At Generation, job attainment — the share of our graduates who are employed within 90 to 180 days after completing our program — is one of our most-watched metrics. It’s in our funder reports, our public commitments, our board materials.

We recently changed how we calculate it, and the new numbers are slightly lower. Here’s the simple version: imagine a graduate we lose touch with after their program ends. Under our old formula, that person disappeared from the math entirely: not counted as employed, not counted as unemployed, just gone. It’s standard practice across employment measurement to count only who you can reach — governments, researchers, and workforce organizations alike build their numbers this way.

However, under the new formula, a graduate’s last known status carries forward to the 90- or 180-day mark, even without a fresh check-in. Only graduates with no post-graduation status at all, those that we haven’t reached even once after they left the program, are excluded.

The result is a more complete picture — and also a lower number. Globally, our job attainment figures moved from roughly 81% to 77% (180-day) and 73% to 70% (90-day). We think that’s a good thing, and we want to explain why.

A different tradeoff, made deliberately

Our data completion (the portion of graduates we contact who complete our surveys) has been consistently strong: 93%+ across all 17 of our countries within the first year post-program, 73% for alumni at 1–2 years after Generation, and roughly 55% at 2–5 years post-program on average — with our largest countries reaching 60%+ data completion in the 2-5 year window. Tracking our graduates’ progress well after they’ve left our programs has always been a top priority for us and we are cost-effective in doing so (it costs us less than 1% of the total cost per learner). 

We’ve always known something important about graduate follow-up: non-response isn’t random. Graduates who haven’t found work are somewhat less likely to respond to a check-in, out of discouragement or simply having less to report. Governments face a version of the same problem in official labor statistics, which is why unemployment definitions are built around confirmed, active job-seeking rather than just “reachable or not.”

Given that, our old formula made a deliberate choice: count only graduates with current status, those who we had just reached for an update, and leave the rest out of the denominator entirely. That kept the metric clean and easy to interpret. But it also meant that we used a smaller denominator in those places where follow-up was hardest.

That was a reasonable tradeoff at the time. But as our data infrastructure and follow-up practices have matured, we decided a different tradeoff now serves us better: count every graduate whose last known status we have, and only exclude those we’ve truly never been able to reach. It’s a shift in what we optimize for, moving from a clean denominator to a more complete one. And it allows us to get deeper answers to our questions.

What good follow-up looks like 

Follow-up at Generation means reaching graduates by phone, WhatsApp, and survey reminders at the following intervals post-graduation: 90 days, 180 days, one year, and then annually up to five years. It requires managing connectivity gaps, survey fatigue, and limited resources. But we’ve seen what’s possible with a concerted effort.

For example, one of our country teams set out to improve data completion rates for their two-year post-program survey. They opened with a personal email appeal from their country CEO, then switched outreach to WhatsApp — the channel they realized graduates were most likely to use — simplified the survey itself, and followed up by phone with non-respondents.

The result: data completion rates jumped from 30% to 70%.

The lesson: data completion isn’t fixed. It’s a product of design choices about channel, message, messenger, friction, and more. That’s why we have always treated data completion as its own metric, tracked alongside job attainment. When we report our outcomes to our Board, we consistently share both data completion and job attainment so that they can see (and query) the rigor of the numbers.

Why we’re sharing this

There’s value in being willing to second-guess your own numbers — even ones that have been stable for years, even ones that are publicly reported, even when the result looks a bit lower. Methodologies that made sense at one point don’t always continue to make sense as your data infrastructure improves and you can see more than you used to. The willingness to look again, and to change course when warranted, is itself a form of rigor.

We’re grateful to our funders and partners who’ve engaged with this change as what it is: a deliberate move toward greater accuracy and transparency. 

Learn more about our impact