A mirror, not a model.
Keystone IQ’s Revenue Quality Architecture™ — you cannot understand what a customer base is worth by looking at the average.
The problem with averages
You have to understand which customer types are generating the revenue, how those relationships behave, and how robust they are when conditions change. Two cohorts can show similar revenue today and carry very different value.
- One is embedded across the relationship, with repeated and diversified engagement that tends to hold up under pricing pressure, new competition or product change.
- The other is narrower, more contingent and more exposed, and can look healthy in the numbers until something small moves against it.
Standard retention and Net Retention Rate cannot separate those two. They also cannot show where customers are strengthening, drifting or moving between states. Our methodology is built to do just that.
The Keystone IQ methodology
We rebuild the base from observed behaviour at individual customer level, using the underlying dimensions that actually create and sustain value in that business. Those dimensions are calibrated to the structure of the base itself, so the picture reflects the real mechanics of customer value, not a generic model applied on top of it.
The methodology does not stop at classification. It tracks established behaviour patterns over time, making movement and trajectory visible early: which customer types are stable, which are strengthening, which are beginning to weaken, and where value is migrating into less resilient parts of the base before that deterioration is visible in the headline numbers.
A forward view, not just a static one.
That creates a forward view, not just a static one. It allows us to attach a number to the base through three forward scenarios:
- What the base will produce if current behaviour continues.
- What changes if at-risk value is stabilised.
- What changes if stronger customer groups are developed further.
Our Revenue Quality Architecture™
The output is Keystone IQ’s Revenue Quality Architecture™ expressed in revenue terms: what the base is supporting now, where it is heading next, and what management action can still change.
The judgement above the method
Customer base analysis done without deep commercial and analytical expertise produces results that look credible and are frequently wrong.
The Revenue Quality Architecture surfaces the picture. Turning that picture into a credible investment view requires judgement the method alone cannot produce. Customer base analysis done without deep commercial and analytical expertise produces results that look credible and are frequently wrong.
- Deterioration signals misread as seasonal noise.
- Customer types that look like growth opportunities but are actually the highest-cost, lowest-return populations in the base.
- Conclusions that confirm what the model was set up to find rather than what is actually happening.
Twenty years inside listed ecommerce, subscription, streaming, SaaS, consumer, retail, loyalty and financial services businesses is what distinguishes a signal from noise, a structural risk from a temporary pattern, and a finding that changes a deal from one that does not.
That judgement sits above the analytics. It is the reason the output is a commercial verdict, not a data report.
The same data. A different picture.
What a retention rate does not show you
Standard cohort analysis produces aggregate metrics: retention rates, NRR, average order value, cohort growth curves. Those metrics describe the customer base as a whole. They do not show what is happening inside it.
The retention rate is accurate. It is also misleading.
The same data. Two readings.
What a standard cohort read shows
70% retention rate.
Sounds healthy. The business is retaining its customers.
What a structural read shows
Within that 70%:
The 30% who left were disproportionately the highest-value customers. The base is getting cheaper to retain and less valuable to own.
Why other approaches produce a different answer
Three analytical approaches dominate commercial due diligence today. Each has a structural limitation that Keystone IQ’s Revenue Quality Architecture is designed to address.
| Dimension | Traditional advisory | Assumed-model providers | Keystone IQ |
|---|---|---|---|
| Data basis | Aggregates, averages, cohort summaries | Probability model (e.g. Pareto/NBD) fitted to aggregate data | Individual-level observed transaction data |
| Time horizon | Point-in-time snapshot or trailing 12 months | Model-projected (forward-looking from a single calibration point) | 12 to 48 months of observed movement |
| Distribution assumption | Implied (averages assume normality) | Explicit (Pareto/NBD assumes a specific shape) | None — the natural distribution is observed, not assumed |
| Interpretation | Analyst-led, framework-driven | Algorithmic, model-output-driven | Operator-led: twenty years of experience reading customer bases from inside the business, baked into proprietary AI models with operator-led interpretation |
| Speed | 3 to 6 weeks | 2 to 4 weeks | 5 to 10 days |
Same business. Same data. The full picture is what changes the investment decision.
Ready in days, not weeks
What we need to get started
We work from anonymised transaction data. No customer names. No contact details. No personally identifiable information of any kind.
What we need
Transaction ID, anonymised customer ID, date, value, and product category. That is all.
Where it comes from
Most businesses can extract this from their ecommerce platform, billing system, or point-of-sale data within 24 hours.
Data quality
We assess usability before the engagement formally starts. If the data is insufficient, we say so before any fees are committed. If it is messy but recoverable, we resolve it during ingestion.
Timeline
Health Check: five days from clean data. Customer Base Diagnostic: ten days from clean data.
Methodology FAQs
Probability models assume a distribution shape before fitting the data. If the real customer base matches that assumed shape, the model works. If it does not, and most do not, the model smooths over the customer type movements and habit changes that matter most for the investment thesis. We observe actual behaviour directly and track how it moves.
Theta fits a statistical model to transaction data and projects forward from a small number of parameters. It treats the customer base as one population. We identify the real customer types through rule-led segmentation built from observed behaviour in that specific business, then track how those types and their habits are moving over time. The starting point is observation, not assumption.
The Revenue Quality Architecture™ is Keystone IQ's proprietary methodology. The competitive advantage is not the framework description. It is the twenty years of operator experience that calibrates it to each business and interprets what the output means for the investment thesis.
The methodology works wherever transaction-level customer data exists: ecommerce, retail, subscription, streaming, loyalty, financial services, and B2B services. The dimensions that drive value differ by sector. The method adapts to the structure of the base, not the other way around.
We assess data quality at the scoping stage before any fees are committed. If the data is insufficient to support a credible analysis, we say so. If it is messy but recoverable, we resolve quality issues during ingestion rather than passing uncertainty through to the output.
Commercial due diligence assesses the market. The Revenue Quality Architecture™ assesses the asset: at individual customer level, across every dimension that drives value, over 12 to 48 months of observed movement. CDD tells you about the competitive environment. We tell you what the revenue is actually made of and where it is heading.