Problem: a client project has gone quiet, the deadline is slipping, and you have three theories about why. Solution: stop theorising and sort what you actually know. Freelancers lose more billable time to unchecked assumptions than to hard technical problems, so this is a diagnosis worth doing on paper. Below is a blunt, repeatable way to run it for remote DevOps work.
The problem in plain terms
Remote work hides information. You cannot glance across the room to see whether the client’s engineer is busy, whether a ticket has been opened or whether the staging server was rebuilt. So you fill the gaps with stories. “They are probably waiting on legal.” “The access request is surely approved by now.” “The pipeline is failing because their repo changed.” Each story feels like background knowledge, and soon you are making decisions with it.
The cost shows up later. You build against an environment that does not exist, or you chase a failing build for two days when the credential was simply revoked. The work is not hard; the starting facts were wrong.
Diagnosis step one: write the stall as a fact
State what is observably true and nothing else. “No reply to my last two messages in five days” is a fact. “The client has lost interest” is a guess. “The deploy job has failed since Tuesday with the same error line” is a fact. “Their team broke the config” is not.
Keep it to three or four lines. If you need a paragraph to describe the facts, you are smuggling opinions in.
Diagnosis step two: audit your assumptions
List every belief your current plan depends on. Typical ones for a freelancer: the scope in the last message is the agreed scope, the access I was given still works, the person I am emailing is the one who decides, the target environment matches what I tested on. Mark each as “seen” or “assumed”. Be uncomfortable about how many land in the second group.
Then do the cheap tests immediately. Log in and see whether access still works. Re-read the last written scope. Ask one short question about who signs off. A beginner-friendly overview such as common questions that new DevOps freelancers ask when getting started covers many of the basics worth checking in this way, and returning to it after a bad week can be sobering.
Diagnosis step three: park the hunches
Whatever is left is hunch territory: motives, internal politics, why a reviewer is slow. Write them down so they stop circulating in your head, and then do not act on them. You can act on a test result or a written statement. You cannot act on a mood.
One rule helps a lot: no hunch is allowed to appear in a message to the client. If you would write “I get the feeling you are not happy”, turn it into a question about something checkable, such as “Is the delivery date in the plan still correct?”
The fix: choose the smallest action that tests the biggest unknown
Rank the unchecked items by how much damage they would do if wrong. Access and scope usually top the list. Pick the one that costs least to test and do it first. This is not about being cautious for its own sake; it is about spending the first hour on whatever could waste the next fifty.
Send a short, specific message, with the facts stated as facts and the open points as questions. For example: “The deploy has failed since Tuesday with the same permissions error. I am assuming the service account was changed; can you confirm?” That message separates seen from assumed in two sentences, and it gives the other side an easy answer.
Where this saves real money
- Quotes. Price work on facts about the environment, and write your assumptions into the proposal as stated conditions.
- Handovers. List what you verified in the client’s setup, so a later dispute is about records, not memories.
- Incidents. During an outage on a client’s system, label hypotheses as hypotheses before you run anything with side effects.
- Referrals. A calm, structured update is what makes people recommend you. Building a reliable support network and finding mentors is one place that reputation can start.
A worked example
A freelancer is asked to fix slow releases for a small e-commerce client. Seen: the last six builds took roughly the same time, and most of it is spent in one test stage. Assumed: the test stage is slow because of the database fixtures; the client’s hosting is unchanged; the client wants speed more than coverage. Hunch: the previous contractor left a hidden delay.
The freelancer tests the second assumption first by asking one question about what must not be skipped. The answer changes the plan completely, since two of the slow tests protect the payment flow. Without that question, the freelancer would have removed them and called it a win.
When the client pushes back
Sometimes a client reads your labelled message as hedging. Hold the line politely: facts are stated flatly, and unknowns are phrased as the next thing you need. Most clients relax once they see that the questions are specific and that each one unblocks something. If they cannot answer, that is also a fact worth recording, along with the date you asked.
Keep learning without the guesswork
It helps to practise sorting claims on material where no client is waiting. One low-pressure option is the DevOpsSchool site, which presents its tutorials, slides and cheatsheets alongside courses and certification tracks for engineers, so you can rehearse the sorting habit on study material instead of on a paying client.
The checklist to stick on your monitor
- Write the stall as two or three observable facts.
- List the assumptions behind your current plan; mark each seen or assumed.
- Test the cheapest, riskiest assumption first.
- Park hunches; never send them as statements.
- Send one short message with facts as facts and unknowns as questions.
It is not glamorous, and it works. The freelancers who get paid on time are mostly the ones who know, at any moment, which of their beliefs they have checked.