Tracking Cohort Progress Without Chasing Founders
Quarterly updates that arrive late and half-empty are a mechanism problem, not a discipline problem. How to ask so the answers come back usable.
The chase is the symptom
Every incubator recognises this. The quarterly update goes out. A third come back on time. Another third arrive after two reminders. The rest arrive in the week before the funder report, filled in from memory.
The usual diagnosis is that founders are busy and undisciplined. That is true and it is not the cause. The cause is that the request is expensive to answer, ambiguous, and consequence-free.
Ask for fewer things
A twenty-field quarterly return does not produce twenty fields of data. It produces four fields of data and sixteen fields of guessing, which is worse than not asking, because the guesses go into your reporting.
Five or six fields is the right size:
- Revenue or usage - one number, defined once
- Headcount - with a stated definition, including whether founders count
- Funding raised this period - amount and type, or none
- Current stage - from your own list, not free text
- Biggest blocker right now - one or two sentences
- What would help most - one or two sentences
Fields 5 and 6 are what make this worth a founder's time. Without them the update is pure extraction: they do work, you get data, they get nothing. With them the update becomes the mechanism by which they ask for help, and response rates change accordingly.
Define each field once, at intake
Ambiguity is the second largest cause of missing data. "Revenue" means recognised revenue to one founder and cash collected to another. "Team size" includes interns for one and not the other.
Write the definition next to the field, fix it at intake, and do not change it mid-year. If you must change it, change it at a year boundary and say so in the report, because an unlabelled methodology change makes your own trend meaningless.
Put the request on a schedule they knew about
Tell founders at intake exactly when updates will be asked for - the four dates, for the year. An expected request gets answered far more often than a surprise one, because it can be planned around.
Ad hoc requests, sent because a funder report is due, produce the worst data you will ever collect: reconstructed, rushed, and impossible to compare with the last set.
Attach a consequence that is not a telling-off
The most effective mechanism is tying the update to something the startup already wants:
- A tranche release requires a current update
- Continued space allocation is reviewed against it
- Mentor matching uses the "what would help" field, so skipping it means getting nothing
None of these is punitive. They are just the natural dependency: you cannot release money against progress you have no record of, and you cannot match a mentor to a problem nobody told you about.
Track who has not answered
This is the operational half, and it is where spreadsheets fail.
A returned form goes into a folder or a mailbox. The absence of a form is invisible - there is nothing to see. So chasing depends on someone holding the expected list against the received list in their head, which works for eight startups and fails at twenty-five.
What you need is dull and specific: for this collection round, who was asked, who answered, who has not, and when they were last reminded. That is the whole requirement, and having it turns chasing from an act of memory into a five-minute task.
Use their own stage definitions to see the programme
If you record stage at intake and at each update, you get the one thing most incubators cannot produce: movement.
Absolute state tells you little - a cohort that ends with eight companies at "first revenue" might have started there. Movement tells you what the programme did, and where it stopped working. If most companies stall between prototype and first revenue, that is a curriculum finding, and it is only visible if stage is recorded repeatedly against the same definitions.
Close the loop, visibly
The single cheapest way to raise response rates: show founders something back.
An anonymised cohort summary - median stage, how many raised, the most common blocker this quarter - takes an hour to produce and changes how the next request is received. It converts the update from a tax into a trade.
It also has a second benefit: the most common blocker across a cohort is usually the most useful thing you can learn about your own programme, and nobody ever reads it if the answers only ever go into a report.
The realistic summary
If updates are late and thin, do not send a firmer reminder. Change four things:
- Cut the form to six fields
- Define every field once, at intake
- Publish the schedule at intake
- Attach one real consequence
And make sure someone can see, at a glance, who has not answered. Most of the chasing problem is not the chasing; it is not knowing who to chase.
Frequently asked questions
- How often should incubators collect startup updates?
- Quarterly for most measures, with a light monthly check-in if the programme is intensive. Monthly full reporting is more than most early startups can produce honestly, and it degrades into copy-paste, which is worse than less frequent data.
- Why do founders not fill in update forms?
- Usually because the form is long, the fields are ambiguous, they cannot see why it matters, and nothing happens when they skip it. All four are fixable and none of them is about motivation.
- What should be in a quarterly update?
- Five or six fields at most: revenue or usage, headcount, funding raised, current stage, the biggest blocker, and what they want help with. The last two are the ones that make the exercise useful to the founder rather than only to you.
- Should updates be mandatory?
- Tie them to something real rather than making them a rule. If a tranche release or continued space allocation depends on a current update, compliance is high without anyone being chased. A rule with no consequence just trains people to ignore rules.
