A spreadsheet has no counterfactual
The instrument itself cannot express the question, so the question quietly becomes a sentence in the memo.
IndustriesDiligence and value creation
The retention spend worked.Prove it.
Retention improved after they invested in it. The chart goes up. Neither fact establishes that the spend caused the improvement, and Nexyron is built to tell the difference.
Deals do not go wrong on recomputation.
Most of diligence is taking management's numbers apart and rebuilding them from source. Cohorts, retention, concentration, price realisation. A good associate does this, and increasingly good tools do much of it faster.
That is not where deals go wrong.
Management spent on retention and retention improved. Was that the spend, or the mix?
The customers a retention team, the save desk, chooses to call are not a random sample. They were chosen because they showed risk. Comparing them to uncontacted customers compares two populations that were never alike, and the number that falls out is the effect of the programme plus the effect of being the kind of customer somebody decided to call.
The data room AI never meets the billing data
Clause extraction at volume is mature. It has no access to the behaviour those clauses govern.
The subscription platform never meets the contracts
It sees the metrics and not the agreements, so it cannot tell you which renewal mechanism churns differently.
Every diligence process in this industry meets that problem. Almost all of them resolve it with a sentence that sounds analytical and commits to nothing.
What Nexyron does instead
Identity first, then the model, then the estimate.
Nexyron takes the billing export, the contract set, the CRM history, the support record and the org structure into one connected model, inside the clean team, on hardware the deal team controls.
Entity resolution runs first, because every downstream number depends on it. Shared attributes weighted by rarity, each proposed merge carrying the evidence that produced it and each one reviewable rather than silently applied.
Then the contracts and the behaviour occupy the same analysis. Whether one renewal mechanism churns differently from another stops being two exports and a hypothesis and becomes a single query.
Your subscription analytics platform computes the metric. Nexyron tells you whether the metric is about the thing you think it is about.
A simulated diligence
Constructed to show the method. No transaction, no target, no benchmark.
- 01
Is a customer a customer
3,140 billing records analysed for shared attributes weighted by rarity. Ninety-one resolve into 34 commercial relationships. Six of those materially change the concentration figure. Every merge carries its evidence and can be rejected.
- 02
Recompute the headline
Reported concentration, 18% by customer name. By resolved relationship, 31%. That is not a rounding difference. It is a different risk profile and a different price.
- 03
Which cohorts are churning
The aggregate looks healthy. One cohort acquired through a particular channel shows markedly weaker retention from month eighteen, and a strong adjacent cohort hides it. The agreements explain part of it: a different renewal mechanism, and the behaviour tracks the mechanism.
- 04
Did the customer success spend work
The room holds which customers were contacted, but not consistently when or why. Without the selection reason the comparison cannot be built, and Nexyron returns exactly that: this cannot be answered from this data, and here is precisely what is missing.
At 2am in week three, that converts an unanswerable question into a specific diligence request with a deadline on it.
- 05
What it looks like when it can be answered
Post-close, with the operating company's complete save-desk record. Comparable populations exist for most treated customers. The effect is real, concentrated in one profile, and close to nothing in another profile that receives a large share of the effort.
- 06
What can this data support
Retention and concentration, with stated caveats. Cohort behaviour, yes. Price realisation, partly. Effect of the retention programme, not from this extract, for the reason given. Produced in week one rather than discovered in week three.
Four things a spreadsheet cannot give you.
Identity comes first
Eighteen per cent and thirty-one per cent are the same data with a different answer to who the customer is.
The refusal is an output
An engine that declines where the data cannot support an estimate beats a confident number that will be wrong in a way nobody notices until year three.
The question becomes reusable
Unanswerable in diligence, answerable post-close. The same Lens, the saved analysis itself, re-run against the operating company's data rather than rebuilt from scratch by a different team.
The model survives the deal
Diligence work normally dies at signing. Here the model, the reports and the Lenses carry into the hold period as the first version of the value creation plan.
The dimension nobody in this category occupies
Every initiative in the plan will be credited with something.
Pricing, cross-sell, churn reduction, procurement, sales productivity. In three years each one will be described as having worked, and the evidence will be a chart with a vertical line on it.
Nexyron carries 42 causal and decision procedures on the same graph the customers live in. Uplift rather than association, propensity overlap checked before a number is believed, off-policy comparison of what a different plan would have produced, and interventions held as first-class objects so that what the operating partner did in March is a fact the model has rather than a slide somebody made.
Nothing else in the diligence stack does this, and it is the only part of the process that determines whether the thesis was actually right.
Running it
Inside the clean team.
Where a target competes with a portfolio company, clean-team obligations, the rules limiting who may see a competitor's data, are real, and uploading a competitor's customer-level data into a third-party service is precisely what counsel spends the whole diligence preventing.
Nexyron installs as an application and runs the entire engine on the deal team's own hardware. Nothing is uploaded to be understood.
Post-close the constraint relaxes and the reason to keep it changes: by then the argument is the analysis, and the model already holds two years of the company's behaviour.
Take a deal you closed three years ago.
One where an initiative was credited with a metric improvement you underwrote to. Ask what evidence the memo actually contained that the initiative caused it, rather than coincided with it.