New e-book: Oracle APEX 26.1 – AI that survives an audit

Application portfolio rationalisation: a scoring model to decide what to retire, retain, refactor or replace

Wojciech Bielawski

Head of Growth

  • November 20, 2026
7 min read

Contents

Most organisations know roughly how many applications they run. Far fewer can say which of them they would build again today, which they would switch off tomorrow and which they keep only because nobody has checked. When the modernisation budget covers a handful of applications and the portfolio holds dozens, that knowledge decides where the money goes.

This article gives you a scoring model you can fill in with data you already have. It starts from Gartner’s TIME model, adds the evidence TIME leaves out, sets weights and thresholds, and works through twelve fictional applications so that you can see how the numbers turn into decisions.

Start from TIME, then add what it leaves out

The most widely used starting point is Gartner’s TIME model, usually traced to Gartner analyst Jim Duggan’s 2010 research on application portfolio triage. It rates each application on two axes, technical fit and functional fit, and places it in one of four boxes (SAP LeanIX summary):

Low functional fit High functional fit
High technical fit Tolerate Invest
Low technical fit Eliminate Migrate

TIME is simple enough to run in a workshop, which is both its strength and its weakness. In practice three things are missing:

  • Usage. Fit is scored by opinion. How many people actually use the application, and how often, rarely enters the score.
  • Granularity. TIME scores whole applications, while the right decision often differs between modules of the same application.
  • Cost and support risk. What the application costs to run and how close it is to the end of vendor support appear only indirectly.

The model below adds all three and maps the result to four decisions:

  • Retire: switch it off and archive the data that retention rules require you to keep.
  • Retain: keep it as it is, with routine maintenance.
  • Refactor: keep the business logic and change the technology or structure underneath it.
  • Replace: move the function to a different system, either a product or a new build.

The scoring model

Score each application, or each major module, from 1 to 5 on six criteria. A higher score always argues for keeping it: more value, more use, better condition, lower risk, lower cost, better-controlled obligations.

Criterion What 1 means What 5 means Where the data comes from Weight
Business value Nothing important depends on it The business stops without it Process owners; process maps 25%
Usage No use across a full business cycle Daily use by many people or processes Database and application logs; single sign-on records; licences assigned versus used 20%
Technical condition Fragile, frequent incidents, hard to change Healthy, tested, easy to change Incident and change records; code analysis 20%
Support risk Out of vendor support; skills hard to find Fully supported; skills available Vendors’ lifetime-support policies; team skills 15%
Run cost relative to value High Low Licence and support invoices, infrastructure, people’s time 10%
Regulatory exposure Heavy obligations, weak controls Light obligations or strong controls Compliance register; audit findings 10%

Combine the six scores into two numbers:

  • Need = (business value × 25 + usage × 20) ÷ 45
  • Health = (technical condition × 20 + support risk × 15 + run cost × 10 + regulatory exposure × 10) ÷ 55

Then apply the decision grid, with 3.0 as the threshold on both axes:

Health below 3.0 Health 3.0 or above
Need 3.0 or above Refactor or replace Retain
Need below 3.0 Retire Retain at minimum cost, or consolidate

Two rules override the grid:

  1. No usage, no exception. An application with a usage score of 1 across a full business cycle is a retirement candidate whatever its value score, until its owner shows where the value comes from.
  2. Retention obligations. If the law requires the data to be kept, retiring means archiving the data and switching off the application, not deleting anything.

For applications in the “refactor or replace” box, ask one more question: is the business logic something that sets you apart, or a standard process? Refactor the first. Replace the second with a product, unless making the product fit would mean rebuilding the old system inside it.

A worked example: twelve applications

The portfolio below is fictional. Scores run from 1 to 5; Need and Health are calculated with the formulas above.

# Application Value Usage Condition Support Cost Regulatory Need Health Decision
1 Contract management (custom Java) 5 5 2 2 3 3 5.00 2.36 Refactor
2 Commission calculation (PL/SQL and reports) 5 4 2 1 2 2 4.56 1.73 Replace
3 Customer portal 4 4 4 4 3 4 4.00 3.82 Retain
4 Legacy BI (Oracle Discoverer) 3 2 1 1 2 2 2.56 1.36 Retire
5 HR training tracker (spreadsheets) 3 4 1 3 4 2 3.44 2.27 Replace
6 Pricing engine (PL/SQL) 5 5 3 3 3 4 5.00 3.18 Retain
7 Document archive 2 1 3 2 2 5 1.56 2.91 Retire, archive the data
8 Field service app (Mendix) 4 4 4 5 3 4 4.00 4.09 Retain
9 Supplier onboarding workflow 3 2 3 4 4 3 2.56 3.45 Retain, consolidate later
10 Old intranet 1 1 2 1 3 4 1.00 2.27 Retire
11 Integration hub (Spring Boot 2) 4 5 3 2 3 3 4.44 2.73 Refactor
12 Management dashboards 3 3 4 5 4 4 3.00 4.27 Retain

Three applications are retired, two replaced, two refactored and five retained. The modernisation budget goes to four applications instead of twelve, and the three retirements free licences, infrastructure and support time.

A few results are worth a closer look:

  • 2 and 1 land in the same box but get different decisions. Commission plans in this example are standard, so a product can cover them; contract rules are specific to the business, so contract management is refactored.
  • 7 is retired by the override rule. Nobody has opened the archive in a full business cycle, but the documents must be kept, so the data is archived and the application switched off.
  • 12 sits exactly on the Need threshold. A score of 3.00 is a reason to look again, not a decision.
  • 1 looks different module by module. Scored as a whole, contract management is heavily used. Split into contract entry, approvals, archive search and reporting, the archive search scores 1 on usage and moves to “retire”, which shrinks the refactoring scope.

Where the numbers come from

  • Usage. Database and application logs show which modules, reports and tables are used; on Oracle estates most of that evidence is already in the database. Single sign-on records give the number of distinct users per application per month, and licence reports show assigned versus active users.
  • Technical condition. Incident tickets per month, failed changes, test coverage and the findings of a code scan.
  • Support risk. Vendors’ lifetime-support pages: for example, Oracle ended Premier Support for Java 17 in September 2026, and open-source support for Spring Boot 3.5 ended in June 2026. Add how easy it is to find people who can work on the technology.
  • Run cost. Licence and support invoices, infrastructure and the share of the team’s time the application takes.
  • Regulatory exposure. The compliance register, audit findings and retention obligations.

Five mistakes that distort the result

  1. Scoring by committee without usage data. The loudest owner wins. Put the usage figures on the table before anyone scores value.
  2. Scoring whole applications only. Dead modules hide inside live applications and get rebuilt with them.
  3. Treating unused as deletable. Retention rules turn some retirements into archiving.
  4. False precision. A Health score of 3.09 against 2.91 is not a decision. Review by hand every application within about 0.3 of a threshold.
  5. Deciding everything at once. “Not now” is a valid outcome. Give it a date and a trigger for the next review, such as the end of vendor support.

How to run it

  1. Build the inventory of applications and their main modules.
  2. Collect usage, incident, cost and support data before the scoring workshop.
  3. Score, calculate Need and Health, and apply the grid and the two override rules.
  4. Review the results with each application’s owner.
  5. Sequence the work: retirements first, because they free budget and shrink the scope of everything that follows.

Rationalisation can also mean consolidation. T-Mobile’s sales consultants worked in three separate CRM systems until a single custom platform replaced all three. Once the decisions are made, our article on phased legacy modernisation covers how to sequence the work.

This kind of decision map is what we build at the start of a modernisation, before anyone estimates a rebuild. If you would like help scoring your portfolio, book a discovery call.

Looking for a software development company?

Work with a team that already helped dozens of market leaders. Book a discovery call to see:

  • How our products work
  • How you can save time & costs
  • How we’re different from other solutions

footer-contact-steps

We keep your data safe: ISO certified

We operate in accordance with the ISO 27001 standard, ensuring the highest level of security for your data.
certified dekra 27001
© 2026 Pretius. All rights reserved.