Bid on Lifetime Value, Not Cost Per Click

The keyword that lies to you

Here is a thing that took me embarrassingly long to properly understand. Two keywords in your Google Ads account can have the same click through rate, the same conversion rate, and the same average order value, and still be worth wildly different amounts of money to your business. Not a bit different. Sometimes twice as much. And if you bid the same on both, which every dashboard quietly encourages you to do, you are overpaying for one and starving the other of budget at the exact moment it is making you the most money.

I want to walk you through why, with real numbers you can plug into a spreadsheet this afternoon, because once you see it you cannot unsee it, and it changes how you spend every euro on paid search. The short version is this: bid to the value of the customer, not to the cost of the click. The click is what you buy. The customer is what you are actually paying for. Those are not the same thing, and the gap between them is where the money is.

Here is the mental shift in one sentence. Your keyword report ranks clicks, and clicks are a cost. What you want ranked is customers, and customers are an asset. The whole point of this piece is to move you from the first list to the second, because they do not sort in the same order, and the difference between them is quietly deciding whether your ad budget makes you money or loses it.

Two keywords, one dashboard, two different truths

Let me make this concrete with a shop I know well in shape if not in detail: a business selling espresso machines and everything around them, beans, grinders, servicing, the lot. Two keywords bring in customers.

The first is the broad one, "espresso machine". Loads of traffic. People click it a lot, so its click through rate looks great in the account, and a healthy share of them buy. It is the keyword the dashboard loves. Green everywhere.

The second is "commercial espresso machine". Less traffic. Slightly lower click through rate, because it is narrower and some of the searchers are just doing research. On the surface it looks like the weaker keyword, the one you might trim when you are tightening the budget.

Now here is the part the dashboard does not show you. The person searching "espresso machine" is very often a hobbyist buying one machine for their kitchen. Lovely customer, buys some beans, maybe comes back once. The person searching "commercial espresso machine" is frequently a cafe owner, a restaurant, an office manager. They buy the machine, then they buy beans every single month, they take a servicing contract, they add a second location. Same shop, same checkout, completely different lifetime value.

If you judge these two keywords by what happens in the first session, they look almost identical. If you judge them by what the customer is worth over the next three years, they are not remotely the same, and you should not be bidding the same on them.

Notice what actually differs here. It is not the ad. It is not the landing page. It is not even the search intent in the narrow sense, both people want to buy a machine. What differs is the shape of the relationship that starts after the sale. One is a single transaction with a bit of a tail. The other is the front door to a recurring account. Your ad platform sees the front door and prices it as if it were the whole house.

What a customer is actually worth

To bid to value you need a number for value, so let us put one down. The cleanest simple model of customer lifetime value, the one I keep coming back to, comes out of the marketing analytics literature, in particular the framework in Cutting Edge Marketing Analytics by Venkatesan, Farris and Wilcox. It says the lifetime value of a customer is their margin per period, scaled up by how long they tend to stick around and discounted for the fact that money in the future is worth less than money today.

Written out it looks like this.

$CLV = M \times \frac{r}{1 + d - r}$

Where $M$ is the gross margin a customer produces in one period, usually a year. $r$ is the retention rate, the probability they are still a customer next period. And $d$ is your discount rate, the annual cost of capital that turns future euros into today's euros.

The fraction is the clever bit. When retention $r$ is high, the denominator $1 + d - r$ gets small, and the whole multiplier gets large, because a customer who keeps coming back is worth a multiple of a single year. When retention is low, the multiplier collapses towards nothing, because there is no future to speak of.

Why that fraction blows up, in one line of maths

If the formula feels like it appeared out of nowhere, here is where it comes from, and it is worth thirty seconds because it explains the explosion. A retained customer gives you margin $M$ this year, then $M$ again next year but only if they survive, which happens with probability $r$, and that future euro is worth less by the discount factor $1 + d$. Keep going forever and you are adding up a geometric series.

$CLV = M \sum_{k=1}^{\infty} \left( \frac{r}{1 + d} \right)^{k} = M \times \frac{r}{1 + d - r}$

A geometric series with a ratio close to one adds up to a very large number. That is the whole story. Retention $r$ sits in the ratio, so as retention creeps towards one, the sum runs away from you in the good direction. A customer is not one sale. A customer is a sale that keeps rolling, and the roll compounds. The formula is just bookkeeping for that compounding.

Running both customers through the formula

Let me run both of our customers through it. The discount rate $d$ is 10 percent, 0.10, for both, because it is a property of the business, not of the customer.

The hobbyist behind "espresso machine": say they throw off 400 euros of gross margin a year in beans and bits, and about 70 percent of them are still buying next year, so $r$ is 0.70.

$CLV = 400 \times \frac{0.70}{1 + 0.10 - 0.70} = 400 \times \frac{0.70}{0.40} = 400 \times 1.75 = 700$

So a hobbyist customer is worth about 700 euros over their life with the shop.

The cafe behind "commercial espresso machine": higher margin, say 450 euros a year because they buy more beans and take servicing, and stickier, because switching supplier is a hassle for a working kitchen, so retention is 0.80.

$CLV = 450 \times \frac{0.80}{1 + 0.10 - 0.80} = 450 \times \frac{0.80}{0.30} = 450 \times 2.667 = 1200$

That customer is worth about 1200 euros. Notice what did the heavy lifting there. The annual margins are close, 400 against 450. The gap in lifetime value, 700 against 1200, comes almost entirely from retention, from that 0.70 against 0.80, because of how the fraction blows up as $r$ approaches one. Small differences in loyalty become large differences in worth. That is the first thing worth sitting with.

A third keyword, because the real world has more than two

Two segments make a clean story, but your account has more than two, and the third one teaches the sharpest lesson. Add the bargain hunter, the person searching "espresso machine offer" or "best value espresso machine". These are lovely people, I am one on most product categories, but as customers they are thin. They buy on price, the margin is slim, and they are the first to leave when a competitor runs a promotion. Say the annual margin is 250 euros and retention is only 0.55.

$CLV = 250 \times \frac{0.55}{1 + 0.10 - 0.55} = 250 \times \frac{0.55}{0.55} = 250 \times 1.0 = 250$

Here is the twist that makes people sit up. The bargain hunter often converts best of the three. They have their card out, they know what they want, they are comparing on price and you have a good one, so the conversion rate might be 4.5 percent against 3.0 for the others. By the logic of the dashboard, this is your star keyword. Highest conversion. And it produces the least valuable customer of the lot. Conversion rate measured how ready they were to click buy. It said nothing about whether they would ever come back, and coming back is where the money lives.

From lifetime value to the right bid

Now we can work out what a click is worth, and this is where it pays off.

You are not going to spend all 700 or all 1200 euros acquiring a customer, or you would make nothing. You decide up front what share of lifetime value you are willing to give away to win the customer. Say you want to keep the bulk as profit and reinvest a third of lifetime value into acquisition. That gives you an allowable cost per acquired customer.

For the hobbyist: 30 percent of 700 is 210 euros. That is the most you can pay to land one and still hit your margin. For the cafe: 30 percent of 1200 is 360 euros. For the bargain hunter: 30 percent of 250 is 75 euros.

But you do not buy customers, you buy clicks, and only some clicks turn into customers. If the conversion rate is 3 percent, it takes about 33 clicks to make one customer. So the most you can pay per click is the conversion rate times the allowable cost per customer.

$\text{max cost per click} = \text{conversion rate} \times f \times CLV$

Here $f$ is that fraction of lifetime value you are willing to spend, 0.30 in our example. Run the three.

Hobbyist keyword: 0.03 times 210 is 6.30 euros per click. Cafe keyword: 0.03 times 360 is 10.80 euros per click. Bargain keyword: 0.045 times 75 is 3.38 euros per click.

Same shop. The cafe keyword deserves a max bid over 70 percent higher than the hobbyist one, and more than three times the bargain keyword, even though the bargain keyword had the best conversion rate of all three. If you had been bidding one flat number across the account, which is the default behaviour, you were capping the cafe keyword so low that you lost those auctions to competitors who did the maths, and you were bidding that same number on the bargain keyword and torching margin on every discount hunter you won.

The whole account on one line each

Here is the whole thing on one line each, the way I would put it in front of an owner.

Keyword group	Click through rate	Conversion rate	Customer lifetime value	Correct max cost per click
espresso machine	5.0 percent	3.0 percent	700 euros	6.30 euros
commercial espresso machine	4.2 percent	3.0 percent	1200 euros	10.80 euros
espresso machine offer	6.1 percent	4.5 percent	250 euros	3.38 euros

Read that table down each column and the story flips depending on which column you trust. By click through rate, the bargain keyword wins, it is the greenest thing in the account. By conversion rate, the bargain keyword still wins. By customer lifetime value, it comes dead last, and the quiet commercial keyword, the one you were about to trim, is worth almost five times as much per customer. Your best keyword by the metrics everyone screenshots is your worst keyword by the only metric that pays your wages. If that does not make you want to re export your search terms report tonight, I do not know what will.

How sensitive is all of this to retention

Because retention did most of the work above, it is worth seeing exactly how much. Hold the annual margin at 450 euros and the discount rate at 0.10, and just move retention. Watch what happens to the value of the customer.

Retention rate	Multiplier	Customer lifetime value
0.50	0.83	375 euros
0.60	1.20	540 euros
0.70	1.75	788 euros
0.80	2.67	1200 euros
0.90	4.50	2025 euros

Ten points of retention between 0.80 and 0.90 nearly doubles the customer again, from 1200 to 2025. The curve is not a straight line, it bends upward, steeply, exactly where good businesses live. This is the single most useful table in the piece, because it tells you that the highest leverage move on your ad bidding is often not in the ad account at all. It is anything that nudges retention up a few points, because every point feeds back through the formula into what you can afford to bid to win that customer in the first place.

The whole chain in one picture

The mistake is optimising the early links in this chain and ignoring the last one. Here is the chain.

Most accounts stop thinking at conversion, box C. They pick keywords, watch clicks, count conversions, and bid to whichever conversion looks least expensive. The value that actually pays your wages lives two boxes further along, at E, and it loops all the way back to what you should have bid in the first place. Bidding to conversion treats every converted customer as identical. They are not. The loop from lifetime value back to the bid is the whole game.

The two things that surprise people

The first surprise is the one we just did. The keyword that looks best in the dashboard can be the one making you the least money, because click through rate and conversion rate both describe the click, and neither of them describes the customer. A shop that ranks its keywords by click through rate is ranking them by how good they look, not by how much they pay.

The second surprise is how much of lifetime value comes from retention rather than from the first order. Look back at the sensitivity table. The margin per year barely moved between our two headline customers, 400 to 450. It was the retention rate, 0.70 to 0.80, that nearly doubled the value, because of the way the fraction compounds a loyal customer's future. Which means the highest leverage thing you can do to your bidding is often not in the ad account at all. It is anything that lifts retention, a subscription for the beans, a servicing reminder, a reason to come back, because every point of retention feeds straight back through the formula into how much you can afford to bid to win that customer in the first place. Marketing and retention are the same conversation. Most businesses run them in different rooms.

A worked mini case, one month in one account

Let me put the whole idea through a single month, because the abstract version convinces the head and the worked version convinces the gut. Same shop, same three keywords, a budget of roughly 2500 euros for the month. First, the world you are probably in now: one flat maximum bid of 6.50 euros across everything, because that is the number that felt about right.

Segment	Flat bid	Clicks won	Spend	Customers	Value harvested
Bargain	6.50 euros	200	1300 euros	9.0	2250 euros
Hobbyist	6.50 euros	120	780 euros	3.6	2520 euros
Cafe	6.50 euros	60	390 euros	1.8	2160 euros
Total		380	2470 euros	14.4	6930 euros

Look at what the flat bid did. It overpaid for the bargain segment, where 6.50 is nearly double the 3.38 those customers justify, and it won a heap of them because the bid was generous. And it underpaid for the cafe segment, where 6.50 is well below the 10.80 those customers justify, so you kept losing those auctions and only scraped 60 clicks. You spent the most money buying the least valuable customers. Value harvested per euro spent is 6930 divided by 2470, about 2.81.

Now bid to value. Drop the bargain keyword to 3.38, lift the cafe keyword to 10.80, leave the hobbyist near where it was. Same appetite for spend, redistributed.

Segment	Value bid	Clicks won	Spend	Customers	Value harvested
Bargain	3.38 euros	90	304 euros	4.1	1013 euros
Hobbyist	6.30 euros	120	756 euros	3.6	2520 euros
Cafe	10.80 euros	150	1620 euros	4.5	5400 euros
Total		360	2680 euros	12.2	8933 euros

You won fewer customers overall, 12.2 against 14.4, and a marketer paid on volume would call that a worse month. It is a much better month. Value harvested climbed from 6930 to 8933, up about 29 percent, on roughly the same spend. Value per euro went from 2.81 to 3.33. You bought fewer customers and far more money, because you stopped paying a premium for the ones who never come back and started paying up for the ones who stay for years. That is the entire argument in one table swap.

The number your finance person will ask for

Sooner or later someone in the business asks for a target return on ad spend, a ROAS number they can hold you to. Value based bidding gives you one cleanly, and it falls straight out of that fraction $f$ we chose. If you are willing to spend a fraction $f$ of lifetime value to acquire a customer, and you feed the platform the lifetime value as the conversion value, then your target return on ad spend is simply one over $f$.

$\text{target ROAS} = \frac{CLV}{f \times CLV} = \frac{1}{f}$

With $f$ at 0.30, the target ROAS is 3.33. That single number, handed to Google or Meta alongside a real conversion value per segment, reproduces every per segment bid we calculated by hand, automatically, in every auction, all day. You do not set the bids. You set the value and the target, and the machine sets the bids for you, correctly, which is exactly what the machine is good at once you stop lying to it about what a customer is worth.

Why this is a moat and not just a tactic

Here is the part I like best, and it is the reason I bother writing any of this down. Value based bidding is not a trick that stops working when everyone copies it, because most of your competitors physically cannot copy it. They are running rented shops that forget where a customer came from, they are sending bare conversion counts to the platform, and they are judging keywords on cost per acquisition in a weekly screenshot. That means the cafe keyword, the one worth 10.80 a click, is sitting in an auction full of people bidding 6.50 because that is what the dashboard told them felt safe. You get to walk in and pay 9 euros for a click worth 10.80 and win it every time, and it still makes you money, and they never understand why they keep losing the customers they most wanted. Meanwhile you quietly hand them the bargain keyword you did not want anyway. The knowledge asymmetry is the moat. It lasts exactly as long as it takes them to connect their orders back to their keywords, which for a rented platform is never. You are not bidding against their budgets. You are bidding against their bookkeeping, and yours is better.

The common mistakes I see

Most accounts that come to me are not broken in exotic ways. They make the same handful of mistakes, and they are all versions of confusing the click with the customer.

The first is sending a bare conversion count to the ad platform. A conversion fires, the pixel logs a one, and every customer looks identical to the algorithm. You have thrown away the single most important fact, which customer this was, before the bidding even starts. Send a value, not a flag.

The second is ranking and cutting keywords on cost per acquisition alone. A low cost per acquisition on a low value customer is not a win, it is a small return on a small asset. The bargain keyword above will always show the best cost per acquisition and the worst contribution to the business. Judge keywords on value harvested minus spend, never on cost per acquisition in isolation.

The third is optimising the ad and ignoring the aftercare. People pour weeks into ad copy and landing pages, which move conversion rate a point or two, and never touch the subscription flow or the reorder reminder, which move retention, which as the sensitivity table showed moves the customer's value far more. You are polishing box B and C while the leverage sits in box E.

The fourth is renting a platform that forgets. If your shop cannot tell you which keyword a customer came from two years after they arrived, you physically cannot bid to lifetime value, because you can never connect the value back to the source. The mistake is architectural, and it is the one that quietly caps everything else.

How to apply this on Monday morning

You do not need a data science team and you do not need to boil the ocean. Here is the order I would do it in.

Start by pulling twelve to twenty four months of order history and tagging each customer with the keyword or campaign that first brought them in. If you cannot do this, that is your real first project, and it is exactly why I keep pushing owners towards shops and systems they control rather than rented black boxes, because you cannot bid to lifetime value if your platform forgets where the customer came from the moment they log out. It is the same argument I make about owning your ecommerce platform rather than renting one.

Then group customers into two or three segments, the hobbyist and the cafe and the bargain hunter, and work out the average margin and the rough retention for each. You are not predicting individuals yet, you are just separating the obvious tiers. Put those into the CLV formula and get a value per segment.

Next, pick your fraction $f$, turn it into a target ROAS of one over $f$, and start feeding real conversion values back to the ad platform per segment, not a bare count. Google and Meta will both bid to a value you send them, but only if you send it. This is the step that turns the automated bidding from working against you into working for you.

Finally, treat retention as a marketing lever, not an afterthought. Every point you add to retention, through a subscription, a reminder, a reason to return, raises what you can afford to bid to win that customer in the first place, and compounds through the fraction. Segment, value, feed back, retain, and review the numbers each month. Precision comes later. The direction is the win.

If you want a hand turning your own search terms and order history into a value based bidding model, that is squarely the kind of thing I do, and it is what my performance marketing work is built around. Bring your data, we will find the keyword that has been lying to you.

The one number to leave with

If you take a single idea from this, make it this one. Stop asking what a click costs. Start asking what the customer behind the click is worth, and let that set the bid. Two keywords with identical clicks can hide a customer worth 700 and a customer worth 1200, and a third that converts best of all and is worth the least, and the only way to see the difference is to look all the way down the chain to the money. Bid to lifetime value, not to cost per click, and you will happily pay more for the clicks everyone else is underbidding, and less for the ones they are all fighting over.