Proposal evaluation

This module is part of Tenders, which is enabled per organization.

When offer reception closes, the tender moves to In evaluation. Sourced combines two complementary engines to decide better: the human evaluation of the technical evaluators and the AI evaluation. Both feed the scorecard on which you award.

Technical evaluators

They are experts, internal or external, who score the technical part of the proposals. They work like this:

• Access without an account: the buyer invites them and they receive a link by email; they get in by verifying their identity with a one-time code, with no need to create a user.

• They only see the technical side: each criterion has a share with the technical evaluator toggle. The evaluator sees only the shared criteria and the technical documentation; the commercial data stays hidden (prices, payment terms, commercial notes) so as not to bias their judgment.

• They score by criterion: they assign a grade from 0 to 100 to each criterion assigned to them, with a confidence level, a justification and the risks they detect.

• With AI assistance: they receive AI suggestions scoped to their criteria, which they can take or ignore.

• Technical questions: bidders' technical queries can be routed to them; they put together a draft answer (with AI help) that the buyer publishes as is with Approve & publish, edits before publishing, or returns with feedback. The answer is credited to the evaluator, approved by the buyer.

• Deadlines: the system doesn't send them automatic reminders; you do the follow-up. The tenders list shows how many evaluators have finished and, when the whole committee finishes, you get a notice (see "Notice when the technical committee finishes" below).

The buyer manages the evaluator pool, which criteria they share and who they assign each one to. More detail in Technical evaluators.

AI evaluation

The buyer triggers the AI evaluation with a button on the Comparison tab. Sourced's AI evaluator reads the specification and all the proposals (including the attached documents: PDF, Excel, Word, etc.) and produces a complete analysis. It works in three passes:

1. Technical analysis: it compares each proposal against the specification (not against the others), item by item. It detects specification compliance, missing information, technical risks and strengths.

2. Commercial analysis: it reviews prices, cost structure, hidden costs, payment terms and competitiveness, and identifies negotiation levers. It flags items well above market and suspiciously low prices.

3. Synthesis and decision: it combines the technical and the commercial into a recommendation per supplier.

The result includes, for each supplier:

• Scores by criterion (weighted according to the tender's weights) and a total score.

• A suggested decision: AWARD, NEGOTIATE or DISCARD, with its rationale.

• Technical and commercial summary, strengths and weaknesses.

And at the tender level:

• Ranking of suppliers and an executive comparative analysis.

• Red flags with severity (critical / warning / informational).

• Negotiation strategy: rounds, concrete tactics and expected savings.

• Compliance check: certificates and annexes submitted, expirations and missing items per supplier.

The AI evaluator cites the evidence for each statement, indicates its confidence level and, if it does not have enough data, does not invent a score (it leaves it empty and reports it as missing). The analysis can be downloaded as a PDF.

The scorecard

The scorecard brings everything together on the Comparison tab: the ranking of offers, a radar chart of scores, the AI insights panel and the evaluators' scores (with an average per criterion and each one's confidence). The criteria and their weights are the ones you defined when creating the tender; the scores come both from the AI and from the evaluators. When a technical evaluator submits their scores, those scores replace the AI's on the criteria they evaluated: the average of the evaluators per criterion is taken and the weighted total is recalculated, so the ranking, the radar chart and what the approver sees stay consistent with each other. The scorecard marks with an icon the scores that come from evaluators, and hovering shows whose they are. If the buyer later corrects a score by hand, theirs prevails; and if the last evaluator score on a criterion is cleared, the one the AI had calculated comes back.

Awarding

With the scorecard in view, you choose the winning supplier and award. The winning proposal is accepted, the others rejected, the tender moves to Awarded and the purchase order is generated. If the negotiation warrants it, you can open a new round before deciding.

If the proposal changes

If the supplier corrects a proposal they already submitted within the same round (or the proposal goes back to draft because you changed the items), the committee's scores are kept but flagged Review: the evaluator sees them pre-filled and confirms or corrects them. Your manual edits and the AI evaluation of that proposal are discarded, since they were based on the previous version.

In a new round, the committee's scores on the supplier's latest proposal carry over to the new one, and how they arrive depends on what changed:

• Only prices or terms (the usual negotiation): they arrive as they were. Confirmed scores stay confirmed and don't count as pending; the evaluator sees them as saved, with the note "Score from round N", and doesn't have to score again.
• Something technical (the technical notes, the delivery lead time of the proposal or of any item, an item's specification compliance or notes, or any non-commercial document): they all arrive flagged Review, so the evaluator confirms or corrects them.

A score that was already flagged Review stays that way in the new round, and scores from an evaluator whose access you revoked don't carry over. The new round's proposal starts without an AI evaluation or edits of yours.

Notice when the technical committee finishes

You don't need to open each tender to know whether the committee has finished:

• In the Tenders list, tenders under evaluation that have technical evaluators and proposals to score show a tag with the committee's progress (for example, 2/3): gray while evaluators are still missing, blue once they have all finished ("Technical evaluation complete: ready to award"). Hovering shows the detail, and clicking takes you to the Evaluators tab (if someone is still missing) or to Comparison (if they have all finished).
• When an evaluator finishes their part and the committee hasn't yet: you only get a bell notification.
• When the whole committee finishes: a high-priority bell notification and an email, once per round. In a new round you get another one if the evaluators have to score again and finish.

An evaluator has "finished" when they have scored all their criteria on every proposal in play in the current round, with no scores pending review. A round where suppliers only changed prices or terms doesn't put the committee back to pending; a new supplier or a proposal with technical changes does.

The notices go to whoever created the tender and are only triggered by evaluators' actions: if you complete the committee yourself (by revoking a pending evaluator or narrowing their criteria), no notice is sent. To silence them, go to Settings → Notifications and turn off the Tenders type (email, in-app or both); that type also covers the notices about proposals received, supplier questions and messages, and the AI evaluation.

Technical view: compliance per item

On the Comparison tab, the per-item price table has two views: Commercial and Technical. In the Technical view, each cell shows whether the proposal meets what was requested for that item according to the AI's technical analysis: Meets, Partial, Does not meet or No data, with the detail and, on hover, the AI's observation if there is one. The data appears once the AI has evaluated the proposal (Fill in with AI); if it hasn't evaluated it yet, or couldn't tell, the cell says No data.

The same per-item compliance appears in the technical table of the AI analysis PDF included in the tender package. It is only for the buying team: technical evaluators and suppliers don't see it.