Describe the four steps of a weighted scoring
covers: lo-1
Answer
-
Collect the criteria that actually matter for this decision.
-
Give each a weight, with the weights summing to 100.
-
Rate every alternative per criterion on a fixed scale, say 0–10.
-
Multiply, sum, compare — the comparison is between the weighted sums.
| Criterion | Weight | Idea A | A·w | Idea B | B·w |
|---|---|---|---|---|---|
Team can build it in one year |
30 |
8 |
240 |
5 |
150 |
Somebody actually needs it |
25 |
6 |
150 |
9 |
225 |
We learn something new |
20 |
7 |
140 |
8 |
160 |
Data protection is manageable |
15 |
9 |
135 |
4 |
60 |
Testable without a real client |
10 |
8 |
80 |
6 |
60 |
Sum |
100 |
745 |
655 |
The order of the steps is the method. Weights come before ratings, because weights set after the ratings are where the desired answer gets engineered.
And the number is not the point. What the table produces is visible reasoning: somebody can now ask "why is feasibility weighted higher than novelty?", which is a question the team can answer — unlike "why did you pick A?".
Points the answer must contain:
-
Collect criteria, assign weights summing to 100, rate alternatives on a fixed scale, multiply and sum
-
The comparison is between the weighted sums
What makes a criterion useful, and what makes one useless?
covers: lo-2
Answer
A criterion is useful exactly when the alternatives differ on it. "Must be software" scores 10 for every candidate and contributes nothing but arithmetic — it feels like diligence and changes no ranking.
The second failure is vagueness. A criterion nobody can rate consistently produces ratings that record who filled in the cell, not what is true:
| Weak criterion | Better criterion |
|---|---|
"Is interesting" |
"At least three of us want to work on it in February" |
"Modern technology" |
"Uses at most one technology none of us has used before" |
"Useful" |
"There is a named person who will use it and can be asked questions" |
The right-hand column has the same property as a verifiable goal: two people can rate it and land close together, because the sentence says what to look at.
One more limit: above roughly seven criteria the individual weights get so small that the sums are noise, and the table starts to look more precise than the judgement behind it.
Points the answer must contain:
-
Useful when the alternatives actually differ on it
-
Useless when everything scores the same — it only adds arithmetic
-
Vague criteria ("interesting") should be replaced by observable ones
Two alternatives end 745 to 738. What does that mean?
covers: lo-3
Answer
That the method has found them equivalent. A gap of one percent is far smaller than the uncertainty in the inputs: any rating shifted by one point, or any weight shifted by two, would reverse it. Reading such a result as "A wins" is reading precision that was never there.
What to do is decide on something the table does not capture — who is willing to maintain it over the summer, which client answers the phone, which of the two the team actually wants — and then write down what decided it, in one sentence, in the project proposal. That sentence is worth more than the table at that point, because it is the only record of the real reason.
What not to do is go back and adjust the weights until the gap widens. That is manipulation with extra steps, and it is visible in the version history.
There is a third possibility worth checking first: a result that feels wrong usually means a criterion is missing. That feeling is information, not stubbornness — find the criterion, add it, and re-run.
Points the answer must contain:
-
The method says they are equivalent — the gap is within the noise
-
Decide on something the table does not capture and write down what it was
-
Do not pretend the number decided it
How can a weighted scoring be manipulated, and what prevents it?
covers: lo-4
Answer
Three ways, in order of frequency:
Adjusting the weights after seeing the ratings. The favourite alternative is behind, feasibility gets ten points more, and it is in front. Nobody experiences this as cheating — it feels like "realising what really matters".
Adding criteria that only the preferred option satisfies. "Uses a technology the team already knows" looks neutral and is a portrait of one candidate.
Rating with the answer in mind. The same property scores 8 for one idea and 5 for the other, and the difference lives entirely in who was filling in the row.
What prevents them is procedure, not virtue. Fix the weights before any rating is entered, and write them down at that moment. Keep the table in version control, where a change to a weight after the ratings exist is a visible diff with a date on it. And let somebody outside the team read the criteria before the ratings — an outsider spots a portrait criterion immediately.
None of this makes the result objective. It makes it traceable, which is the actual claim: the decision can be questioned at the point where it was made.
Points the answer must contain:
-
Adjusting weights after seeing the ratings
-
Adding criteria that favour the preferred option
-
Prevention: fix weights before rating, keep the table in version control where changes are visible
Why is the result written into the project proposal at all?
covers: lo-1, lo-3
Answer
Because a project choice has to be defensible to somebody who was not in the room — the supervising teacher now, and in the diploma thesis the assessment board later. "We discussed it and picked A" cannot be examined. A table of criteria, weights and ratings can: every step of it is open to a question, and each question has an answer.
The traceable part is the reasoning, not the number. Nobody will contest that 745 is more than 655; what can be contested is whether feasibility deserves 30 and novelty 20 — and that is exactly the conversation worth having, because it is about what the team values, not about arithmetic.
It also pays off inside the team. In February, when the project is hard and a different idea looks more attractive in hindsight, the table is the record of why this one was chosen. Without it the choice looks arbitrary from the inside too.
Points the answer must contain:
-
It makes the choice traceable for someone who was not in the room
-
The reasoning, not the number, is what can be questioned and defended