Rubric deep-dive: past-work review (Pillar 02).
20% of the composite. Screenshots and slides are not evidence. We grade shipped work you can open, outcomes the team can prove they caused, and how long clients stay. What we look for per discipline.
Past-work is where most agencies think they will shine and where most get their lowest score. Twenty percent of the composite. Judged by one senior engineer and one liaison, working from a case-study dossier, an engagement timeline, and outcome scoring with sources. Screenshots and slides are not evidence.
The bar is verifiable, not impressive
Impressive is a bar we do not grade against. Verifiable is. The question we ask on every case study. Can two reviewers, working independently, confirm the team did what they say they did, and that the outcome they claim is real and attributable? If both land on "yes" without conditions, the case scores. If either lands on "maybe," it does not.
That framing kills a lot of the mockup-and-slide portfolios that offshore agencies default to. It also kills the marketing case studies that quote ROAS improvements with no baseline. It also kills the design portfolios that show hero screens from projects the team briefly consulted on but did not ship.
What counts as shipped, by discipline
Software. Live URLs the team owned, repositories the team wrote to, or recorded walk-throughs where the team's contributions are named. Not "we consulted on the architecture." Not "we built parts of this that are no longer visible." Real, running artifacts with clear scope.
Marketing. Campaigns or programs with brief, creative, and reporting access. Reporting access is the hardest to satisfy. Most agencies show platform dashboards (ROAS, CTR, impressions) rather than reconciled business metrics (pipeline, revenue, CAC). We accept both, but a case with only platform metrics scores lower than one with CRM reconciliation.
Design. A product surface, a brand system, or a research report, with source files. Design portfolios without source files score lower than ones with. Source files show the version history and the collaboration record.
Product and strategy. Engagements with the actual artifacts the client received (PRDs, market-entry memos, ops playbooks) plus a decision-to-outcome trail. What was recommended, what shipped, what changed.
Outcomes: named, sourced, tied to your scope
Twenty-five percent of the pillar. Three tests. Named (what specific metric moved). Sourced (where does the number come from and can it be verified). Tied to the team's scope (was the team actually working on the thing that changed). If any test fails, the score drops.
The most common miss is the third. A marketing agency shows a 40% increase in signups over a quarter, but the client also launched a new pricing page during the same quarter, and the marketing team was not involved in the pricing page. The 40% might be real, but the team's contribution to it is not isolable. That reduces the outcome to a coincidence, not evidence.
Retention is the loudest signal
Twenty percent of the pillar, and probably the highest-signal criterion in the entire rubric. How long do clients stay, and why. Renewals and scope expansions count more than headline logos. A team that has kept the same three clients for two years is more credible than a team that has cycled through fifteen logos in the same window.
We look at the timeline for the last five engagements. Median engagement longer than nine months is the bar. Two renewals across the last five is the substance. Cluster patterns (clients who all left around the same time, or all left when the same team lead moved on) are read carefully. A single departure is noise. Three around the same event is a signal that something happened.
Scope discipline is scored on the ability to say no
Fifteen percent of the pillar. The test is simple. Can the team cite one engagement where they pushed back on a client request and the relationship survived. If they cannot, they have either never pushed back or the relationship did not survive. Both are red flags.
Scope discipline shows up in specific behaviors. A refused request in writing, with reasoning. A re-priced SOW after a scope change. A client who tried to expand scope without a new SOW and was walked back with a written note. These are not rare. Almost every team has one. What is rare is the team that has documented one.
IP and handoff: the ending-well test
Ten percent of the pillar. Deliverables, documentation, and access transfer cleanly at engagement end. A non-author must be able to stand the system up from the handoff alone.
Two red flags we watch for hardest. Credentials, files, or accounts withheld at engagement end to pressure renewal (a serious escalation, always disqualifying). No documented ending checklist at all (fixable, but flagged).
- 01The bar is verifiable, not impressive. Two independent reviewers must be able to confirm the work.
- 02Outcomes must be named, sourced, and isolable to the team's scope. Platform metrics without CRM reconciliation score lower.
- 03Retention is the loudest signal. Median engagement over nine months and two renewals across the last five is the bar.
- 04Scope discipline is scored on one documented pushback that survived the relationship.
- 05Handoff hygiene tests whether a non-author can stand the system up alone. Withheld access at engagement end is disqualifying.
Where does your shop sit on this pillar?
The self-assessment mirrors the pillar weights and gives you a private draft score before you commit to the full intake.