Which CLV formula should I use if my store is less than a year old?

Use a short fixed window rather than a projection. Three or six month cohort CLV is observable after a single quarter and lets you compare cohorts immediately. Avoid the lifespan formula at this stage, because you have no evidence for the lifespan term and the result will mostly reflect whatever number you guessed. Extend the window to twelve months once your earliest cohorts have actually completed twelve months of trading.

Should CLV be calculated on revenue or on profit?

On profit, using gross margin after refunds. Revenue CLV is easier to pull because your platform reports it by default, but it overstates what a customer is worth by roughly the inverse of your margin. In the worked example above the same cohort produces 2,160 TL on revenue and 800 TL after refunds and a 40% gross margin. Only the second figure can safely be compared against acquisition cost.

What about the discounted CLV formula with a retention rate?

That version applies a retention rate and a discount rate to projected future cash flows, and it belongs to subscription and finance modelling where the horizon runs several years. For advertising decisions made on a twelve month window it adds machinery without changing the decision much, since the uncertainty in your retention estimate is larger than the effect of discounting a single year. Reach for it when your window genuinely spans multiple years.

How often should CLV be recalculated?

Once a quarter is enough for most accounts, plus a recalculation whenever something structural changes: a price change, a new product line with a different margin, or a shift in channel mix that brings in a different kind of customer. Each recalculation should add the cohorts that have newly completed your window rather than rebuild the whole history against a new definition.

How the customer lifetime value formula works

The customer lifetime value formula converts a customer relationship into a single number: the profit one customer generates across the whole period they keep buying from you. The awkward part is that there is no single formula. Three different calculations travel under the name CLV, they answer three different questions, and applied to the same dataset they return figures that are not comparable with one another. Choosing the wrong one, or quietly comparing two of them in the same planning conversation, is the most common way this metric gets misused once it drives ad budget decisions.

The three versions of the customer lifetime value formula

Before any arithmetic, decide which question you are asking. Are you measuring what customers have already been worth, projecting what they will be worth, or comparing groups of customers over an identical stretch of time? Each question has its own formula and its own failure mode.

Historic CLV

Add up the gross profit on every order a customer has placed since their first purchase. Nothing is estimated, which makes it the most defensible version and the least useful for planning. Tenure ruins the average: a customer acquired three years ago has had three years to accumulate orders, one acquired last month has had four weeks. Average the base together and you mostly measure how old your customer list is. Historic CLV earns its keep when you are looking at your top accounts individually, not when you are setting an acquisition budget.

Average order value times frequency times lifespan

This is what most people mean when they say the CLV formula. Multiply average order value by how often a customer buys in a year, then by how many years the relationship lasts. Suppose your average order value is 1,200 TL and your customers place 1.8 orders a year. Assume a three year relationship and you get 1,200 × 1.8 × 3 = 6,480 TL of revenue per customer. It is quick and forward looking, and two of its three inputs come from your data while the third is an assumption. Label it carefully: 6,480 TL is a three year figure. It is not a twelve month number and it must never be set beside one.

Cohort CLV inside a fixed window

This is the compromise most ecommerce teams settle on. Take everyone acquired in a single month, freeze a window such as twelve months, and measure only what that group spent inside it. Say 500 customers arrived in January and over the following twelve months placed 900 orders worth 1,080,000 TL in total. Divide revenue by customers: 1,080,000 ÷ 500 = 2,160 TL of revenue per customer over twelve months. The same answer arrives from the other direction, since 900 ÷ 500 = 1.8 orders per customer and 1,200 × 1.8 = 2,160 TL. Nothing is predicted here. The window is arbitrary, but because it is identical for every cohort, February can be compared with January honestly.

The margin correction almost nobody applies

Everything above is revenue, and revenue does not pay for advertising. Two corrections stand between the cohort figure and a number you can act on.

Refunds and cancellations. If 80,000 TL of that 1,080,000 TL came back, the cohort actually kept 1,000,000 TL, so value per customer falls to 1,000,000 ÷ 500 = 2,000 TL.
Gross margin. At a 40% gross margin, the twelve month margin CLV is 2,000 × 0.40 = 800 TL.

That 800 TL is the figure that belongs in a budget conversation, and it sits a long way from the 2,160 TL you started with: the uncorrected number is about 2.7 times the corrected one. Which cost lines belong inside that margin is a separate accounting question. The point here is narrower: skipping the correction inflates every decision downstream by roughly the inverse of your margin, and it does so silently, because the revenue version is the one your ecommerce platform reports by default.

Why the lifespan term is where the number breaks

Return to the projection. Everything in 1,200 × 1.8 × 3 = 6,480 TL is measurable except the 3. Change that one assumption to two years and the result becomes 1,200 × 1.8 × 2 = 4,320 TL, two thirds of the previous answer, from a business that behaved in exactly the same way. A store trading for eighteen months has no evidence for either figure. It has not existed long enough to watch a cohort stop buying, so any multi year lifespan it plugs in is an assumption wearing a decimal point.

That is the practical case for the fixed window version. Twelve months is short enough to actually observe after a year of trading, long enough to capture a second and third purchase in most consumer categories, and identical across cohorts so the comparison carries meaning. If your product is bought once every couple of years, pick a longer window and hold it everywhere. A figure derived over twenty four months is a twenty four month figure in every sentence that follows it, including the sentence where you divide it by acquisition cost. Relabelling it as an annual number is the single error most likely to turn a careful calculation into a bad budget.

What the finished number is for

There are two uses and only one is interesting. The dull one is reporting CLV as a health score in a monthly deck, where it drifts and nobody acts on it. The useful one is as a ceiling on acquisition cost. If your twelve month margin CLV is 800 TL and you pay 400 TL to acquire a customer, the ratio is 800 ÷ 400 = 2 to 1 on a twelve month basis. Had you used the uncorrected 2,160 TL, the same account would read 2,160 ÷ 400 = 5.4 to 1, roughly 2.7 times the honest figure, which is exactly how an account talks itself into scaling a loss.

Where that threshold should sit, and how a strong ratio built on a long horizon can still leave you short of cash, are separate questions. Both consume the number you have just computed, which is why the window and the margin correction have to be settled before the ratio is worth arguing about. The other side of the ratio has its own levers, covered in how to optimise customer acquisition cost. And note that first purchase ROAS answers a narrower question than CLV: it measures the return on one transaction, which is why an account can look thin on ROAS and still be healthy once repeat orders are counted.

The number also decides what growth decisions optimise toward. Whether a person or an automated scaling process moves the budget, a value signal built from revenue pushes spend toward revenue, and one built from margin pushes it toward margin. The formula you choose quietly becomes the objective, so settle it before you scale.

Producing your first cohort figure

None of this needs a data warehouse. An export and a spreadsheet will get you a defensible first number in an afternoon.

Export every order from the last eighteen months with a customer identifier, an order date and an order total. Guest checkouts without a stable identifier break the join, so decide up front whether you match on email address and accept the error that introduces.
Group customers by the month of their first ever order. That month is their cohort, and they stay in it permanently.
Choose one window and write it at the top of the sheet, then include only cohorts old enough to have completed it. A cohort acquired four months ago cannot produce a twelve month number, and padding it with a projection destroys the comparison.
Subtract refunds and cancelled orders from revenue before you divide by customer count.
Apply gross margin last. If margin varies sharply between product categories, compute margin CLV per category instead of one blended rate, because a blended rate hides the categories that are quietly losing money.
Put the cohorts side by side. The movement from one to the next tells you more than any single value, because it shows whether the customers you are buying now are worth more than the ones you bought last quarter.

Expect the first honest figure to land below whatever number had been circulating internally. That is the method working, not failing. A smaller number you can defend line by line survives the moment somebody asks you to justify raising the acquisition budget, and a large one nobody can reconstruct does not.