Substitution, Comparison, Remove-and-Restore and Rollback
Where you are: Module 6 — Isolation Techniques · Lesson 6.2 · Based on Chapter 6 (§6.2, §6.3, §6.5, §6.6) of Dr. Tahseen Al Doori's Troubleshooting Essentials.
What you'll be able to do
- Use substitution with a known-good part and interpret the result.
- Use comparison against an identical working system or a baseline to identify a likely cause.
- Apply remove-and-restore, including the role of safe mode, to find a fault caused by an added item.
- Use rollback when a fault follows a known change, and explain why rollback is not a permanent fix.
Watch: Substitution, Comparison, Remove-and-Restore and Rollback
The video introduces four techniques that all rely on a known-good reference, then applies each one: a comparison table between two office PCs, a step-by-step remove-and-restore, and a graphics driver rolled back after a screen begins to flicker. As you watch, consider:
- What does "known-good" mean, and why does it make substitution conclusive?
- In remove-and-restore, why is the last item restored the cause?
- Why does rolling back an update prove the cause but not finish the job?
One idea, four techniques
Half-split (Lesson 6.1) needs a chain and test points. The four techniques in this lesson work differently: each compares the faulty system with a reference that is known to be good.
| Technique | The known-good reference |
|---|---|
| Substitution | A part confirmed to work |
| Comparison | An identical working system (or a recorded baseline) |
| Remove and restore | The system at its minimum |
| Rollback | The system as it was before the last change |
Substitution: swap with known-good
The book's glossary defines known-good as a component or system confirmed to work, used for comparison or substitution. Substitution replaces a suspected part with one known to work: the book's examples are a different cable, charger, mouse, browser, user account or device. If the problem disappears, the removed part is at fault.
Substitution is fast and conclusive when spare parts are available. A few rules keep it reliable:
- Swap one part at a time. Swapping a cable and a charger together breaks the principle change one thing at a time: if the fault disappears, you will not know which swap fixed it.
- Be sure the replacement really is known-good. A "spare" from a drawer may be faulty too. Ideally use a part taken from a working setup.
- Substitution works in both directions. You can also move the suspect part to a working system: if the fault follows the part, the part is at fault.
- Stay with external, swappable parts. Cables, chargers, peripherals, browsers and test accounts are fair game. Internal components are not a quick swap: the book's safety guidance requires shutting down, unplugging, discharging and anti-static protection before opening any computer, and authorisation for equipment under warranty.
Comparison: find the difference
Comparison places the faulty system beside an identical working system and compares settings, versions, cables and behaviour. The book's key sentence: the difference between them is usually the cause. It is especially effective in organisations where many devices share a standard configuration, because the "identical working system" is the next desk along.
(Fictional example.) Two office PCs are built from the same standard. One cannot open websites by name; the other can.
| Setting | Faulty PC | Working PC |
|---|---|---|
| Operating system build | Same version | Same version |
| Network port | Wall port 12 | Wall port 14 |
| Default printer | Floor 2 printer | Floor 2 printer |
| DNS server | 203.0.113.53 | 192.168.10.1 |
Not every difference is a cause. The network ports differ, but both are known to work, so that difference is expected. The DNS server, the service that translates names into addresses, is different and has nothing to do with the office standard, and it matches the symptom. It becomes the leading theory, which you then test (Step 3) by correcting it and retrying.
If no working twin is available, use a baseline (Chapter 4): a record of how the system looked and performed when it worked correctly. A baseline is a working twin on paper.
Remove and restore
Remove-and-restore suits systems with many extras. The book's procedure:
- Reduce the system to its minimum. Disconnect accessories, start in safe mode (a start-up mode that loads only essential drivers and software), and disable add-ons and startup programs.
- Test. If the minimal system works, the fault lies in something you removed.
- Restore items one at a time, testing after each.
- When the fault returns, the last item restored is the cause.
Safe mode supports this technique by providing the minimal starting point automatically: if a fault disappears in safe mode, the cause is very likely one of the drivers or programs that safe mode did not load.
Beyond the book (2026 note): Microsoft's support guide to a clean boot applies remove-and-restore to Windows: it starts Windows with only essential drivers and startup programs, and Microsoft describes it as giving more control than safe mode for isolating a problem. To find the culprit, the guide suggests re-enabling half of the disabled services at a time, which eliminates half of them as the potential cause in each test: half-split inside remove-and-restore. The guide also warns that the tool can make a computer unusable if misused, and that functionality may be temporarily reduced. Practise it only on a training machine or virtual machine, with permission. Source: Microsoft Support — How to perform a clean boot in Windows.
Rollback: undo the last change
If a problem began after a change, reverse that change: roll back the update, restore the previous setting or disconnect the new device. If the problem disappears, the change was the cause, and a permanent solution can be planned.
Rollback is the technique that puts Module 2's question, What changed?, directly to the test. It is also why Step 4 of the method insists on recording current settings and planning how every change will be undone.
Rollback proves the cause and gets people working again, but it is rarely the end. Chapter 10 of the book is explicit: when an update causes a fault, roll it back, report the problem to the vendor and pause automatic updates for that component until a corrected version is released, but never leave a system permanently unpatched as a workaround, because updates fix security problems as well as bugs.
Beyond the book (2026 note): In Windows 11, Microsoft documents uninstalling an update through Settings > Windows Update > Update history > Uninstall updates. Microsoft notes that some updates cannot be uninstalled and advises understanding the risks before removing a security update. In an organisation, follow your change-management process rather than removing updates yourself. Source: Microsoft Support — How to uninstall a Windows Update.
Worked example
(Fictional scenario.) On Monday afternoon a design laptop receives a new graphics driver. On Tuesday morning the user reports that the screen has flickered ever since.
- Identify. Since Monday's driver update, the laptop screen flickers constantly; an external monitor connected to the same laptop also flickers.
- Theory. The fault began immediately after a known change, so the book's table points to rollback first.
- Test. The current driver version is recorded, and the driver is rolled back to the previous version. The flicker stops; after re-installing the new driver on a test basis, it returns. The change is confirmed as the cause.
- Plan and implement. Return to the previous driver, pause that driver update, and report the problem to the vendor with the version numbers.
- Verify. The user works for a day with no flicker. When the vendor releases a corrected driver on Wednesday, it is installed and the screen is checked again.
- Document. The ticket records both driver versions, the rollback and the vendor report, so colleagues with the same laptop model are warned.
Notice the bonus test: re-applying the change and seeing the fault return makes the conclusion very strong.
Try it
Goal: practise substitution and comparison safely.
- Substitution. Pick an accessory with a swappable part, for example a phone and its charging cable. Write down how you would test whether the cable is faulty using a known-good cable, and how you would test the phone with the original cable on a second charger. Note what each result would mean.
- Comparison. With a classmate who uses the same operating system, compare five settings on your two devices (for example operating system version, browser version, time zone, default printer, screen resolution) in a table. Do not change anything.
- For every difference you find, write one sentence on whether it could explain a hypothetical fault of your choosing.
You have finished when you have a written substitution plan with interpreted outcomes and a five-row comparison table with a judgement for each difference.
Knowledge check
You swap a suspect charging cable for a known-good one, and the laptop starts charging. What does this show?
Comparison is described as especially effective in which situation?
During remove-and-restore you restore add-ons one at a time. The fault returns after the fourth add-on. What is the most likely cause?
A fault began straight after a driver update, and rolling the driver back removes the fault. What should happen next?
Answers explained
- B. Only one part changed and the problem disappeared, so the removed part (the original cable) was at fault. The battery and socket were not swapped, and only one thing changed.
- A. The book says comparison is especially effective where many devices share a standard configuration, because an identical working system is readily available. With no second device, comparison needs a baseline instead.
- C. The last item restored before the fault returned is the cause. The first three were restored without the fault returning, and safe mode is the minimal state that worked.
- B. Rollback proves the cause; a permanent solution must be planned. The book warns never to leave a system permanently unpatched, and every incident should be documented.
Key takeaways
- Substitution: swap one suspect for a known-good part; if the problem disappears, the removed part is at fault.
- Comparison: the difference from an identical working system (or a baseline) is usually the cause.
- Remove and restore: strip to the minimum, restore one item at a time; the last item restored is the cause.
- Rollback: undo the last change to prove it caused the fault, then plan a permanent fix and never stay unpatched.
Sources
- Dr. Tahseen Al Doori — Troubleshooting Essentials, University Edition (2026), §6.2, §6.3, §6.5, §6.6 (with cross-references to §2.3, §3.4, §4.5, §8.6 Safety, §10.3 and the Glossary)
- Microsoft Support — How to perform a clean boot in Windows
- Microsoft Support — How to uninstall a Windows Update
