Know what every customer is worth. Then act on it, every day.
Agentic Customer Lifetime Value for every customer: a predicted contribution margin over the next 90, 180 or 365 days, refreshed every day, wired to the campaigns, offers and budgets that use it.
The forecast stops living in a dashboard and starts changing decisions. Two identical customers acquired on the same day can end up with three times different lifetime values, depending on how well your retention worked for each. That gap is where the money is.
Runs in your own cloud. You own the model, the code and the data, and every agent decision is logged.
Customer value, all customers Updated tonight · 02:00
Top 20% of customers79% of revenue
Customer #10400CLV · next 12 months
€1,240
Next purchase likely in 2 to 3 weeks
Value sent to Google and Meta
Offer timed to the next purchase
Illustration · example values
In one minute
Customer lifetime value in one minute
What CLV is, how it is calculated for each customer, why the top 20% of customers carry most of the revenue, and what it changes for marketing: budgets, Google and Meta, retention and discounts.
Voice and music are AI-generated. Customer values shown are illustrative. Plays via YouTube. Clicking play loads the video from YouTube (Google), which may set cookies. See our privacy policy.Read the transcript
Customer lifetime value is the margin a customer brings you in the coming months, after the cost of winning and serving them.
We calculate it for every single customer: what they spend per order, when they will likely buy again, and what they cost. So you also know the right time to get in touch.
Sorted by that value, the top twenty percent of customers usually bring seventy-five to eighty-five percent of revenue.
Most companies plan by cohorts. But inside every cohort, customers differ, so we group the high-value customers together.
For marketing, an agent puts it to work every day: budget moves to channels that bring valuable customers, Google and Meta find more customers like your best, retention goes where value is at risk, and discounts only where they pay back.
Know what every customer is worth. Book a free call at prodata.ai.
What CLV is
The definition most CLV conversations skip
Most casual uses of “CLV” are something else. The rest of this page only makes sense once the base concept is solid.
Customer Lifetime Value (CLV, also CLTV or LTV) is a predicted number: the contribution margin a customer will generate for your business over a defined future window, net of the cost of serving and acquiring them. Not revenue from the first purchase. Not revenue over a quarter. Contribution margin, over a horizon you set, as a value in euros.
How CLV is calculated
Margin per order€155
×
Expected orders, next 12 months9
−
Cost to win and serve€155
=
CLV, next 12 months€1,240
Example for one customer. The margin per order comes from past orders and returns; when and how often the customer will buy again is predicted from their recent behaviour; cost is allocated to the individual customer. Calculated for every customer, refreshed daily.
Three things separate real CLV from the imitations
Forward-looking
CLV predicts future value. Historical cohort spend is a useful floor and belongs in board reporting, but on its own it is not CLV.
Profit-based
Contribution margin, net of returns, payment, shipping, and acquisition and service cost allocated per customer. Revenue-based “CLV” overstates the number and leads to overspending on acquisition.
Time-bounded
Over the next 90, 180 or 365 days. A value over an unspecified “lifetime” cannot be tested. A value over a fixed window can be checked against what actually happened.
Six decisions CLV makes better than any other metric
1
Acquisition ceiling
How much you can spend to win a customer, by channel and segment, before you stop making money. Acquisition cost alone cannot answer this.
2
Channel valuation
Which channels bring customers worth winning, not just clicks. A €30 channel bringing €90 customers is weaker than a €50 channel bringing €220 customers.
3
Retention focus
Retention budget goes to customers whose predicted value is at risk, not to the ones who were going to stay anyway.
4
Segmentation
Which customers get premium care, which get rich automation, which get efficient automation only.
5
Finance alignment
A number marketing and finance can share: contribution margin, tested against actuals. Budget talks become investment talks about payback.
6
Early warning
A rising CLV curve on new cohorts is the hardest-to-fake sign the product is winning. A falling one shows up quarters before revenue moves.
Your priorities
What a CMO actually needs from a CLV programme
Everyone agrees CLV matters. Almost nobody uses it day to day. These six concerns come up in every conversation with a CMO.
“Our CLV is stale by the time it lands.”
Computed once a quarter, read once, then forgotten.
How Agentic CLV helps
The forecast refreshes daily, per customer, and feeds the systems that spend money, not just the deck that reports it.
“One CLV for segments with very different economics.”
The average hides the customers who matter most.
How Agentic CLV helps
Forecasts per customer show the five to ten times spread that averages hide, so you can treat segments differently on purpose.
“Acquisition cost keeps rising, and finance asks why.”
The ad platforms bid for the cheapest first purchase.
How Agentic CLV helps
Predicted CLV goes to Google and Meta, so they bid for lifetime profit and find more customers like your best, not just cheap first orders.
“Retention budget goes to customers who were never going to leave.”
Discounts land on loyal customers who would have paid full price.
How Agentic CLV helps
The agent targets customers whose predicted value is at risk, not those who already opened the last email.
“We lose most customers in the first 90 days.”
Most retention programmes start after day 180, long after the drop.
How Agentic CLV helps
Separate plays, budgets and measures for days 0 to 30, days 30 to 180 and beyond day 180.
“We built a CLV model last year. Nothing changed in the P&L.”
A good model on its own changes nothing.
How Agentic CLV helps
The model is about 30 percent of the work. The agent, the plays and the links into ad platforms and CRM are the other 70 percent, and they move the margin.
The problem
Why most CLV programmes stop at the dashboard
The problem is almost never the model. The model is usually right. Everything around it cannot keep up.
It is stale
Computed monthly or quarterly, it describes behaviour that is weeks or months old. Bids, emails and offers are decided in seconds.
It is blended
The company-wide average hides the five to ten times spread across segments that would justify treating them differently.
It is advisory, not operational
Ad platforms, email tools, CRM and personalisation mostly never see CLV. They keep optimising for cheaper proxies: clicks, first purchases, open rates. That gap is one of the largest sources of wasted acquisition spend.
Five ways to work with CLV
Approach
What it is
Why it stalls
Average CLV
One number for the whole customer base
Hides that a minority of customers carries most of the margin
Historical CLV
Each customer's past spend, extended forward
Assumes the future looks like the past, which stops being true once your actions shape what customers do
Rule-based segments
RFM buckets, loyalty tiers
Stable and easy to read, but blind to customers whose behaviour is about to change
Predictive CLV in a dashboard
A model, a chart, a quarterly review
The forecast is right; the decision never happens
Agentic CLVOur approach
A forecast per customer, refreshed daily, with an agent that takes the matching action in the systems that spend money
The forecast becomes a decision, and the decision shows up in contribution margin
Results
What changes when the forecast becomes a decision
Drawn from work in retail, e-commerce, telecom and B2B subscription. Specifics are shared under NDA on a call.
Contribution margin
Budget shifts from treating everyone the same to treating each customer by predicted value. Clients typically see a mid single-digit margin lift within two quarters, without raising marketing spend.
Acquisition quality
With predicted CLV as the conversion value, Google and Meta bid for lifetime profit rather than first-purchase revenue, and find more people who look like your best customers. Done well, this is often the single highest-return move in the programme.
The first 90 days
Most of the value a business ever loses is lost in the first 90 days. Rebuilding that window around the second purchase is where the fastest P&L improvement usually shows up.
Retention cost
Incentives stop going to customers who would have stayed anyway and go to those who were about to leave.
Operating rhythm
Daily agent runs replace quarterly CLV reviews. The CMO sees what moved yesterday, which plays fired and what finance can expect this quarter.
The next use case
Segmentation, churn and next-best-offer reuse the same data foundation, so they ship in weeks, not quarters.
Most companies can report CLV. Few use it. We usually find a CLV number on demand, but rarely wired into the systems that spend money.
The spread is bigger than leadership expects. Five to ten times between the top and bottom predicted-CLV deciles is the normal shape, not the exception.
In B2B, a few accounts carry most of the margin. The concentration varies by industry, but the pattern holds, which is why the CLV lever is often largest in B2B.
Cost allocation decides the outcome. Clients who allocate cost per customer properly tend to see payback within two quarters. Those who skip it get a nice dashboard and flat margin.
Our approach
A model that forecasts. An agent that acts. A loop that learns.
“Agentic AI” gets used loosely. In a CLV context it means three specific capabilities working together.
1
Prediction per customer
For each customer, a predicted contribution margin over a fixed horizon (90, 180 or 365 days, matched to your planning cycle), refreshed daily, with a confidence range finance can defend. A value in euros, not a probability.
2
Action within guardrails
The agent bids, sends, offers, personalises, escalates or holds back, inside caps and rules your team sets. No marketer has to open a dashboard for it to happen.
3
Learning from holdouts
Every play has a holdout group. Every decision is logged with its expected uplift. Each recalibration learns from what actually worked, so the agent gets better instead of drifting.
What goes into the number
Nobody can know how many years a customer will stay. What can be predicted is whether and when a customer will buy in the coming months, and what that will be worth. So CLV is one value per customer for a fixed window ahead, for example the next 6 or 12 months, learned from a rolling window of past behaviour (typically the last 24 months) and refreshed daily as new orders come in. Our method, refined over 12 years of CLV work, builds it from three sides.
The value side
How much a customer spends per order: margin history, order frequency and whether the gaps between orders are shrinking or growing, category preferences, returns, complaints and payment history.
The activity side
Whether and when they will buy again in the window: days since the last order, the usual gap between orders, tenure, and the signals that someone is drifting away, such as a change of address, falling email response, a switch in assortment or, in B2B, product usage tailing off. It also tells you the right time to get in touch.
The cost side
Marketing and service cost allocated to the individual customer: cost per order directly, cost per click to the sessions it caused, newsletter and shop costs to the customers they served. Without this, cheaply acquired customers look better than they are.
Three levels of CLV, for three different jobs
Level
What it is
What it is for
Level 1: Historical CLV
Realised margin per cohort, by acquisition channel, first product and month
Board reporting, unit economics, an auditable baseline. If this is wrong, nothing above it can be trusted
Level 2: Formula-based CLV
Order value × frequency × margin × retention, discounted, by cohort and channel
Annual planning, acquisition cost targets per channel, media mix
Level 3: Predicted CLV per customer
A predicted contribution margin per customer over a fixed horizon, net of allocated cost, refreshed daily
The signal the agent uses to make decisions every day
Most organisations stop at Level 2. Agentic CLV runs on Level 3, with Levels 1 and 2 kept consistent alongside it.
Modules
Customer Analytics, module by module
Customer Analytics is a programme, not a tool. Each module reuses the same data foundation, pipeline and cloud setup, so the second module costs a fraction of the first.
Module 1Available now
Agentic Customer Lifetime Value
Predicted CLV per customer in euros, over a horizon you set (90, 180 or 365 days), net of acquisition and service cost, refreshed daily
Companion signals for the agent: likelihood and expected timing of the next purchase, expected order value, frequency trend, confidence range
Decision agent with configurable plays, thresholds, caps and consent handling
Connections to CRM, marketing automation, Meta and Google conversions APIs and the sales queue
Weekly CLV monitoring by cohort and segment
Finance view: predicted against actual contribution margin, with holdout results
Module 2Coming next
Customer Segmentation
Behaviour-based segments on the same data foundation
CLV, churn risk, price sensitivity and channel response per segment
Segment-aware playbooks, so segments drive decisions instead of reports
Drift monitoring, so segments stay useful quarter after quarter
In operation
How the agent runs, day to day
A fixed daily loop, four plays that do most of the work, a morning view for marketing and a monthly view for finance.
The daily loop
1
Collect
Overnight: transactions, service contacts, campaign responses, web and app events, and attribution results where available.
2
Forecast
CLV, when the next purchase is likely and expected order value, refreshed for every customer.
3
Decide
Each customer is checked against the active plays at the right moment, close to their likely next purchase: at most one play per customer per day, ranked by expected uplift, within frequency caps and consent.
4
Act
Decisions go to the systems that execute them: audiences in Meta and Google, CRM tasks, offers, sales alerts.
5
Report
A morning view for the CMO: what moved, which plays fired, what finance can expect.
6
Learn
Holdout groups measure each play's uplift; the next recalibration uses it.
The four plays that matter most
Play 1
Send predicted CLV to Google and Meta, to win more customers like your best
Most growth teams still bid on acquisition cost or first-purchase return, which overpays for channels that bring cheap, low-value customers. The agent sends each customer's predicted value to Google Ads and Meta. In Google, value-based bidding and the new-customer goal in high-value mode then pay more for prospects who resemble your most valuable customers, across Search, Shopping and Performance Max. In Meta, a customer list with a lifetime-value column builds value-based lookalike audiences, and campaigns optimise for value instead of for any purchase. Acquisition budget goes to people likely to become high-value customers. We start with a well-calibrated segment-level value and move to per-customer predictions once the pipeline is proven.
Play 2
Rebuild retention around the first 90 days
The steepest drop sits between the first and second purchase. Days 0 to 30: a sequence tuned to each customer's signals, aimed at the second purchase, not at more sends. Days 30 to 180: category expansion, loyalty, cross-sell, and early win-back where value is fading. After day 180: VIP service and referrals for the top, efficient automation for the rest.
Play 3
Treat segments differently, on purpose
High predicted value: premium onboarding, human support, early access. Middle: rich automation, targeted upsell. Low: efficient automation with a good experience, but no investment beyond what their predicted margin can cover. Sometimes the honest answer is not to acquire a segment at all.
Play 4
Catch at-risk VIPs before they lapse
The agent watches each customer's own rhythm, not a generic 60-day cutoff. The strongest signal is a flip in order frequency: gaps that were shrinking start to grow. Together with category drift, returns, complaints and falling email response, it shows up weeks before “they stopped buying”. For a high-value customer the agent escalates to sales or starts a tuned reactivation within hours. In B2B the same works for accounts: falling usage, support signals, contact changes.
The CMO, every morning
Yesterday's decisions by play and segment, with expected and actual uplift
Customers newly flagged as at-risk VIPs, escalated in one click
Cohort CLV over the last 90 days, against plan
Which segments gained or lost predicted value, and why
Audiences updated overnight in Meta and Google
Finance, every month
Predicted against actual contribution margin
Payback curves by acquisition cohort
Budget shifts the agent recommends, with assumptions and ranges
Plan against actual, same template every month
People stay in charge
You decide how much the agent decides. Most clients start with the first level and move up play by play, once the results are in.
1
Recommend only
The agent proposes; a person in your team approves each batch before anything goes out. This is how every pilot starts.
2
Approve by exception
Routine plays run on their own within caps. A person approves anything above a threshold: deeper discounts, larger budget shifts, any new play.
3
Run within caps
Fully automatic inside the limits you set, with the morning report and a one-click pause per play, segment, channel or customer.
Always decided by a person
At-risk top customers go to a salesperson, not to an automated message
Budget shifts are approved by finance
Credit limits are reviewed by a person; for consumers, credit scoring is high-risk under the EU AI Act from 2 December 2027, and human oversight is required
New plays, thresholds and caps are set and changed by your team
Guardrails are part of the design
Your team defines the plays, thresholds and frequency caps
Minimum service levels, so no customer is left without a good experience
Caps on discount depth, so loyal customers are not over-discounted
An exploration budget, so the agent keeps learning
Any play can be paused per segment, channel or customer in one click
Every decision is logged with its reason and main drivers
More uses
Where else CLV pays off
Once every customer has a value, many decisions outside the ad platforms can use it too: in marketing, service, sales, purchasing and finance. These are the ones we see most often.
By team
Leadership
A view of future revenue, and decisions based on where value will come from.
Marketing
Budget by customer value, SEA and SEO aimed at valuable customers, campaigns timed to the next purchase.
Key accounts and sales
Accounts prioritised by value, underdeveloped accounts spotted early, upsell and cross-sell where the potential is.
Website and e-commerce
How top customers use the shop, and search and recommendations tuned to raise customer value.
Customer service
Tickets and shipping for top customers first, automated answers where that fits.
Purchasing and marketplace
Customer value by product, category and brand for range and seller decisions.
By decision
Customer service and shipping
Who gets the dedicated line, a faster response or a human instead of a bot, whose tickets and parcels go first, and where service for low-value customers can be automated to cut cost to serve.
Campaign and channel evaluation
Which campaigns and acquisition channels brought customers with a high value, not just many first orders. Together with attribution, this sets where acquisition budget goes.
Occasional buyers to regular ones
Customers who buy once or twice hold large upside. Lifecycle plays timed to their likely next purchase turn part of them into frequent buyers.
Credit limits and payment terms
Invoice purchase, payment terms and credit limits sized by expected value and risk together, so good customers are not held back. Case study: credit limits in fashion retail
Discounts and promotions
Discount depth, coupon eligibility and free-shipping thresholds set by predicted value, so incentives go where they change behaviour.
Returns and fraud
Generous returns for valuable customers, and a closer look at patterns that destroy value, such as serial returns.
Tier thresholds and benefits funded by predicted value, not by past spend alone.
Sales and key accounts
In B2B: where sales capacity, account managers and service levels go next quarter.
Assortment, brands and marketplace
Value by product, category and brand: which ones attract customers who later become high-value, and which sellers to bring onto a marketplace.
Finance and company value
Cohort payback, budgets and forecasts, and customer equity: the value of the customer base as a basis for company valuation, investor reporting and due diligence.
We add these uses once the marketing and sales plays are stable, and only where the legal framing is clear, for example for credit decisions.
Is this for you?
Honest fit criteria
CLV pays off where customers come back and where someone can act on the forecast.
A good fit if
You have a direct customer relationship: B2C, direct-to-consumer, subscription, or B2B with repeat purchases
Roughly 50,000+ identifiable customers, or €10M+ customer-driven revenue
At least two years of transaction history
A CRM or marketing automation platform, and at least one paid channel (Meta or Google) the agent can write to
A CMO, Head of CRM or Head of Growth who wants decisions, not another dashboard
Probably not yet if
You sell one-off purchases with no repeat behaviour
You have less than a year of clean transaction data
There is no channel to act on the forecast: no CRM, no automation, no media buying
You only need one CLV figure for a board slide: a lighter Level 1 analysis is cheaper and faster
Not sure? That is what the free first call is for.
A faster start if
You already run a modern data warehouse (Snowflake, BigQuery, Databricks, Redshift), a CDP or a GA4 export; our Marketing Attribution solution is live; and contribution margin and per-customer cost (cost per order, cost per click, shipping, returns) already sit in your warehouse. That last foundation is the least glamorous and the most often skipped. Fixing it pays off even before the agent arrives.
How we work together
From first call to running agent
A fixed-fee assessment, a pilot on time and material with an estimate up front, then a monthly run fee. You never commit to more than the next step.
1 2 to 3 weeks
Assessment
What happens
We check your data, build a first segmentation view and a historical CLV baseline by channel, and list the plays worth automating first.
What you get
Data check and first segmentation view
Level 1 CLV baseline by acquisition channel
A short list of plays, ranked by expected value
Your decision
Is a pilot worth it?
2 8 to 12 weeks
Pilot
What happens
We build the model on your data, stand up the agent and wire it to two plays, usually predicted CLV to Meta and Google and retention of high-value customers in the first 90 days. Live for one segment or business unit, with holdouts from day one. The agent starts in recommend-only mode: your team approves what goes out.
What you get
A working model and agent on your data
Two live plays with holdout tests
Results in contribution margin, not model fit
Your decision
Did it move the margin? Then extend.
3 monthly
Run
What happens
The agent runs daily. We refresh the model, add plays, tune thresholds and extend to new segments. You own the model, the code and the cloud setup.
What you get
Daily agent runs and model refreshes
New plays and segments as you grow
Monthly finance report, same template
Your decision
Stay as long as it pays off.
Data you bring
Transactions with customer ID, date, amount, product and margin where available
Customer master data, including address change history
Marketing contacts and campaign responses (opens, clicks, unsubscribes)
Service contacts, complaints and payment history where relevant
Returns and refunds, consent and preference flags
Margin inputs (cost of goods, payment, shipping, returns) and acquisition cost per customer
At least two years of history
What you receive
Model, agent, pipeline, dashboards and a finance report template
Documentation, validation report and holdout results
A playbook catalogue with thresholds and expected uplift
Everything in your cloud, under your control
Trust and data
Your data stays yours
Runs in your cloud
The pipeline runs in your own cloud: AWS, Azure, GCP, Snowflake or Databricks.
Pseudonymised
Pseudonymised identifiers; no raw personal data leaves your environment.
GDPR from the start
A data processing agreement under Art. 28 GDPR. Consent and preference flags are respected when the decision is made, not afterwards.
EU AI Act and works council
The agent supports decisions inside plays and caps your team defines, and you choose how much it may do without approval; it does not profile people on its own. Credit-limit uses get human review, as the EU AI Act requires for consumer credit scoring from December 2027. We document purpose, scope and human oversight for your data protection officer and, where relevant, the works council.
Logged and auditable
Every agent decision is logged with its reason.
Owned by you
You own the model, the code and the data foundation.
Your contact from the first call to the live agent. 20+ years in data and AI leadership.
FAQ
Frequently asked questions
How is this different from the CLV feature in our CDP or CRM?
Packaged CLV features give you an average or a rule-based score inside their own platform. Our model is trained on your full data, lives in your cloud, writes to any system (including Meta and Google through their conversions APIs) and comes with an agent that acts on the forecast. If a packaged feature already fits your case, we will tell you on the call.
How is “agentic” different from “automated”?
Automation runs a fixed script. The agent combines a daily forecast per customer, action within guardrails and learning from holdouts. Your team owns the plays and thresholds; the agent handles the work across every customer, every day, and improves quarter after quarter.
Will the agent act without approval?
Only if you want it to. You choose the level per play: recommend only (a person approves every batch), approve by exception (a person approves anything above a threshold) or run within caps. Every pilot starts in recommend-only mode. Some steps always stay with a person (human-in-the-loop): at-risk top customers go to sales, finance approves budget shifts, and credit limits are reviewed by a person. Any play can be paused per segment, channel or customer in one click.
How accurate does the forecast need to be?
Accurate enough to beat the current decision. We measure value, not model fit: every play has a holdout group, and we report the uplift in contribution margin.
How do you calculate customer lifetime value?
For each customer: expected margin per order × expected number of orders in a fixed window (for example the next 12 months) − the cost of winning and serving that customer = CLV. Example: €155 margin per order × 9 expected orders − €155 cost = €1,240. The simple version uses averages; the useful version predicts each customer's orders from their own behaviour and is refreshed daily.
What is the difference between CLV, CLTV, LTV and customer value?
None in substance: CLV (customer lifetime value), CLTV and LTV are names for the same idea, and in German it is called Kundenwert or Kundenlebenszeitwert. What matters is how it is defined: margin, not revenue; a fixed window, not an unknown lifetime; one value per customer, not one average.
Is CLV a probability?
No. CLV is a value in euros: what a customer will be worth in contribution margin over the next 90, 180 or 365 days, net of the cost of serving and acquiring them. The probability that a customer will be active in that window is one of its inputs. The agent also uses it on its own, together with other signals such as the likelihood of accepting an offer.
Why a fixed horizon rather than “lifetime”?
A value over a specific window can be defended, tested and compared across customers. A value over an unspecified lifetime cannot. We match the horizon to your planning cycle and check it against actuals at every refresh.
What about customers with little history?
New customers get a starting estimate from similar customers, with a wider confidence range. The agent is more cautious with them until history builds up.
Does this work in B2B?
Yes, with adjustments. The customer is often an account with several contacts, the cycle is longer and sales matters. The model weighs company data, product mix, product usage and engagement more heavily, and the agent writes to the sales queue and account-based marketing. The first 90 days become the window from onboarding to the first expansion.
Why send CLV values to Google and Meta?
To win more customers like your best ones. Both platforms accept your own conversion values and customer lists. Google uses them in value-based bidding, and its new-customer goal can bid higher for prospects who resemble your high-value customers. Meta builds value-based lookalike audiences from a customer list with a lifetime-value column and optimises campaigns for value. Fed with predicted lifetime profit instead of first-purchase revenue, both start buying the customers who will be worth the most, not just the cheapest first orders. We start with a well-calibrated segment-level value and move to per-customer predictions once the pipeline is proven.
Can we start with CLV and add segmentation later?
Yes. Segmentation is the next module and reuses the same data foundation and pipeline; expect weeks, not quarters, for the second use case. Sometimes we recommend the other order, segmentation first, so marketing gets something to use while the CLV model is built.
Who owns the model?
You do. Model, code, pipeline, data foundation and documentation sit in your cloud. We hand them over and stay as long as you want us to.
How long until it changes a decision?
First decisions in the market within the pilot (8 to 12 weeks). The first measurable lift in contribution margin typically within two quarters after go-live.
We built a CLV model last year and nothing changed. What is different?
The model is roughly 30 percent of the work. The other 70 percent is the operating model: the agent, the plays, the connections to Meta, Google, CRM and the sales queue, the guardrails, the holdouts and the finance view. We deliver that layer, not just the model.
Is this only for very large companies?
Size is not the gate; spread is. The agent earns its keep when customers differ enough that treating everyone the same leaves money on the table. We have built this for companies with tens of millions of customers and for focused businesses with tens of thousands. What matters more is a repeat relationship and a channel the agent can act in.
How much history do we need?
Usually at least two annual cycles, so seasonality and the drop after the first purchase show in the data. If the base is too thin or too new, we start with Levels 1 and 2 and deliver the segmentation and planning value right away. We would rather tell you to wait than fit a model to too little data.
How fresh is the number?
A full refresh for every customer every night, plus updates during the day on the signals that matter most: a purchase, a return, a service contact, a clear sign of disengagement. A CLV recomputed quarterly is a report; one recomputed daily and written into the systems that spend money is an operating signal.
How do historical CLV, predictive CLV and the CLV in our CDP relate?
Historical CLV sums each cohort's realised margin: the auditable floor for board reporting. Predictive CLV estimates each customer's margin over a fixed future window: what the agent uses. The CLV score in a CDP or marketing tool is usually a simplified version of one of the two, usable only inside that tool. Our pipeline keeps all three consistent and makes the number available to every system you want.
Does this replace any of our systems?
No. It runs in the cloud you already use and writes to the CRM, marketing automation, ad platforms and sales queue you already have. We are deliberately vendor-neutral.
We looked at a packaged CLV feature from a large vendor. Why build this instead?
Packaged features are a sensible start when the goal is a score on a card. When the goal is a decision, the limits show quickly: data preparation is limited to what the vendor exposes, cost per customer usually cannot be modelled, the score stays inside that vendor's platform, and the agent layer is missing or tied to the vendor's playbook. If a packaged tool fits your case, we will say so.
What should we expect in P&L terms?
In our own projects the typical shape is a mid single-digit lift in contribution margin within two quarters, without more marketing spend. Most of it comes from two places: predicted CLV as the conversion value in Meta and Google, and a redesigned first 90 days. The result depends on careful preparation, especially cost per customer.
What goes wrong in CLV projects, and how do we avoid it?
Five patterns: one blended average that hides the spread (segment first, calculate second); a “lifetime” nobody defines (a fixed horizon, tested against actuals); revenue dressed up as CLV (contribution margin, net of all costs); a model that quietly drifts (monthly plan against actual, scheduled retraining); and a good number that never changes a decision (ship the agent and the connections with the model, and measure margin from day one).
Where does the lever land harder: B2B or B2C?
Both work; the shape differs. B2C pays back through marketing efficiency at scale: better bids, targeting and retention timing. B2B pays back through sharper priorities: a short list of accounts carries most of the margin, and the agent shifts where sales, account management and product attention go next quarter.
Does the CLV number feed anything beyond marketing and sales?
Yes: customer service levels, credit limits and payment terms, discounts, returns, recommendations, loyalty tiers, sales capacity, assortment and finance, up to valuing the customer base itself. See More uses. We add these once the marketing and sales uses are stable, and only where the regulatory framing in your jurisdiction is clear.
How is this priced?
The assessment has a fixed fee. The pilot runs on time and material, because data preparation is often what takes most of the time, and nobody can know how much before looking at the data. At the end of the assessment you get an estimate for the pilot straight away. Once live, there is a monthly fee. Sometimes we recommend starting with customer segmentation before CLV, so marketing has something to use early.Talk to us about your customers
Put the value of every customer to work.
Two identical customers acquired on the same day can end up with three times different lifetime values, depending on how well your systems treated each. Thirty minutes on a call, and you will see what your customer base is really worth, and what an agent could do about it.
