Choosing partners with weighted criteria

A partner scorecard compares candidates on the same criteria, so the choice rests on reasons rather than on who made the best impression. It only works if two decisions come before any scoring: which way each criterion points (is a high score good or bad?) and how much each criterion matters.

Skip them and the total lies. Teams scored three candidates 1–5 on fit, reach, trust, conflict risk and onboarding effort. The training agency got 5, 4, 3, 5, 3 and the logistics association 5, 5, 5, 1, 4: both total 20. But a 5 on conflict risk means a lot of conflict, and a 4 on effort means a lot of work. The plain sum rewarded the agency for being risky and the association for being hard to onboard.

A finer scale, fewer criteria or letting the founder break the tie don't fix this. Fixing direction and weights does.

Some criteria are things you want more of: fit with your ideal client, reach, trust. Others are things you want less of: conflict risk with your own sales, onboarding effort, setup cost. Before adding anything, turn every criterion into "higher is better". On a 1–5 scale that means using 6 − score for the lower-is-better ones: a conflict risk of 5 becomes 1, an effort of 4 becomes 2.

The agency's 5 on conflict is real: it has its own English trainers and sells to the same HR managers as Teams. Flipped, it is a 1. With equal weights the agency's scores 5, 4, 3, 1, 3 average 3.2, and the association's 5, 5, 5, 5, 2 average 4.4. The tie is gone.

Write the direction next to each criterion on the sheet. Otherwise the next person who updates it will add the raw scores again.

Weights say how much each criterion matters, and they should come from the gap the partner is meant to fill. Teams needs reach and trust with companies that don't know it, and partners must bring clients who fit. Marta's weights follow from that: fit with the ideal client 0.3, reach 0.2, trust 0.2, conflict risk 0.2, onboarding effort 0.1. They add up to 1, so a weighted score stays on the same 1–5 scale.

The timing matters as much as the numbers. Set the weights before looking at the candidates' scores. Once you know who you like, it is tempting to nudge a weight here and there until your favourite wins, and the scorecard just dresses up a gut decision.

If the weights don't add up to 1, totals stop being comparable with the 1–5 scale; if they were tuned after seeing the scores, nobody else can trust them.

A weighted score multiplies each (flipped) score by its weight and adds the results.

The association's scores are fit 5, reach 5, trust 5, conflict 1, effort 4. Flip conflict to 6 − 1 = 5 and effort to 6 − 4 = 2. Then:

0.3 × 5 + 0.2 × 5 + 0.2 × 5 + 0.2 × 5 + 0.1 × 2 = 1.5 + 1.0 + 1.0 + 1.0 + 0.2 = 4.7.

The HR consultancy (fit 4, reach 3, trust 4, conflict 1, effort 2) comes out at 4.0 the same way, and the agency at 3.4.

Three common slips: forgetting to flip (the HR consultancy would get 3.0), adding raw scores (14 for the consultancy, a different scale altogether), and flipping the wrong criteria. Do the flip first, write each product down, then add.

With direction and weights in place, the ranking is association 4.7, HR consultancy 4.0, training agency 3.4. Leo's objection is natural: the agency has the biggest sales team of the three. But the scorecard already counts reach, and the same sales team would sell the agency's own English courses alongside Klasna's. Its conflict risk and middling trust pull it down more than its reach lifts it.

A low rank is not a verdict of "never". It says that, on current terms, the agency brings more risk than value. Terms can change that: if the agency drops competing courses, or both sides agree clear rules on who sells to which client, its conflict score falls and the rank moves.

So report a low-ranked candidate fairly: the score, the one reason that moved it down, and what would change the picture. That keeps the door open and the decision honest.

A score without evidence is an opinion with decimals. Next to each score, write the fact it rests on, so anyone can check or challenge it.

• Reach: how many of the companies the partner works with match Teams' ideal client profile.
• Trust: buyers who acted on the partner's recommendation. The association's 5 rests on Pavlo's account: Veloria bought its safety course because the association recommended the provider.
• Conflict risk: whether the partner sells something like Klasna's to the same buyers.
• Onboarding effort: what the partner needs to start: training for its people, materials, joint events.

Counts and facts are enough. You don't need a partner's client list or anyone's personal data to score it, and you shouldn't ask for them: collecting other people's data has to have a purpose and a legal basis, and scoring a partner needs neither.

A scorecard ranks fit and risk; it doesn't say what a partner earns. For the top names, estimate the money: deals a year × lessons per deal × what Teams keeps per lesson − the cost of setting the partner up. A referral partner leaves 259 UAH a lesson, a reseller 179, so the same deals are worth different amounts in different models. A partner that brought two deals of 50 lessons a month on a referral would add 2 × 600 × 259 = 310,800 UAH a year before setup.

Then test how stable the ranking is. Move a weight by a small step, say 0.05 from one criterion to another. If the winner changes, the two candidates are close, and the scorecard hasn't settled anything: the evidence behind the disputed scores, and the money, should decide.

When the ranking and the money disagree, don't pick one blindly. Find out why: a high-scoring partner may bring few deals; a profitable one may carry a risk the scores flag.

The order of the steps is what keeps a scorecard honest.

1. Choose the criteria from the gap. For Teams: fit, reach, trust, conflict risk, onboarding effort.
2. Set each criterion's direction and weight, before any candidate is scored. This is what stops weights from bending towards a favourite.
3. Score each candidate with evidence next to every score.
4. Compute the weighted totals and rank.
5. Check the leaders' economics and whether small weight changes flip the winner.

Done this way, the three candidates come out as association 4.7, HR consultancy 4.0, agency 3.4, and you can explain each number in plain words.

How it fits together: the previous lesson named the gap and the model; this one turns the gap into criteria and weights and ranks real candidates. The next lesson sets the terms you sign with a partner, and the lines competition law doesn't let you cross.