Channel Profitability Analyzer

What this tool does

CM8-368 compares acquisition channels on the profit their customers actually returned over a fixed window. It takes what each channel cost, what its customers went on to spend, and the margin on that spending, and works out which channels earn more than their share of the budget.

Everything runs inside this single file — no account, no upload, no network request of any kind.

Which question this answers

Four questions sit close together and get confused constantly:

• What does a customer cost? — an acquisition cost calculation.
• What is a customer worth? — a lifetime value calculation.
• How long is my money tied up? — a payback calculation.
• Which channels should get the budget? — this tool.

The distinction matters because the answers frequently disagree. A channel can have the lowest cost per customer and the worst profitability, if the customers it brings are worth less. In the sample, the channel with the best return per pound and the channel with the best customers are different channels.

The window, and why it must be the same

Every row must cover the same number of months from acquisition. A channel measured over eighteen months will beat one measured over six, every time, regardless of quality — the customers have simply had longer to spend.

Each row records the window behind it and the tool flags any that disagree with the setting, because the mismatch is invisible in the result and fatal to the comparison. Twelve months is the usual choice: long enough to capture repeat business, short enough that the data actually exists.

The tool also checks the window against the cohort date — you cannot have twelve months of data on a cohort acquired four months ago, and a surprising number of channel comparisons quietly do.

Cohort revenue, not period revenue

This is the discipline that makes the analysis worth doing. Cohort revenue is what those specific customers spent within the window — including everything they bought again. It is not revenue attributed to the channel in a period, which mixes new customers with old ones and tells you nothing about acquisition.

Getting this figure is the hard part and it is where most channel analysis stops. If your systems cannot follow a cohort, model it from your retention curve and mark the row as modelled; the tool records which rows are measured, modelled or estimated, and the reader deserves to know.

The calculation

Refunds are separated because channels differ enormously and it is rarely compared: a channel selling on a discount to people who were not sure often refunds at three times the rate of one selling to people who came looking. In the sample, paid social refunds at 9.2% against 1.2% for referral.

Return on spend

Gross profit returned for each unit spent. Below 1.0 the channel loses money over the window: it returned less gross profit than it cost to run.

Note it is gross profit per unit spent, not revenue. A return on ad spend calculated on revenue at a 30% margin needs to exceed 3.3 just to break even, and quoting it without the margin is how a loss-making channel gets celebrated.

The target should come from your own cost structure: what multiple of acquisition spend does gross profit have to reach to cover overheads and leave a profit? Three is a common answer and not a universal one.

Earning your share of the budget

Above 1.0 means a channel produces more of the profit than it consumes of the budget. Below 1.0 means the reverse. This is a cleaner reallocation signal than return on spend alone, because it is relative to what the rest of your portfolio achieves rather than to an abstract target.

A portfolio in balance has every channel near 1.0. Wide dispersion means money is in the wrong place — which is good news, because it is the cheapest improvement available.

Return against contribution per customer

These two charts frequently disagree, and both are right.

Enterprise outbound in the sample has a poor return on spend and by far the best contribution per customer. Cutting it because the return is low would remove the most valuable customers in the business. The correct reading is that it is expensive and worth it, and the question is whether it can be made cheaper rather than whether to run it.

Headroom is the whole point

A channel with a wonderful return and no headroom cannot absorb more money. Referral is the classic case: extraordinary returns, and no way to buy more of it. Reallocating budget into it achieves precisely nothing.

The tile for the best channel, and the reallocation table, both consider only channels with headroom. Where none has any, the table says so, and the honest conclusion is that money taken out of a weak channel should stay out rather than be moved.

Reading the reallocation table

It takes the weakest channel and the best one with headroom, and shows what moving several fractions of budget would be worth if returns held.

They will not hold. Every channel has diminishing returns: the next pound buys a slightly worse customer than the last, and doubling spend on a channel typically raises its cost per customer by a fifth or more. The table is an upper bound and says so in its own footer.

Use it to size the opportunity, then move in steps and re-measure. Moving a fifth of a budget and checking is a good decision; moving all of it on this arithmetic is not.

Attribution, and saying so

Every figure here rests on deciding which channel won each customer, and that decision is a judgement. The customer who saw an advert, read two articles and then asked a friend is claimed by three channels and belongs to one of them in this register.

The tool cannot resolve that. What it does is make you record how confident you are in each row — measured, modelled or estimated — and show it in the table, so a channel whose profitability rests on a guess is visible as such. The sample marks trade shows as estimated for exactly this reason.

What this cannot do

• It cannot do attribution, and no tool can do it correctly.
• It cannot see channels assisting each other. A channel that generates no direct customers but makes every other channel work better will look worthless here.
• It stops at the window. A channel whose customers stay for years is undervalued by a twelve-month window, and that is the price of a comparison that is possible at all.
• It assumes constant returns to scale, which is why the reallocation table carries a warning rather than a recommendation.
• It has no view about strategic value, brand, or what happens to a market position if you stop.

Printing and sharing

The Report tab prints the tiles, charts and both tables with a title block you fill in. The profit and loss table is the one for a budget meeting: it shows every step from revenue to contribution, so the argument stays about the inputs.

Saving your work

Channels are held in this browser, on this computer, and stay there between visits. Use the backup button to write a JSON file you control.

Accuracy & disclaimer

Every figure here is one you entered, and the most important of them — which customers belong to which channel — is a judgement rather than a measurement. The tool does the arithmetic and reports how confident you said you were; it cannot make an attribution correct.