How many times have you looked back at a quarter and thought, I did a lot of good work there, only to realize you can’t actually reconstruct most of it?
That’s the problem.
Not that the work didn’t happen.
Not that it wasn’t valuable.
The problem is that by the time review season comes around, the details are gone. The exact metrics. The decisions you pushed for. The tradeoffs you navigated. The people you unblocked. The work that made everything else easier.
And once that information fades, your promotion packet gets weaker.
That’s what I mean by memory fade.
It’s the gap between the impact you actually had and the impact people can still remember when it matters.
The part nobody tells you
Engineers tend to assume good work will speak for itself.
In practice, it rarely does.
Some of your work disappears because people forget. Some of it disappears because it was invisible in the first place. And some of it disappears because the promotion process itself is built on partial memory and incomplete narratives.
That’s a bad combination.
If you are doing Staff-plus level work, this gets even worse, not better. The higher you go, the less your value is tied to tickets closed and the more it depends on influence, leverage, judgment, and cross-functional impact.
Those are all real contributions. They are also the easiest ones to lose if you do not document them as they happen.
So the core lesson is simple:
Your career does not run on memory. It runs on evidence.
Why memory fades
There are two forces at work here.
The first is cognitive. People forget. A lot.
The details of a project feel obvious when you are inside it, but a few weeks later they start to blur. That is the basic shape of the forgetting curve: memory drops quickly at first, then more slowly over time. Replications of Ebbinghaus’s work still show that early decay is real, not just a classroom metaphor.
The second is organizational.
Performance reviews are not a perfect archive of your work. They are a reconstruction exercise. Managers, peers, and calibration groups are trying to reassemble months of activity from memory, notes, and whatever evidence survived the cycle. Performance-appraisal research has documented primacy and recency effects, which is exactly why the work from last week can crowd out the work from six months ago.
The work that gets rewarded is not always the work that mattered most.
It is often the work that was:
- easiest to explain,
- most recently visible,
- best advocated for,
- or most neatly documented.
That is a very different game.
What gets lost
Memory fade hits some kinds of work harder than others.
The obvious things are easier to remember:
- a big feature launch
- a production incident
- a metric spike
- a visible cross-team project
The less obvious things decay fast:
- debugging the issue nobody else could isolate
- mentoring a junior engineer through a tough quarter
- reviewing design docs that prevented a bad architecture decision
- unblocking product, design, or infra at the right moment
- keeping a system stable so nobody ever noticed the fire you prevented
That last category is where many strong engineers get hurt.
If your work is mostly reliability, leverage, mentorship, or coordination, it may be essential and still invisible. People tend to remember events, not maintenance. They remember outcomes, not the scaffolding that made the outcomes possible.
That is why “I was just doing my job” is such a dangerous sentence.
Sometimes it is technically true.
It is also exactly how important work vanishes.
The promotion problem
This gets sharper during calibration.
At that point, the room is not asking, “Who worked hardest?”
It is asking, “Who can we confidently defend?”
That is why a promotion packet is not just a summary of work. It is a case file.
If the evidence is thin, people fill the gaps with assumptions:
- remember the most recent work.
- trust the person who speaks most confidently.
- overrate visible wins and underrate invisible labor.
- conflate team success with individual contribution.
None of that is malicious by default.
It is just what happens when the record is incomplete.
And once the discussion becomes subjective, you are fighting a different battle.
The structural problem is not unique to one company. Research on invisible labor in open source ecosystems shows that essential work often sits outside the usual measurement systems, which means it can be undercounted even when it keeps the whole system functioning.
That also fits broader work on network visibility: when work is relational, coordination-heavy, or infrastructural, it is easier to miss unless someone deliberately captures it.
What to do instead
The answer is not to work harder.
The answer is to leave a better trail.
You need a system that captures impact before it fades. That means tracking your work while it is still fresh enough to describe accurately.
At a minimum, every meaningful win should record:
- What changed
- How you know it changed
- What you did to cause it
- Who benefited
- Why it mattered to the business or team
That structure matters because it turns “I helped with X” into evidence that can survive a review cycle.
Here is the difference:
- “Improved search”
- “Reduced search latency 65% by identifying cache bottlenecks and implementing targeted indexing changes, improving retention 12%”
The second version has memory resistance.
It gives the reviewer a story, a metric, and a mechanism.
That is what survives the fade.
The hidden work deserves a record too
This is especially important for glue work.
Glue work is real work.
It is also the easiest kind of work to forget because it often does not leave an obvious artifact.
If you helped unblock another team, did a deep design review, mentored someone through a technical transition, or held the social and technical pieces of a project together, write it down.
Do not wait for somebody else to notice it.
If you do, it becomes much easier for the organization to classify that effort as “nice to have” instead of “career-moving.”
That is one of the more frustrating parts of engineering careers: the work that keeps systems and teams healthy can be the work that gets the least recognition.
A practical system
If you want to beat memory fade, keep a running log with three levels.
1. Weekly notes
Capture raw wins while they are fresh.
This can be simple:
- incidents handled
- decisions made
- metrics moved
- people unblocked
- docs written
- risks avoided
You are not writing the final version yet. You are collecting evidence.
2. Monthly synthesis
Once a month, turn those notes into impact statements.
Ask:
- What was the business effect?
- What would have happened otherwise?
- What did I do that was not already obvious?
This is where vague activity becomes promotion-ready material.
3. Quarterly packet
Before review season, convert the best items into a clean narrative.
At that point you should be able to answer:
- What are my top 3 to 5 wins?
- What themes connect them?
- What level of scope do they prove?
- What evidence backs each claim?
If you cannot answer those questions quickly, your packet is probably too thin.
The real point
This is not about self-promotion for its own sake.
It is about making your work legible.
If you are doing good engineering, but nobody can reconstruct the impact later, then the system is not actually seeing your contribution. That is not a talent problem. It is a record-keeping problem.
And record-keeping is solvable.
The engineers who grow consistently are not always the loudest. They are usually the ones who build a durable trail of evidence and keep that trail current.
That is what protects you from memory fade.
Not hope.
Not charisma.
Not a manager’s recollection three months later.
Evidence.
Bottom line
If you want your work to count when it matters, you have to preserve it before it disappears.
Track the work.
Translate the work.
Tie the work to outcomes.
Do that consistently, and review season stops being a guessing game.
Good engineering is not enough if the organization cannot remember it.
I'm building careercraft.ing in public. It is an MCP-powered vault that captures your engineering wins as they happen and turns them into promotion-ready bullets. If you have ever lost a win to memory fade, follow along for the next draft and the product updates around it.
Top comments (0)