Conjoint Simulator Assumptions

It's tempting to not want to make any assumptions. However, not making any assumptions is the same as the default assumptions which is usually the wrong way to go.

For conjoint, the assumptions we need are either about the costs and accounting of the company or the supply conditions in the market.

By providing better assumptions, the outcome of the simulator should better match the marketplace outcomes.

By default the simulated outcomes assume that everyone is aware of all of the options. But that is rarely the case.

Awareness comes from outside of the choice tasks. Most of the time we will have a separate survey question asking which brands the respondent aware of. The aggregate proportion of aware consumers can then be piped into the simulator.

In many consumer packaged goods (cpg) categories distribution is easy. Syndicated sales data has a variable called ACV% for each sku.

ACV stands for all commodities volume. It is a measure of the proportion of sales channels each product is in. It is weighted by proportion of the category purchases that occur in each channel.

A common mistake is to figure that our products are on Amazon.com therefore distribution doesn't matter. That would be true if all of the category sales were via amazon. That is never the case.

In our experience a ball park guess on distribution numbers is far superior to leaving everything at the default 100% assumption.

For each component of a product there is a cost associated with delivering an additional unit.

In a 48g Snicker's bar, there might be 11g of peanuts. One scenario might replace the peanuts with peanut butter. Including the costs of the material in the cost calculations provides a way to get a better understanding of the profitability of different product moves in the marketplace.

If all of the levels have the same cost to include/deliver than you can skip including variable costs in the simulator.

The variable cost for delivering each feature could look like this:

We left off the price column, because it doesn't ever change the cost to deliver the product.

Some features do not cost any extra when additional units are sold. Think about features for a piece of software - no variable costs.

When a feature does not have variable costs it is more likely to have a fixed cost. So a software feature would cost $25,000 to develop regardless of the number of users it supports.

Another example of fixed product costs would be a donation or contribution to a cause. You only pay that once regardless of how many candy bars you sell.

The fixed costs for the product features could be laid out like this:

There are a bunch of other costs that a brand carries that are not directly related to the products/services they offer. For example payroll would have a really week connection to the number of units sold and would not be needed for most conjoint studies.

These fixed brand costs are not needed for the conjoint simulator. Including them will have no impact on the revenue or profit optimizing price points. When you subtract the fixed brand costs, these curves shift downward and keep their peak at the same price point.