The (un)Common Logic Approach to Lifetime Value

Every team says they care about Lifetime Value. Yet most dashboards show a single average number, presented with the same confidence as a bank balance, then used to set media budgets or justify discount-heavy promotions. Six months later the finance team wonders why cash is tight and the media team wonders why the bidding model stalled. The problem is not that Lifetime Value is wrong. It is that LTV, when treated as a static, universal constant, ignores the mechanics that actually create value.

The (un)Common Logic approach treats LTV as a working model that earns trust by predicting what really happens to customers, cash, and margin. It is not a complicated formula designed to intimidate. It is a set of practical decisions about data, segmentation, and how to use uncertainty when you spend money. Done well, it lets you make faster bets with more courage and fewer regrets.

LTV earns its keep when it changes a decision you would make today. If a forecast of customer value nudges your paid search bids, pushes a sales rep to call a customer tomorrow instead of next week, or convinces the product team to shorten a trial, then the number did its job. If it sits on a slide for board meetings, politely ignored, it failed.

This means an LTV model should:

Notice what is missing. You do not need the perfect estimate for the next ten years. You need a credible forecast for the next four to eight quarters, with enough segmentation to act on. You also need the discipline to revisit the model as reality unfolds.

Behind every LTV is a simple cash engine. The engine has inputs and leaks. When you diagram it, people stop arguing about definitions and start fixing the right problems.

The essential map looks like this. A customer signs up, perhaps with a coupon. They generate orders or invoices over time, some at full price, some discounted. Every order has variable costs: cost of goods, payment processing, shipping, handling, and sometimes the cost to serve, like support minutes or success hours. Some orders return or refund. Some customers churn then return. A subset buy add-ons, upgrades, or higher priced bundles. Your business collects cash on these orders at different speeds. Finally, there are step-fixed costs, like the shift you added in the warehouse that does not scale per order.

An LTV you can operate from is the present value of expected gross profit over a sensible time horizon, net of variable servicing costs, with returns, discounts, and credit losses accounted for, and with time to cash explicit. You can argue rounding on any of those components. You cannot ignore them.

Two ground rules make the map actionable. Use contribution margin, not revenue. And measure value in cohorts, not in the aggregate. Contribution lets you compare offers and channels honestly. Cohorts let you compare June signups to July signups when you changed the landing page.

If you want to know how a garden grows, you track patches, not the whole yard. Same with LTV. Cohort LTV shows how a group of customers who started under the same conditions behave over time. It picks up the effects of price tests, copy changes, a new app version, or a change in your returns policy.

A direct to consumer apparel brand I worked with had a clean looking average LTV. It masked a mess. New customers acquired on branded search repurchased at twice the rate of social prospecting customers, which is not surprising. The surprise was that social customers acquired in October with a 30 percent discount kept buying through spring, but November social customers with a 40 percent discount ghosted in January. A single average LTV would have kept money flowing into the wrong month with the wrong promo. Cohorts, built by acquisition month and promotion type, made the seasonal trap obvious.

B2B has the same pattern, only slower. A SaaS company selling to mid market accounts had a rising average retention rate, which looked like progress. Cohorts by sales rep showed two different businesses. One rep closed quickly with a discount and had first year churn at 25 percent. Another rep sold slower, no discount, and his accounts would renew even after an outage. The LTV for those two streams was not just different, it demanded different lead routing, coaching, and quotas. Cohorts surfaced the decision.

Most LTV errors start with optimistic retention curves. It is tempting to fit a straight line through a couple of months of repeats and assume the slope holds. Two uncomfortable truths help avoid trouble.

First, retention is not a single number. Define the retention curve you care about. For subscriptions, it is the share of customers active at each period. For reorder businesses, it is the probability of another purchase by time since last order, not months since signup. For B2B, it might be logo retention and dollar retention separately.

Second, your data is censored. You have customers who have not had time to churn, because they joined recently. Survival analysis exists to handle exactly this. You do not need to publish an academic paper. You do need to avoid pretending that six months of data can guarantee what happens at month twelve. A simple Kaplan Meier style approach, even in a spreadsheet, keeps you honest by showing how much of the tail is an assumption.

When you expose the assumed part of the curve, executives tend to ask better questions. What is driving the visible part of retention now, and which levers exist to change it? Would we rather invest in acquisition or shorten time to second value? With a clear view, product and marketing begin to solve the same problem.

Treat contribution margin like oxygen, not a footnote. Count everything that scales with orders or customers. For ecommerce, that means cost of goods, pick and pack, packaging, outbound shipping, payment processing, customer service contacts, refunds and chargebacks, and the cost of free returns. For SaaS, include hosting costs that scale with use, customer support volume, implementation hours, and third party pass through fees.

Two recurring mistakes are worth calling out. First, ignoring the cost to serve heavy users. If your LTV model rewards users who open support tickets ten times a month, you are paying to acquire someone who will cost you more than they pay back. Second, hiding discount depth. A BOGO can double AOV and make early cohorts look like heroes, while clawing back all the margin you thought you gained. Fold the discount into your unit economics, not as a marketing line item later.

I like to plot LTV in contribution dollars, not revenue, for each cohort at 30, 60, 90 days, then quarterly. When a line flattens too early, you look for margin leaks or a stale product moment. When a line is healthy but slow, you ask if faster onboarding or a better reorder nudge reduces time to value.

The board cares about LTV to CAC because it implies a return on investment. The bank cares about cash timing. Your model should bridge them. Two companies with the same LTV and CAC can have completely different cash stress. If Company A collects cash at checkout and Company B invoices net 45, A can recycle marketing dollars faster. If both fund acquisition off the same credit line, A can scale into higher bids and seasonal spikes that B simply cannot afford.

Build a cash ladder for each cohort. When do you pay for traffic or sales commissions. When does the first purchase settle, net of chargebacks. When do repeat orders land, especially for subscription trials that ship before the first full price cycle. You do not need minute level precision. You do need a clear view of payback windows and a policy for how aggressive you are willing to be. I have seen teams turn a 9 month LTV payback into a 5 month cash payback by pulling forward first repeat with a smart email series and making cancellations self serve but with a grace offer. The LTV did not change, the timing did.

There are many ways to forecast LTV. The right choice depends on data volume, purchase cadence, and how often you plan to pull the lever.

Heuristics work when you are small or move fast. If you have a single product and most value arrives in 90 days, a simple rule like cumulative 90 day contribution times a modest multiplier can guide bids. You will leave some precision on the table, but you will avoid false certainty.

Deterministic cohort models suit mid stage teams. Build a spreadsheet or a simple warehouse model with cohort rows and period columns, fill in observed rates for the early periods, and apply conservative tails based on older cohorts. You can segment by channel, offer, and device without overfitting.

Probabilistic or Bayesian models help when you have high customer heterogeneity and long tails. A buy till you die model with hierarchical priors can borrow strength across segments and express uncertainty directly in your bidding policy. This is powerful once you have the plumbing, but overkill if your biggest problem is data hygiene.

Pick the simplest model that still explains the differences you see in the real world. If February Facebook customers behave differently from May Facebook customers, you should not treat them the same because your model is elegant.

An LTV model that never changes a bid or a sales script is a cost center. Tie it to operating levers. If you run paid media, use predicted 90 day contribution at the ad set level to set target CPA or ROAS, and refresh weekly as cohorts mature. If you run a sales org, use predicted first year contribution with an uncertainty band to set discount authority. If you run product, aim experiments at shortening time to second value, because everything downstream improves when that interval shrinks.

Before you trust a new LTV model, run a short checklist to keep your feet on the ground.

The discipline of this checklist prevents most expensive misinterpretations. It also builds credibility with finance, which is the team that will save you when macro conditions change.

Every board deck eventually shows an LTV to CAC ratio. The ratio gets waved around as a health indicator. It can be, but only if you pin down the terms. What time horizon is the LTV measured on. What costs are in CAC. Are you measuring blended CAC or new customer CAC by channel. What discount rate did you use. If you cannot answer those questions in one sentence each, the ratio is theater.

I encourage teams to express a policy instead of a single ratio. For example: we target a payback of under 5 months on cohorts with predicted 12 month LTV:CAC of at least 3, and will accept 6 months for audiences with lower uncertainty bands and strong upsell rates. That policy guides bids, calendar decisions, and headcount ramps much better than a single number.

A subscription skincare brand followed a hard 3 to 1 rule and strangled growth for a quarter. When we rebuilt LTV and showed that email triggered add ons at month 3 drove a third of total value, they adjusted the policy to accept 2.3 to 1 on channels with reliable onboarding into add ons. Growth recovered without a collapse in contribution.

Discounts attract, but they also sort. A steep discount can change the mix of customers who convert and the way they behave later. Early in a paid social program for a household goods client, we watched October and November cohorts with identical sizes diverge in December. The only difference was that one group saw 20 percent off and the other saw 35 percent off. The deeper discount grew faster in week one, https://privatebin.net/?b4c3a964c88fea57#4r2y1S7Evr5592ypGyVB9bhoEW44kTEBDJ298Py785hB then flatlined. When we stitched the margin math to the retention curve, the 20 percent cohort delivered 30 percent more contribution by day 90, even with lower top line.

Do not fold promotional strategy into a single knob called CPA. Segment LTV by offer. Track first repeat with and without promo. Be honest about cannibalization, especially around events like Black Friday when your organic buyers would have purchased anyway.

Most LTV problems are not statistical. They are accounting or plumbing.

Reacquisition double counting. If a churned customer returns and your ad platform claims credit, decide whether that is reacquisition or retention. Your LTV should not include reacquisition spend as free.

Gift cards and store credit. If you count gift card redemption as revenue in month one, you will inflate early contribution then starve later periods.

Prepaid or annual plans. These pull cash forward and mask churn risk. Make sure your model accounts for revenue recognition and the risk of non renewal at the right intervals.

Fraud and promo abuse. High first order AOV with different shipping names and the same IP is not a windfall. Exclude known fraud from cohorts or you will train your model to love it.

Channel blend shifts. If branded search grows faster because of your TV ad, your LTV by channel view needs to reflect the real driver. Otherwise you overfund the cheap looking channel and ignore the engine that made it cheap.

Spotting these traps early is part of the (un)Common Logic approach. You do not need perfect truth. You do need to be consistently less wrong in the same direction.

Pricing changes LTV in two ways. It changes margin per order, and it changes behavior. A price increase that lifts margin but lengthens time to second purchase can leave you net worse off. Conversely, a small price decrease that increases attach rates for profitable add ons can raise long term contribution. Treat pricing experiments as LTV experiments, not just AOV moves.

Packaging matters too. A meal kit brand discovered that three recipes per week produced higher 6 month contribution than two or four. Two was too small to build a habit, four created fatigue and cancellations. The difference showed up in week eight churn, not week two. A short A/B test would have missed it. Cohort LTV picked it up and let the team scale the right SKU mix.

A reliable LTV forecast admits error and improves with feedback. Three habits help.

First, compare predicted and realized LTV by cohort at regular intervals. A simple plot with prediction intervals and realized points forces conversations about model drift, seasonality, and operational changes. If your predicted 180 day contribution runs 15 percent high for two cohorts in a row, treat it as a fire, not a footnote.

Second, write down the current set of assumptions. Discount rate, cost to serve, estimated tail behavior beyond observed data, and how you treat reacquisition should be explicit. When the CFO challenges a spend ramp, you can show what would have to be true for the bet to be wrong.

Third, keep the model close to the operators. If the only people who can change LTV assumptions sit in analytics, everyone else will work around the model instead of with it. The best teams I have worked with give marketing, product, and finance seats at the LTV table. They own different levers, but the same view of value.

You do not need a complex MLOps stack to run a disciplined LTV program. You need a cadence and a place where the truth lives.

Weekly: refresh cohorts, update 30, 60, 90 day contribution, and push predicted 90 day contribution to ad buying or lead scoring. Review exceptions, such as a cohort underperforming its prior three neighbors by more than a set tolerance.

Monthly: revisit tails, variance by channel and offer, and time to cash. Adjust payback policy if macro or logistics change. Bring finance into the room and align on credit line and headcount implications.

Quarterly: audit assumptions, validate predicted vs realized for the oldest cohorts, and decide on two to three large experiments aimed at shifting LTV, not just conversion rate. Examples include a change to the onboarding sequence, a returns policy adjustment, or a new bundle.

This cadence keeps LTV from drifting into academic analysis. It also creates a culture where people expect the number to move when they take action, which is the entire point.

Perfection is seductive. You can always add another parameter, another segment, another lag structure. Resist it when the action would not change. If a channel shows predicted 90 day contribution of 42 dollars, plus or minus 3, and your payback policy supports bids up to a 40 dollar CPA, the decision is to spend. Do not delay to shave a dollar off uncertainty if it means missing a seasonal window or an inventory position.

On the other hand, invest in precision when small errors swing the decision. If your margin is razor thin or your cash runway short, make the model sweat. Include processing fees accurately, split shipping zones, and get returns timing right. In one case, simply moving from assumed 4 percent refunds to the observed 7 to 10 percent range, by cohort, prevented a seven figure overspend.

The test is always the same. Will better precision change a near term decision you can actually implement. If not, ship the simpler model and revisit.

The phrase (un)Common Logic captures a posture more than a formula. Treat LTV as a living instrument, not a vanity metric. Make the cash map explicit. Show your work on retention. Use contribution dollars. Segment where it matters. Prefer a clear, falsifiable policy to a single triumphant ratio. And tie the output to the levers you control today.

An apparel team used this approach to pull out of a slump. They had a dashboard that said LTV was fine and a bank balance that said otherwise. We rebuilt cohorts by channel and promo, measured contribution not revenue, and drew a cash ladder. Two truths fell out. November social customers on steep discount never paid back, and free returns were masking margin erosion that accelerated at scale. They changed offer strategy for Q1, limited free returns to VIPs, and moved onboarding emails earlier. The next two cohorts recovered to a payback under four months, with a predictable path to profit by day 120. Not a miracle, just better logic applied to the way the business really works.

That is the heart of an effective LTV practice. It respects uncertainty without being paralyzed by it. It meets finance where they live, inside timing and margin. It gives marketing and product room to experiment with purpose. Most of all, it keeps the company honest about which customers create value, how quickly, and at what cost. When a team runs on that kind of logic, the average LTV on a slide finally matches the money in the bank.