Skip to content
EncryptTechBD
Home Industries Our Work Blog Pricing About Contact Discuss Your Project
বাংলা (BN)
Software Engineering 6 min read

Technical debt management: a buying guide for business software

How to buy technical debt management without funding a rewrite: visibility, supplier models, scorecard evidence, pilot boundaries and handover terms.

E
EncryptTechBD · 0 comments
Programmer writing code with unit tests on screen
Photo: Joonspoon / Wikimedia Commons (BY-SA)

Technical debt is a useful metaphor until it becomes a budget line. At that point the argument shifts from engineering craft to procurement: what will we fix, who will do it, and how do we know the work reduced risk rather than rearranged code?

That shift matters because debt is not one thing. It includes shortcuts that slowed delivery, dependencies that no longer receive security fixes, tests that were never written, documentation held in one person's memory and architecture that no longer matches the business. A supplier who treats all of it as a rewrite opportunity is not selling management.

Make the debt visible before you shop

Start with a register rather than a code review. For each item, record the affected system, the business process it supports, the symptom it causes, the age of the workaround, the effort to fix and the cost of leaving it. Symptoms are more persuasive than code smells: a month-end report that needs manual correction, a checkout change that takes three weeks, or a security patch that cannot be applied without breaking an integration.

Group items into three buckets: fix now because they create operational or security exposure, fix next because they slow every change, and monitor because they are tolerable at current scale. This register becomes the requirement document you give suppliers, and it prevents a vendor from defining the problem for you.

Requirements that separate advice from a rewrite pitch

  • Inventory: repositories, services, integrations, databases, scheduled jobs and the owner of each.
  • Evidence of interest: how long changes take, how often releases fail, how many incidents trace to the same component, and where security findings cluster.
  • Constraints: systems that cannot change this year, contractual limits, data residency, peak trading periods and team capacity.
  • Safety net: test coverage, staging parity, rollback and backup verification before refactoring begins.
  • Knowledge transfer: documentation, pairing and a named person who can maintain the result.
  • Success measures: lead time, change failure rate, incident count, dependency age and the number of manual workarounds removed.

Ask every supplier to respond to the same register. Proposals that ignore half the items and price a platform replacement are answering a different question.

The supplier models compared

Code quality platforms scan repositories and track issues. They create visibility but do not remove debt, and they work best when an internal team already has capacity. Specialist consultancies assess, design and often execute remediation; they bring pattern knowledge but can leave a dependency if handover is weak. Staff augmentation adds engineers to your team and keeps knowledge inside the business, though it requires management capacity. Managed modernisation takes responsibility for a defined outcome over a longer period, which suits businesses without an engineering manager but carries the highest lock-in risk.

There is also a fifth option: an external review combined with internal execution. It is often the cheapest way to start because it tests whether your team can absorb the work before you commit to a programme.

Scorecard criteria and the evidence to demand

Weight the criteria before proposals arrive. Useful categories are diagnosis quality, remediation depth, safety practices, knowledge transfer, commercial fit and exit. For each, write the evidence you will accept.

  • Diagnosis: a sample of the register traced to code and runtime behaviour, not a questionnaire score.
  • Remediation: a named engineer, a work breakdown and a definition of done for each item.
  • Safety: test strategy, staging approach, rollback plan and change freeze handling.
  • Transfer: documentation standard, pairing hours and who owns the backlog afterwards.
  • Commercial: what is fixed, what is time-based and what triggers a change request.
  • Exit: repository access, credentials, runbooks and a transition period.

Do not score a vendor on the elegance of the target architecture. The relevant question is whether the proposed sequence reduces a named business constraint within a period you can fund.

Commercial structures and what they reward

A fixed-price assessment is useful because the output is bounded. Remediation is harder to fix-price honestly: the work reveals more work. Capped time-and-materials with a written stop point, or a series of small fixed-scope releases, usually produces better decisions than one large transformation contract. Retainers can work for continuous improvement, provided the backlog and acceptance criteria are visible to you.

Ask how the supplier is compensated when they find more debt. A model that rewards discovery without a cap encourages an ever-growing programme. A model that rewards only delivery speed can encourage shallow fixes. The contract should tie payment to accepted remediation and evidence, not to activity.

A bounded pilot on one service

Choose one component with a clear symptom and a measurable outcome. For example, a checkout service that requires manual intervention, or a reporting job that fails at month end. Give the pilot a fixed window, a named engineer and a rollback plan. Acceptance should include a passing test suite for the changed area, a runbook, a before-and-after measurement and a review by the person who maintains the system.

During the pilot, watch how the supplier handles an unexpected dependency. That moment reveals whether they diagnose or simply replace. Also check whether they document as they go or leave it for the final week.

Handover, documentation and exit

Debt work creates knowledge. If that knowledge leaves with the supplier, you have converted one dependency into another. Require repository-level access from day one, architecture decision records for significant changes, runbooks for anything operational and a named internal owner for each service. Agree in advance what happens at the end of the engagement: credentials, backlog, monitoring configuration and a support period.

Warning signs in a proposal

Treat a rewrite recommended before any code has been read as a sales position. Be cautious about a single maturity score with no underlying evidence, a plan that removes tests to move faster, an architecture diagram with no migration sequence, and a team that cannot name the people who will do the work. Also question any proposal that cannot explain how the first release will be rolled back.

Choosing by constraint, not by fashion

If the constraint is security exposure, prioritise remediation depth and patch discipline. If the constraint is slow delivery, prioritise test coverage and the removal of coupling. If the constraint is knowledge concentration, prioritise pairing and documentation over code volume. If the constraint is budget, buy a diagnosis and one bounded fix rather than a programme you cannot sustain.

Our part in this work

We assess and remediate business software inside our Custom Software division. The custom web applications service covers the application layer, including integrations and handover. The pricing page explains how assessments and delivery are separated. For the budgeting context around this decision, read how SME IT budgeting changes as the business grows.

Sources

Services that deliver this

Want this implemented properly?

Pricing →

Every engagement starts with a written scope. If a requirement is outside it, we quote it separately before starting.

Responses (0)

No comments yet. Add the first one below.

Leave a response

Your email address will not be published. Required fields are marked *.

Keep reading

More field notes