I’m writing this new book on pricing in eCommerce. I’ve written one chapter so far and like everything else I do in life, I want to product manage this. There are two things I’m in the dark on:

  1. Is what I’m writing of interest to anyone?
  2. If yes, what more would people like to read?

So, I decided to publish the first chapter and table of contents that I had in mind on this blog. If you like reading this, please share your feedback via this Google form.

Northstar Metric: contribution margin/CAC

Say you paid me $100 for a bottle of my finest Islay single malt. This is my revenue. I’d bought the bottle from a distiller earlier for $40, called cost of goods sold (COGS). This makes me a cool $60 as profit, also called gross profit. However, this is only a part of the transaction.

I had to pay a logistics firm to ship the bottles from the distiller to me and then to your house. I also had to securely pack them to avoid any damage during transit totalling up to $20 for each bottle I sold. This makes my actual profit per bottle 60-20 = $40, also called contribution margin.

But wait, I paid $1000 for an Instagram ad last week which is how you knew about my fine scotch collection. I’ve been selling to ~40 customers weekly and I spent 1000/40 = $25 as Customer Acquisition Cost (CAC). But wait, I also take my own salary, employ people to maintain the office, pay my utility bills and rent. All of that is another $10 per bottle so I really only made 40-25-10 = $5, my actual-actual profit. Heck, all this effort and I still have to pay tax to the government! Appropriately, this profit is called Earnings Before Interest, Taxes, Depreciation, and Amortization (EBITDA). Visually, this is what the breakdown looks like.

An important observation here is that the example above aren’t real numbers for an online store selling whisk(e)y. There is just no way that a product with no fixed expiry date can be sold for 2.5x the price from a distiller to the end consumer. A wise man once said “your margin is my opportunity” 1 and proceeded to dominate all of ecommerce. Let’s be honest though: what kind of person would I be if my imaginary store sold potatoes? No sire. Not under my watch. The numbers I’ve quoted are a ‘truer’ reflection of an online grocery store but it is obvious that in the early stages of a firm, there is a lot more money being spent on logistics, marketing and general expenses resulting in negative EBITDA margins. It doesn’t (necessarily) mean that the firms are run poorly, they’re just more forward looking than traditional firms. But, I digress. Now that I have, let’s go down one more path!

Let’s say the bottle’s quoted price was $110 but I offered you a $10 discount for being a loyal patron, I never received that money so it isn’t counted in the revenue, appropriately called “net” revenue. Similarly, any refunds I had to issue for a bad customer experience also isn’t net revenue. The $110 figure is called Gross Merchandize Value (GMV) which is an important measure if I was doing multiple types of business. For instance, say my store also stocked Speyside single malt from a friend’s distillery. She manages the entire logistics and shipment herself once I get a confirmed order but just pays me $2 per bottle for placement (my revenue). It would be unfair to evaluate the value of my virtual shelf space basis revenue alone so GMV helps. The interesting bit here is that I didn’t have to do any additional work for that $2 so it is directly added to my contribution margin. In e-commerce parlance, I am operating a ‘marketplace’ where my friend could sell her inventory.

My core argument in this book is that every pricing team should maximize the ratio of their contribution margin earned per customer and the cost of acquiring this customer.

Let’s create 3 imaginary businesses: a Product firm run by Priyanka, a Services (SaaS) firm by Smriti and a Marketplace run by Manu. All three hire me as a pricing consultant and we start a conversation on what northstar metric to optimize for.

Continuing the previous example, Priyanka’s product firm is a retailer of whiskey and sells in grocery stores, 3 branded stores and on their own website. She creates a first draft of her firm’s OKRs (Objective and Key Results) 2 and sends me, her pricing and marketing leads this email

Objective: Become the biggest online whiskey retailer in the world in 5 years
Key Result 1 (pricing team): Increase _northstar metric_ by 100% this year
Key Result 2 (marketing team): Reduce Customer Acquisition Cost (CAC) by 10% 

We need to pick the best northstar metric based on three criteria. This metric should

  1. Disincentivize short-term tradeoffs for long-term growth
  2. Be measured daily and allow for trend evaluation in weekly catch ups
  3. Be directly controlled by the pricing lead with minimal overlaps

Smriti is running a SaaS business creating data analytics tools and her questions are exactly similar. The major difference is that her cost of goods and fulfillment are so low that revenue and contribution margin are pretty close to each other. What she worries more about is customer retention and recurring revenue. Manu who operates a marketplace for photographers has few real costs; namely marketing and server hosting. So, for simplicity of narrative, we’ll run with the product example till we evaluate contribution margin/CAC as a metric and then detail all the three types of businesses.

Northstar Metric = GMV

On receiving Priyanka’s email for her product firm, the first exercise we ran together was to evaluate together which metric would work best and started with GMV. I sent her back a simple equation.

GMV in one year = number of customers * number of orders per customer in one year * average marked price of each order

This is a good measure of how well her business is doing. However, there is a predicament to solve for. The number of customers in the store each year is largely determined by marketing spend 3 and the marketing lead has to cut her spend per customer by 10%.

Increasing the number of orders per customer is a function of available inventory width 4 (i.e., can we serve all possible customer needs from cigars to wine openers?). This is completely outside the control of the pricing lead.

The third variable, marked price is a static number printed on the bottles already. So, how should he move GMV?

He pulls out last week’s data on my request. Of every 100 people who walk into the online store (traffic), only 3 people buy anything (conversions). The rest are just window-shopping. Let’s call this 3% as Conversion Rate (CR). He benchmarks this with some of our peers to realize that 3% is pretty good. It’s not like we’re serving an urgent need like an on-demand cab service. Whisk(e)y is a luxury after all and can be replaced by other vices. His lightbulb 💡 moment is this: if we run a promotion to drop prices even further, he can convince more of the window shoppers to buy something! Not just loyal consumers, any consumer. He makes an assumption that by dropping prices by 10%, he can convert exactly 10% of the window shoppers and writes down this table as his first pricing proposal.

Heads Now Proposal % change
Marketing cost 1000 1000
Traffic 1333 1333
Marked price 110 110
Retail price 100 90 -10%
Conversion Rate 3% 3.3% +10%
Conversions 40 44 +10%
Revenue = conversions * retail price 4000 3960 -1%
GMV = conversions * marked price 4400 4840 +10%
CAC = marketing cost / conversions 25 22.7 -9.1%

This creates three obvious challenges for Priyanka and she creates a three point rebuttal for the proposal

  1. The pricing lead has no incentive to ever raise prices. In his proposal, revenue is dropping and the firm is making even lesser profit per bottle with cheaper prices but GMV is 10% higher. This is an unacceptable distortion.
  2. His peers love this idea. For the marketing lead being measured on reducing CAC, this is lovely news. She’d already met her CAC target because of the price drop. The management team should ideally have some healthy conflict and tradeoff on OKRs which push the overall firm forward.
  3. The assumed conversion rate increase seems too low. During past promotions giving 10% off, she’d seen much higher revenue increases but that could also be because of marketing spends and she hadn’t separated out the effects cleanly

Hence, picking GMV is sub-par. Maybe we should experiment with revenue?

Northstar Metric = Revenue

In the above example, revenue seemed to reflect a much healthier tradeoff. It dropped with the price drop and seemed harder to game but since Priyanka was skeptical, the pricing lead ran a simulation with different values of conversion rate for the same price drop of 10%. Elasticity is the ratio of the conversion rate increase to the price decrease. A minor change to elasticity meant they’d either lose or gain a lot of money.

Conversion rate Elasticity # conversions Revenue
6% -0.6 42.2 3802
8% -0.8 43.1 3881
10% -1 44.0 3960
12% -1.2 44.9 4039
14% -1.4 45.8 4118

He reached out to the marketing lead to inquire what the range of this number has been in past promotions and received an unexpected insight. Whenever they’d run a promotion in the past, the prices were reduced 2-3 hours before the marketing spends really kicked in. In this period, the big-name brands which accounted for most of the revenue flew off the virtual shelves. Her assertion was that everyone into the big-name brands knew the product well and that the price is cheaper than their usual store. Not just that, people bought multiple bottles in one purchase to gift friends for the holidays. This behaviour was similar for both cheaper mass brands and much pricier niche brands.

She did some back of the envelope math for the last promotion and suggested that for big-name brands, conversion rate had increased by 20%-30% for a 10% price drop. Elasticity is actually closer to 2-3 for these brands, way larger than the original calculation! In some promotions, the average value of the order had also stayed steady indicating that people either bought more bottles or larger sizes. For the smaller distilleries where consumers didn’t have a set expectation of the product quality, they’d seen a 10%-15% increase in demand for a 10% price drop, again, higher than the initial guess.

This mirrors the author’s own experience across industries. In the digital channels, elasticity is always greater than 1 and sometimes as high as 4. It is highest in online travel (specifically, flights where it is easily >3) since aggregators make it easy to surface promotions and the travel experience is exactly the same. At the opposite end, elasticity is lowest for private-label fashion products and groceries but is greater than 1 at nearly all price levels. This means that revenue keeps increasing with price drops unlike what economics textbooks predict. The sane textbook prediction is that revenue should reach a maximum and then start decreasing which the real world refuses to accept. Answering “why” is hard but user research suggests that at each price level, retailers unlock a different market segment altogether and demand isn’t fixed. For instance, it becomes profitable for small retailers to buy from larger retailers and become resellers. Someone might counter that this party has to end at some point and I would humbly concede the argument. Literature validation for this is also available at 5 and 6.

In my weekly catchup with the leadership team, we concluded that this created three major challenges.

  1. If elasticity is always greater than 1, revenue will increase on dropping prices and profitability will drop. This is a race to the bottom.
  2. The lightbulb 💡 moment from this chat for Priyanka was that price drops cause people to buy more stuff and this puts the firm on a higher growth trajectory. In nearly all cases, she has volume incentives from suppliers which look like this:
Number of bottles per week Per bottle price
1,000 40
5,000 38
10,000 36

This implies that if the weekly volumes are close to these thresholds, she could offer a discount to her users without losing any profit. Just measuring revenue is hiding this part of the picture.

  1. Priyanka’s largest discomfort in these two conversations was that each $ of revenue and GMV was being considered equal. In reality, one $ of revenue from the marketplace business where she enabled her friends is more profitable than her owned inventory. The user experience for the marketplace is worse off since marketplace boxes typically took longer to deliver and were damaged more frequently!

Maybe we should be using gross profit which also accounts for these volume incentives?

Northstar Metric = Gross Profit

The pricing lead setup some time with the operations head to try to understand what affects the cost of purchasing whiskey. The operations head was quite knowledgeable and pointed out that they buy whiskey at different prices from multiple wholesalers, whenever they find the best deal.

At the point in time that the user makes a payment, they just know that the bottle(s) would ship from the warehouse in city X since it is closest to the user’s pincode. But they don’t have a way of assuring that this particular one which had cost $Y to buy would be picked up. They get to know of this only once the barcode on the bottle being shipped is scanned. Hence, actual costs come in after ~48 hours and if we wanted a more frequent report, we’d have to estimate costs.

All ecommerce firms the author has worked for try to barcode items individually and stack them in specific shelving such that one could follow a specific operational strategy e.g., ship the item with the nearest expiry date first or favour items from specific sellers where we’re close to meeting our volume targets. It makes a metric tonne of logical sense but it requires a level of capital expenditure in automation, both while stocking and retrieval which most firms struggle to achieve. Being able to build such a platform is a game-changing cost advantage.

The key takeaway 🔑 from the meeting was that unlike revenue which was clearly for each order, any cost element would first need an estimation which would become more accurate after a few days.

The pricing head then refined his proposal to incorporate the higher elasticity we generally see (~2.6) and in our next catch-up, created this proposal only to discount our big-name brands to suggest that we can actually gain market share without losing profitability.

Now Proposal % change
Net revenue per order 100 90 -10%
Cost of goods per order 50 50
Gross profit = Revenue - COGS 50 40
Conversion Rate 3% 3.8% +26%
Conversions 40 50 +25%
Total revenue 4000 4500 +12.5%
Total gross profit 2000 2000 0%

As we discussed this in more depth, we realized this proposal has a major flaw: gross profit and price are tightly linked. A $10 reduction in price is exactly a $10 reduction in gross profit since the whiskey has already been purchased in the past. That cost doesn’t change because of our decisions today. As the marketing lead had suggested earlier, running a promotion leads to people buying more bottles in each order. Heavier boxes are costlier to ship and need more packing material. This means that the future spends we’re about to make for fulfilling the order would be better reflected in contribution margin.

One of the key reasons the author picked a product that doesn’t expire easily was to highlight this contribution margin linkage. For products that do expire, what a firm typically does is put an estimate on their value over time something like this:

This implies that if Priyanka as a CEO expected to make some gross margin in the first few days, she’d revise her expectations down over time to make sure she isn’t left with inventory no one wants to buy. This has no impact on any cost element till the value drops to zero; which happens frequently in very short expiry businesses like groceries. The firm then has to spend money to dispose of the goods. What this would mean for the pricing lead is just a different set of margin targets for different time windows. For instance,

Age of inventory (days) Gross margin target
30 60%
60 50%
90 40%

An exception to the example above is if the retailer bought the item on credit from a seller (e.g., a used car retailer) and can return the unsold merchandize. This implies that if a sale doesn’t happen, both the original seller and retailer are willing to take a haircut on price. This would mean that the gross margin target might stay unchanged despite a “for sale” sign on the front door.

Northstar Metric = Contribution Margin

The pricing lead setup a second meeting, this time with the CFO as these cost estimations were running more into the finance domain. The CFO explained that there are three types of cost which go into selling the whiskey

  1. Direct cost: The cost can be directly attributed to fulfilling customer orders and that is directly linked to each bottle shipped. For example, material costs in packaging, labour costs for assembly, logistics, etc.
  2. Fixed Costs: Purchase costs for machines used for handling or repacking. These costs are typically distributed over time since machinery could be used for 10-15 years and looking at the entire cost in the month of purchase would make that month and year look quite unprofitable when it clearly isn’t. Accountants call the process depreciation and amortization.
  3. Overhead costs: Mostly people and utility costs like salaries, travel expenses, rent, electricity bills, etc.

She argued that for pricing, we should only care about the first set i.e., direct costs. By definition, fixed and overhead costs don’t change with the volume of orders placed and pricing decisions should have no impact on them7. Secondly, some indirect costs become clearer only on specific dates (e.g., utility bills) and estimating everything back to a specific order was tricky because the order could also be refunded. The administrative complexity of generating a daily EBITDA estimate for each product sold is too high. In comparison, we get the contribution margin by just deducting the direct costs from gross profit. When the CFO interacts with her peers or with the government, she would include all types of cost and share EBITDA margins for deeper scrutiny but this would be after each month’s closing.

Using this framework, the pricing lead refined his earlier proposal to realize that we’d actually lose contribution margin and he would need a different price point if we wanted to maximize profits.

Now Proposal % change
Net revenue per order 100 90 -10%
Cost of goods per order 50 50
Direct costs per order 4 5
Contribution margin per order 46 35 -24%
Conversion Rate 3% 3.8% +26%
Conversions 40 50 +25%
Total revenue 4000 4500 +12.5%
Total contribution margin 1840 1750 -4.8%

As he sent this learning out to the same audience, the marketing lead had a counter. Her argument was that there are some channels e.g., television ads where she has to spend a lot for acquiring customers and some like the referral program where she doesn’t have to spend any money at all. If we look at the cost of generating this profit, this same proposal might still work for some of her low cost marketing channels.

Northstar Metric = contribution margin / CAC

The pricing lead then jumped into how we estimated CAC. It seemed quite easy for the online marketing channels where all of the marketing partners agreed on a standard tracking methodology using UTM parameters 8. Every single transaction that happened could be attributed back to a certain advertisement that a user clicked on.

The online marketing team could also show multiple types of attribution models for users who took a few weeks to make a purchase. For instance, it could be traced that the user clicked on an online ad the first time they came and registered as a user but didn’t complete the payment. Only once they received a discount voucher in their email did they finally convert. The marketing team showed a comparison of attribution models to suggest that some channels (e.g., brand advertising) are important for building awareness while some are critical for driving sales (e.g., targeted discounting emails). It was also quite clear how much money was being spent on each customer to drive a purchase.

For offline advertising though (e.g., TV or podcasts), it was a little trickier since the only signal present was a discount code which might have accompanied the advertisement. For some channels (e.g., posters or hoardings), there was some innovation being used like a QR code stamped on each of them to differentiate them but most offline transactions that didn’t have tracking parameters were hard to attribute.

This created a nice balance for the pricing lead where he could argue that for offline channels with limited targeting, we could use the smallest discount and share the best deals with the best customers. Arguably, if it cost $10 to acquire a customer and the firm made a profit (contribution margin) of more than $10 once they bought something, the business is quite healthy. Measuring contribution margin / acquisition cost and keeping a target of greater than 1 is a great northstar metric.

Just re-evaluating the three parameters for a northstar metric we’d outlined earlier

  1. Disincentivize short-term tradeoffs for long-term growth ✅

    • Keeping contribution margin / CAC at a higher threshold (say 2x) allows for a long-term outlook
  2. Be measured daily and allow for trend evaluation in weekly catch ups ✅

    • Both contribution margin and CAC for individual products sold and users can be estimated daily and trends evaluated
  3. Be directly controlled by the pricing lead with minimal overlaps ✅

    • The promotion proposals are controllable directly by the pricing team with minimal overlap with marketing or operations

Pausing here for a moment and picking up a conversation with Smriti for her subscription business of analytics tools. What does contribution margin for her look like?

Contribution margin = number of customers * number of orders per customer * (revenue - COGS - direct costs)

The equation here is exactly similar apart from the number of orders bit. She expects her customers to pay her a fixed amount each month if they’re happy with the service. This means that she could’ve started with a free trial period where she made no money and then a profit after a few months. Visually, this is what her profit for such customers would look like:

Two points of note:

  1. Contribution margin starts out negative since she gave a free trial (or an equivalent promotion). This means that unlike most product firms that have to be profitable on each order, subscription firms can make a loss on their first order in the hope of recovering this money later. This is not an obligation, more a norm.
  2. The straight line indicates that all customers keep paying her. However, not everyone will. Some would cancel their subscription if they weren’t happy with the service. This would extend her expected break-even time. Her profit curve across a cohort of customers would trend downwards like this

This implies that for each of her pricing proposals, she needs to estimate what would happen to user retention and her contribution margin over time. For instance, if she decides to evaluate a pricing proposal of the form below, what she’s looking for is a reasonable time window (typically 1 year but could be longer depending on the duration of subscription) where one of the proposals definitely wins out.

Now Proposal
Monthly price: $10 Monthly price: $10
Yearly price: $120 $99

For the yearly subscription, the user received a $21 discount but had to pay the entire amount upfront after the free trial. Given the large sticker shock, fewer customers might convert at the end of the trial. Whereas users on the monthly subscription got no promotion apart from the free trial and kept paying over time. At some point during the year, they might have shown a higher profit.

Again, re-evaluating the three parameters for the northstar metric of contribution margin / CAC for this business:

  1. Disincentivize short-term tradeoffs for long-term growth ✅

    • Measuring contribution margin over the retention period ensures that the overall profitability is accounted
  2. Be measured daily and allow for trend evaluation in weekly catch ups ✅

    • Recurring subscription revenue which comes in everyday for past acquisitions help refine the models for the initial retention estimates
  3. Be directly controlled by the pricing lead with minimal overlaps ✅

    • The incentives for both pricing and marketing teams are quite aligned even in case of overlaps

Now evaluating the case for a marketplace which Manu operates to hire wedding photographers. Manu operates an agency that employs a few hundred artists directly listed on the platform while other freelance artists and agencies can create their own profiles.

Always optimizing for contribution margin gives Manu a distortionary incentive to sell artists from his own agency instead of others. This deliberate influence on consumer choice hurts the platform longer term. He needs guardrails for seller experience metrics to protect marketplace health9. These guardrails could look like artificial volume caps (e.g., private labels can only be maximum x% of daily sales)10 or happiness metrics for sellers (e.g., churn rate, net promoter score) or taking all control of pricing and promotions away from the pricing team and handing it to individual artists putting their listings11.

The second aspect unique to marketplaces is a tradeoff that can be made between contribution margin and quality of customer service. For instance, let’s compare two photographers on the platform who offer the same price to the end-consumer:

Marketplace Commission User Rating
Artist 1 $100 5/5 (100 users)
Artist 2 $150 2/5 (100 users)

One point to keep in mind is that since people typically don’t get married very frequently 🤞, the marketplace is offering a one-time service and not a subscription. Giving the pricing team an incentive to promote Artist 2 who consistently provides a worse service is harmful longer term. This puts forth another set of guardrails on the customer experience metrics e.g., user ratings, referral k-factor12 (a measure of how many new users are being referred by existing customers), etc.

The third aspect is that it costs money to acquire both the buyer and the seller. Most marketplaces are such that one seller serves hundreds of buyers e.g., cab services, doctors, restaurant delivery, etc. However, a few are well balanced e.g., apartment rentals where one seller serves fewer buyers.

Hence, marketplaces need a northstar metric which takes all these factors into account. The cleanest way of doing this is a two fold exercise:

  1. Numerator: convert all of the guardrails into an equivalent contribution margin estimate
  2. Denominator: add the cost of acquiring the seller into CAC

The core idea with converting guardrails into a contribution margin estimate is that simplicity trumps accuracy during major organizational changes. Using past data, it can be estimated how one individual parameter (e.g., adding an additional seller, # of escalation calls in the call center, etc.) impacts acquisition or retention and hence, profitability. It is far easier for every team member to index on a single metric and codify this into their algorithms; both processes and code. Statisticians will correctly argue that this estimation process is bound to be error prone and I’d counter that in two ways:

The core argument with adding seller CAC into the denominator is a concept already accounted for by the product firms like the aforementioned whiskey retailer. Assume that the product firm loses a great employee in the warehouse. They’ll spend money to hire her replacement and train them which would count in the direct costs. Extending that argument, the cost for acquiring and onboarding a new seller into the marketplace should count in both direct costs and CAC. However, since one photographer can serve tens of customers, we can just try to estimate how much to count against each transaction like this:

CAC for sellers = Actual cost for acquiring this seller / estimated # of transactions this year

One template for doing this is to classify the sellers by size and account for their costs accordingly

Size of seller Estimated # of transactions CAC CAC per transaction
Small 100 500 5
Medium 500 1500 3
Large 1000 2000 2

Again, re-evaluating the three parameters for the northstar metric of contribution margin / CAC for a marketplace:

  1. Disincentivize short-term tradeoffs for long-term growth ✅

    • Indexing long-term impact parameters to contribution margin helps make the tradeoffs transparently
  2. Be measured daily and allow for trend evaluation in weekly catch ups ✅

    • The estimates can be made daily and the accuracy of the past estimates shall keep getting better
  3. Be directly controlled by the pricing lead with minimal overlaps ✅

    • The stakeholders involved in a marketplace are higher but the functional overlaps are of a similar complexity

One parting note: why contribution margin / CAC as a ratio and why not the absolute value of contribution margin - CAC? It can be any. It doesn’t make a difference. The ratio just makes intuitive sense. I spend $1 to acquire a customer and if I make more than $1 as profit, I’m golden. The simple mathematical proof is this:

Contribution margin - CAC > 0
Contribution margin > CAC
Contribution margin / CAC > CAC / CAC
Contribution margin / CAC > 1

Future table of contents

What I want to write next:

  1. A/B testing the northstar metric
  2. Pricing individual products
  3. Pricing a product portfolio
  4. Pricing a subscription
  5. Designing promotions
  6. Case study: pricing this book

I’m amazed that you got this far. If you liked this, please share your feedback via this Google form. You can make any suggestions via the form; on the current or future content. Thanks for being awesome!

Regards, Ravdeep

  1. Jeff Bezos. 7 Jeff Bezos Quotes That Outline the Secret to Success. I frequently quote Amazon in this book as a benchmark for how efficiently eCommerce can be run.↩︎

  2. Read John Doerr’s excellent book, Measure what matters for a more detailed view on OKRs, why and how to set them.↩︎

  3. Having an excellent product that sells itself is the best growth strategy. However, that is a problem statement for the distillers making the elixir, not the retailers!↩︎

  4. Incentivizing people to keep ordering also has multiple marketing levers e.g., running a loyalty program covered in the promotions section later.↩︎

  5. Fisher, Marshall and Gallino, Santiago and Li, Jun, Competition-Based Dynamic Pricing in Online Retailing: A Methodology Validated with Field Experiments PDF↩︎

  6. Granados, N., A. Gupta and R. Kauffman. “Online and Offline Demand and Price Elasticities: Evidence from the Air Travel Industry.” Inf. Syst. Res. 23 (2012): 164-181. PDF↩︎

  7. This is technically inaccurate. You could have a situation where for generating more sales, you need to run a facility for longer and spend more on electricity or water. However, unless this is a core part of the product itself, for instance your firm makes ice or soda, this aspect would still be a fixed cost.↩︎

  8. Urchin was a startup acquired by Google in 2005 that created Urchin Tracking Module (UTM) parameters now deeply embedded in Google Analytics. These parameters are simple and scalable, an envious quality of a solution to a vexing problem. Because of this, it has become the standard adopted by all online advertisers where marketers just need to configure some tags into the link of a page and all reporting tools respect those tags. More details at the wikipedia link explaining UTM parameters↩︎

  9. Amazon quite famously got into trouble for biasing it’s visibility algorithms towards profitability Amazon changed search algorithm in ways that boost its own products↩︎

  10. In a congressional hearing in Sep 2020, Amazon disclosed that 9% of it’s sales in clothing are from its private label brands↩︎

  11. Wherever possible, marketplaces put the onus of pricing and approvals for promotions on the sellers themselves encouraging competition. Amazon’s fair pricing policy↩︎

  12. i= number of invites sent by each customer (e.g. if each new customer invites five friends, i = 5). c= percent conversion of each invite (e.g. if one in five invitees convert to new users, c = .2). k-factor = i * c. Wikipedia page on k-factor↩︎