You will want to represent this on a unit basis, i.e., a single, average user. If you have distinctly different classes of products and customers, you can separate accordingly to add clarity.
For CAC, you should look at the total acquisition-related costs of bringing an average new user into your service and converting that user to a revenue-generating user (acquisition-related costs would only include expenses directly attributable to driving new users to your service). If you have an advertising-based revenue model, this would be the cost to generate an impression, a click through, an action, etc... based on your revenue structure. Users acquired organically have a zero cost (SEO, viral, etc...) and will average your CAC down. You will likely want to model different conversion funnels by acquisition channel, and your expected mix by acquisition channel, to come to a blended average rate. It will be helpful to reflect these assumptions as you explain your model.
For LTV, you will need to estimate the total revenue generated by the average revenue generating user (i.e., the user from the above CAC profile). For an advertising model this would be frequency and magnitude of engagement over time; e.g., in a simplified form this could be pageviews per month over the total number of months the user will engage with your service, using your revenue model to calculate a total dollar amount generated.
As an example:
Let's look at a single channel for simplicity. Let's say you've struck a partnership deal where you will spend $10,000/mo to have prominent place on a website in an adjacent market segment. Over the course of the month, one-million unique users will see your promotional unit and you expect 1.5% to click through. Once a user clicks through, they hit your site, immediately see ads on your site, and now are officially generating revenue.This would produce 15,000 revenue-generating users at a CAC of $0.67 ($10,000 spend / 15,000 acquired users).
Let's say over their lifetime, these 15,000 users will generate 1MM cumulative pageviews, your ad revenue model is CPM based, and your rate is $5/CPM. This would yield $5,000 in total revenue for the lifetime of these users, and an LTV of $0.33 ($5,000 revenue / 15,000 acquired users).
That would produce a negative ROI, not viable.
Now let's look at incorporating your ecommerce model and update the LTV definition to reflect the change.
Let's say of the acquired users who hit your site, 5% will convert and transact against your eCommerce offering, with an average purchase price of $250, and an average of 2 purchases over their lifetime. With these assumptions applied to the above 15,000 acquired users, you would generate an additional $375,000 in revenue producing an eCommerce-based LTV of $25 ($375,000 eCommerce spend / 15,000 acquired users).
Combined with your ad revenue, your total LTV is now $25.33 against a $0.67 CAC, now a stellar ROI of ~3,700% (and you may question the continuation of your advertising revenue efforts).
Take this same framework for evaluating a partnership and apply it to your broader marketing efforts and you will have CAC and LTV at an enterprise level.
Note that your assumptions and economics may change over time, so you may find a time-based model (e.g., month-by-month over 3-5 years) best reflects your customer economics. This will also lend itself to reflecting the launch of your ecommerce model in the future, as some % of users will transact against that product, as well, driving their LTV up, as illustrated.
Having actual data to drive your model is ideal, but not always possible. Where it is not possible, use defensible industry comps with sources. If you end up relying heavily on assumptions, you may want to reflect a range for CAC and LTV and be prepared to speak well to the rationale behind it.