Sign in
Set up for Financial services

AI for insights and analytics

AI analytics for financial services

A products-per-customer figure moves from 2.4 to 2.9 after duplicates are resolved and back to 2.3 after unused products are removed, and neither correction was visible in the headline. AI analytics reads the customer master and product files, resolves the view, and answers the segment head's questions with figures that say what they include.

Ask it in your own words

These are the questions financial services firms ask. Each one maps to a measure our tools compute from your files. The AI model chooses the measure and explains the result; the arithmetic is done by our code, and every total is checked.

The questionThe measure behind itWhat comes back
How many customers do we really have, and how many products each?Single customer view coverageCustomers after duplicate resolution, products per customer before and after, with the duplicates listed.
Which products were opened and never used?Opened and unusedProducts with no activity since opening, by product, channel and month, and the customers holding them.
Which segments hold fewer products than their norm?Products held against segment normProducts per customer against the segment median, with the gap in customers and value.
Who is showing signs of leaving?Balance and activity attritionCustomers whose balances and activity fell against their own baseline, ranked by value and pace.
Which high-value customers have not been contacted?Contact recency by valueDays since last contact by customer value band, with the untouched top band listed.
Which customers are moving between segments?Segment migrationCustomers whose value or activity moved them across a segment boundary, in both directions.

A conversation, with the figures

Every figure below was computed by a tool from the rows and checked before it was shown. The lines sum; the percentages match; each line opens to the rows that make it.

How many products were opened and never used last year?

12,400 products opened last year have had no activity since, across 3 channels. Branch channel accounts for 6,400 of them, 52% of the total.

Branch channel6,40052% of the total
Digital channel4,10033% of the total
Contact centre1,90015% of the total
The 3 lines sum to12,4000 unexplained

Which branch products?

3 products account for 5,600 of the 6,400 unused openings in the Branch channel, 88% of it. The savings account is opened alongside a current account as a matter of routine and then never funded.

Easy saver account3,100opened with a current account; 71% never funded
Overdraft facility1,600arranged and never drawn
Credit card900issued and never activated
These 3 are5,60088% of Branch channel

What should product and branch do about it?

Stop counting unused openings as sales, and start the funding conversation at opening. The tool gives the branch list, the product list and the customers, so both teams work from the same figures.

Product head, savingsChange the opening flow to include a first depositThis quarter
Branch networkContact customers with an unfunded saver opened in the last 90 daysThis month
Data leadReport products per customer after duplicates and unused removedMonthly
How many customers do we really have, and how many products each?Which products were opened and never used?Which segments hold fewer products than their norm?Who is showing signs of leaving?

The measures behind the answers

Each measure has one formula, one source and one meaning. They are computed per segment and relationship owner and in total, and every one carries an identity that must hold before it is shown.

MeasureFormulaFromWhat it tells you
Single customer view coverageAccounts, and revenue, linked to a resolved customer ÷ total, by match methodProduct systems; customer masterWhether any per-customer measure can be trusted
Products held against segment normActive products held ÷ median for the customer's segmentHoldings; segment fileCustomers thinner than their peers
Opened and unusedProducts with no funding or transaction within the window ÷ products opened, by product and originProduct systems; transaction fileCross-sell that did not happen
Tenure against depthCustomers by tenure band and products held; long tenure, low depth listCustomer master; holdingsLoyal customers never developed
Segment migrationCustomers moving up or down a value band in the period, with the norm re-appliedBalances; revenue; segment rulesCustomers outgrowing their service model
Primary relationship indicatorsSalary or main income credited; share of outgoings through the accountTransaction fileWhether the firm is the main provider
Balance and activity attritionBalances and transaction counts, last 3 months ÷ own prior yearBalance and transaction filesCustomers leaving before they close anything
Contact recency by valueRevenue of customers contacted within cadence ÷ revenue of managed customersCRM; revenue fileHigh-value customers nobody has spoken to
Repeat complaints and contactsCustomers with a second complaint or the same issue within 30 days ÷ customers who complainedService logService failures that precede attrition
Revenue concentrationTop decile's share of revenue; largest relationshipsRevenue file, by resolved customerDependence, visible only once customers are resolved

Each measure is worked through, with the export it comes from and what to drop, in Customer KPIs for financial services firms.

What the AI model does, and what our tools do

The AI model chooses the measure

From your question and the measures declared for financial services, the model picks the one that answers it, and the period and comparison the question implies.

Our tools do the arithmetic

Deterministic code reads the rows, computes the measure, and checks the identities below. The same question on the same data gives the same answer, every time.

The AI model explains, and cites

The AI model writes the sentence around the result, naming the segment or account behind it. It states no figure that is not in the result, and every figure links to its rows.

The identities that must hold

Customer viewAccounts = linked to a resolved customer + unlinked; revenue likewise
HoldingsProduct revenue by customer sums to the revenue ledger
Opened and unusedOpened = active + unused + closed within window
MigrationCustomers at start + new − lost = customers at end, per band, with moves netting to zero

What it reads

The exports financial services firms already produce. Column names are mapped once and the mapping is reused. A file is the way in; scheduled delivery and connections to your systems come with the plan, and every source is listed here.

  • Product systems
  • Customer master
  • Holdings
  • Segment file
  • Transaction file
  • Balances
  • Revenue
  • Segment rules
  • Balance and transaction files
  • CRM
  • Service log

Who owns each answer

An answer is a list with an owner and a cadence, or it is a chart nobody works.

MeasuresOwnerCadence
Single customer view coverageData or operations lead, with each product headMonthly
Attrition signals; contact recencySegment heads; relationship ownersWeekly to monthly
Products against norm; tenure and depth; opened and unusedHead of customer or distributionQuarterly
Migration; concentrationSegment heads; financeQuarterly

Questions financial services firms ask about AI analytics

How does it resolve duplicate customers?

By the rules you give it on the data map: matching identifiers, then names and addresses within the tolerance you set. Every merge is listed, so the operations lead can see which records became one customer and undo a match that is wrong. The products-per-customer figure is shown before and after.

Why does it show a headline with and without unused products?

Because the two figures answer different questions. Products held says what the book looks like; products used says what the selling achieved. Reporting only the first hides openings that were never funded, and reporting only the second hides the book. The tool shows both and the difference.

What does attrition look like in the data?

A fall in balances and in activity against the customer's own baseline, over a pace the segment head sets. It is a signal, not a prediction: the customer is listed with the figures, and the relationship owner decides what to do. No score is produced by a model.

Which files are needed?

The customer master, products held by customer, balances or activity by month, and the contact log. Names can be replaced with identifiers before anything is uploaded, and every file is matched to your column names once.

See it on your data

Bring a few thousand rows. The data map opens next, every column mapped once, and the first question is answered in minutes. Free, in your browser, no account.