correlation. Once this table is created in Excel, you can compute the correlation by applying the CORREL function to the 2 nd and 5 th columns, which are highlighted in blue. The cool thing is that you can apply this method to compute the sales correlation for any metric, provided that: you can measure the metric per day and per vehicle in your inventory. 5

6 VDP view case study results The following table summarizes the results of computing the correlation between VDP views per day and sales at four different dealerships in North America. The case study was conducted over a 1-2 year period and all correlations computed were significant (p < ). For all dealerships in the study, the correlation between VDP views per day and sales never exceeded 6%, which is a very weak correlation. In general, VDP views tell you very little about future sales. However, there can always be exceptions to this rule. Any dealership interested in knowing the correlation between VDP views per day and sales should compute the correlation themselves. The case study s results should be treated as a rule of thumb, not the letter of the law. 6

7 Influences on VDP views sales correlation There are many factors that affect the correlation between VDP views and sales. Here s a simple example to illustrate that idea. Imagine we have two dealerships, each with 10 vehicles each, as shown in the figure below. Dealer A Dealer B Dealer A contains ten, 2016 Toyota Priuses, each with similar trim and mileage. Dealer B contains only one 2016 Toyota Prius and nine different vehicles. In this scenario, Dealer A s VDP views will likely correlate less to sales than Dealer B s VDP views. This is because, a person who goes to one VDP on Dealer A s website is likely to look at more VDPs because the entire inventory has similar cars. However, the person will almost always buy only one of the cars, not all ten. In other words, a shopper will likely visit all ten VDPs before choosing to buy one of the cars. Conversely, if someone views a car on Dealer B s website, they likely won t view another VDP, because there aren t similar vehicles that would interest them. Meaning, a shopper will visit one VDP before buying the vehicle. Since Dealer A s shoppers look at most vehicles before choosing one to buy, Dealer A s correlation between VDP views and sales should be lower than Dealer B s correlation. 7

8 How to modify Google Analytics to start capturing all your online customers clicks on your VDPs Google Analytics has the ability to record and track events. In order to start tracking events, you have to tell Google Analytics what clickable elements of your VDPs to watch. Clickable elements can be VDP elements like links, buttons, pictures, and form boxes. Your webmaster or web provider should be able to help you add events to the clickable elements that you desire tracked. Once Google Analytics is recording the events you want, you can view this new event information in the Google Analytics web interface. If you don t have login credentials to your Google Analytics account because it was setup by a third-party, it is highly recommended you obtain them. 8

9 A basic approach to predicting your inventory car sales step 1 Once you start collecting events on your VDPs, you need a way to analyze this information and make predictions. Here s a simple, but effective way to do this: First, assign point values to each event that occurs on your VDP that you re tracking. The trick is that you have to decide how much an action suggests someone is interested in buying a car. This is not an easy task and typically requires you to come back later and experiment with different point values. In the figure below, we ve listed some example VDP events in the first column and assigned point values to each in the second column. You can assign any point values you want as long as the ratio among the point values assigned to each event makes sense to you. Figure 1: Assigning point values to VDP events 9

10 A basic approach to predicting your inventory car sales step 2 Once you have assigned point values to all your events, you need to perform the following calculation for each VDP in your inventory: 1. Collect the event totals for each event and multiple each total by the corresponding point value. 2. Sum up the products to obtain a single number for a single VDP. In the figure below, we re totaling up the points for a single VDP of a 2008 Audi A4. Column 1 lists all the events we re tracking. Column 2 contains the point values we assigned to each action in step 1. Column 3 contains the number of times each event occurred on the VDP in the last 2 weeks. Column 4 contains the result of multiplying point values by number of events - in other words, the product of multiplying columns 2 and 3. After computing the values in column 4, sum them up to obtain a point total for the vehicle. In our case, the 2008 Audi A4 attained 744 points. A spreadsheet program like Microsoft Excel can be very helpful in performing these calculations. Figure 2: Totaling up points for a 2008 Audi A4 VDP. 10

11 A basic approach to predicting your inventory car sales step 3 The final step is to sort your entire inventory by these point totals that you ve calculated for each VDP in step 2. You should have a list similar to what you see below in figure 3. Over time you should see that vehicles over a certain point value will sell noticeably faster than the rest of the inventory. In general, the higher the point total, the faster the car should sell. As noted above in step 1, in practice it will typically require significant experimentation and adjustment of the point values before you can produce consistent and accurate predictions. However, with some deliberate and careful assignment of point values, you should achieve accurate predictions on when cars will sell fast and when cars will sit. Figure 3: A final list of point values for each vehicle. 11

12 How our dealership benefited During the presentation, we talked extensively about a dealership, Canalcar, that decided to invest time in predicting car sales. Here s a summary of some of the noteworthy findings. These findings are the result of 3 years of work and research with this dealership. The point value sales prediction metric correlated 28% with sales. We outperformed our older, previous repricing metrics, such as VDP views, s, or calls. The average improvement was 5 times better. While this metric does not give us clairvoyance, it certainly gives us a much more confident metric to use in our repricing decisions. Monthly rotation increased by 22% over 2 years without affecting net profit margin. Obviously, other factors contributed to the increase in rotation as well. However, this increase was due in part to the use of car sales predictions. We adjust prices in smaller amounts (1%-3%) at higher frequency. We will drop a car s price as early as 3 days after listing it. Goal is to see online user clicks that demonstrate buying interest, then hold at price. In the past, the dealership would set the price of a vehicle using market factors and third-party market tools. Then the GM would wait 14 days before revisiting the vehicle if it hadn t sold. The GM would investigate if the vehicle had any action by talking with the sales department. If the GM decided the vehicle needed a price drop, he would reduce the price and wait another fixed amount of time. Now, the dealer sets the price of a vehicle the same way. However, he no longer waits 14 days before revisiting the vehicle s price. He will refer to the sales prediction of the car on an almost daily basis and lower the price of the vehicle slightly if the likelihood to sell is below a certain threshold. This proactive pricing strategy gained traction in the dealership over time as the GM slowly became more confident in the predictions. In the first 8 months of 2016, the dealer increased the price of 92 cars for an average of 700 euros. 67 of them sold at the increased price within 10 days of the increase. Achieved a total of 46,900 euros in added profit. After many months of using the predictions, the GM very reluctantly raised the price of a vehicle by 500 euros that was predicted to sell very fast. Much to the dealer s surprise and delight, the vehicle sold 2 days later. Over the course of the next year, the dealer continued to raise the prices of cars that were predicted to sell soon. He began very slowly as he was extremely skeptical. Now, his skepticism is almost gone and he regularly trusts the predictions to raise the prices of vehicles as a new way to generate profit. 12