The January past-performance backfill ritual
Two hours in mid-January to refresh last year's past-performance references so they actually carry the weight they need to for this year's bids. The shortlist, the edits, the re-approvals.
Every January, the proposal team blocks two hours on a Saturday and does the past-performance backfill. Nobody writes new content. Nobody responds to a bid. The only work is confirming that last year’s reference engagements still read correctly as evidence for this year’s bids.
The ritual came out of a specific failure. A bid last March cited a customer engagement from 2024 in the present tense — “we operate a managed-SOC practice for [customer]” — when the engagement had actually wrapped in Q4 of the prior year. A reviewer on the buyer side flagged it on a reference call. The deal survived, but it survived on a call where the proposal team had to walk back the framing. That was avoidable.
The shortlist
Every past-performance block tagged with a calendar-year reference gets pulled into a list. At a mid-sized shop, that’s usually 30 to 80 blocks. Each one has three questions asked of it, in order:
- Is the engagement still live, or has it ended? If it ended, past tense. “We delivered” replaces “we deliver.” The end date goes in the block.
- Do the scale numbers still apply? An engagement described as “18 seats” that has grown to 40 needs the new number, or the block needs to be split into “2024 footprint” and “2025 footprint” blocks the retrieval layer can rank against the year of the reviewing RFP.
- Is the contact person still the right reference? The named point of contact on every past-performance block gets a quick confirm. They left the buyer’s company? The reference is retired, or transferred to the internal team lead who can speak to the work on their own authority.
The edits
Most blocks get a one-line update and a re-approval. Some get a tense shift and a date-stamp. A few get retired outright — the engagement ended badly, the relationship cooled, the customer was acquired. Those don’t come out of the KB; they get a no-reference-available flag so retrieval doesn’t surface them as citable evidence in this year’s responses.
The freshness-score gets a manual boost after the backfill. A reviewer-confirmed block in January 2026 is more trustworthy than a block whose last touch was the auto-ingest date from some forgotten past bid. The system is structured to treat the manual boost as signal; the retrieval layer will weight January-reviewed blocks above stale ones even when the semantic match is slightly worse.
Why it’s worth the Saturday
Two hours in January buys a year of cleaner citations. The writers who pull from the past-performance pool during the year never surface a block that says “we operate” for an engagement that ended 14 months ago. The reviewers never catch the tense mismatch. The reference calls don’t surprise anyone. Shelf made the same case for knowledge bases broadly: stale content “actively undermines user trust.” Past-performance sits at the sharp end of that.
The ritual is on the team calendar through 2028. It is the cheapest quality move we make all year.