Stop paying for stale documentation
Built from your team's actual size, region, and AI adoption, with multipliers backed by research from Atlassian, GitLab, and Stripe — plus a PDF report you can share with your team
How we calculated this
eng_drift = eng_count × annual_salary × rate pm_count = max(1, round(eng_count / 8)) pm_doc = pm_count × hours/week × 47 weeks × pm_hourly_rate total = eng_drift + pm_doc — Where — rate = 5% low / 10% medium / 20% high (AI adoption) hours/week = 4 low / 8 medium / 12 high (AI adoption) — With Specsight — eng_recovered = eng_drift × 70% pm_recovered = pm_doc × 90% specsight = $129 × 12 = $1,548 / year (Standard, flat) net_savings = (eng_recovered + pm_recovered) − specsight
Every multiplier above is grounded in customer research, layered with public data where available.
- Up to 20% of engineering time goes to interpreting unclear specs and re-deriving feature behaviour from code. The actual rate varies by AI adoption: 5% low / 10% medium / 20% high. Aligned with public stats — Stripe 2018: developers spend 42% of time on maintenance. GitLab 2024: 7 hours/week lost to inefficient processes including poor knowledge sharing.
- 4 to 12 hours per week per PM goes to maintaining docs that go stale anyway. Rate varies by AI adoption: 4h low / 8h medium / 12h high (higher AI throughput = more behaviour changes for PMs to chase). Atlassian 2025: teams waste 25% of the workweek searching for answers, citing insufficient documentation.
- Roughly 1 PM per 8 engineers is typical across SaaS organisations from Series A through mid-market. We derive your PM count from this ratio (rounded up) and show it openly in the methodology.
- AI adoption drives both eng interpretation time and PM doc maintenance hours within the bounds above. Higher AI = more PRs per week = more behaviour change to chase = closer to the upper bound of each range. This calibration is from customer conversations; not yet backed by a cited public study, and we are honest about that.
- Typical recovery falls in a 60–80% range on engineering interpretation time and 80–95% on PM doc maintenance. The figures here use the midpoints (70% / 90%) so the headline number stays single-valued; the methodology states the range honestly.
- 47 working weeks/year = 52 minus 4 weeks PTO and 1 holiday week. 40 hours/week. 1,880 productive hours/year.
Common questions
It's an honest model, not a precise forecast. The multipliers (20% eng interpretation, 8h/week PM maintenance, 70%/90% recovery) come from Specsight's research with product teams plus published data from Atlassian, GitLab, and Stripe. Real teams sit in a 60–80% / 80–95% range on recovery; we use the midpoints here so the page shows a single dramatic figure. Open the methodology section below for the full sourcing.
Documentation drift is a cross-functional cost. Engineering loses time interpreting unclear specs; PMs lose time maintaining the specs themselves. Specsight reduces both — we are the documentation, so PM maintenance time effectively goes away. We derive your PM count from a typical SaaS ratio (1 PM per 8 engineers) and show it openly in the breakdown so the math is transparent.
AI-coded changes ship faster than humans can update docs — more PRs per week, more behaviour changes, more drift volume. We model this as a 1.0× / 1.15× / 1.3× multiplier on the engineering interpretation cost. This is Specsight's hypothesis; we're honest that there's no cited public study yet backing the exact multipliers. If you have AI tools but disable them mid-flow, pick low.
Four pages: your numbers on the cover, the breakdown table, what the cost means in human terms ('that's 5 senior hires of time per year'), and three drift-reduction tactics tailored to your team's AI adoption level. People who run our calculator forward the PDF to their CTO or board — it's formatted to make the budget case for you.
Stop paying the drift tax
Specsight reads your repository and gives your team a spec that stays accurate with every release. Free to try, no card