What your first 100 days records must let you do

No products are named here. The choice of software barely affects the first hundred days; what affects it is four fields that either exist on day one or can never be reconstructed afterwards.

So this page is about capabilities and about the specific properties that make a record readable in week nine. Take the requirements to whatever you already pay for.

These records are the raw material for testing the idea itself. Whether people stay, what they keep asking, and why they leave are the answers the first 100 days are meant to produce, and none of them can be reconstructed later.

What to take away

The four fields, and why they cannot wait

Arrival date. Turns your customers into cohorts. Without it, every retention question you ask later has to be answered across the whole population, which is dominated by whoever arrived most recently and stays flattering while you are growing.

Route. Where this person came from, recorded at the moment they arrive rather than inferred later. Inference after the fact is guesswork dressed as data.

Warm or cold. Whether this person knew you before. In the first fortnight almost everyone is warm and the field feels pointless. In week eight it is the difference between a readable result and an unreadable one.

Exit reason, in their words. Free text, never a dropdown. A dropdown collapses "too expensive because I never used it" and "too expensive because I use it daily and cannot justify it" into one label, and those are opposite findings.

Any tool that cannot hold these four should be supplemented rather than replaced. A spreadsheet beside it is fine.

What the arrival record must let you do

Filter by arrival week and show a count of who is still active at the end of each subsequent week, for that week's group alone.

That is the whole requirement, and a surprising amount of software cannot do it. If yours cannot, export weekly to a file and do it by hand: five minutes, and it is the central measurement of the period.

Disqualifying property. Any view that reports retention across all customers as a single number, with no way to split by arrival week.

What the question log must let you do

Hold every inbound question from every channel in one searchable place, with a date and a customer.

You need two readings from it. Questions per active customer per week, which should fall as comprehension improves. And a grouping by subject, so you can name the largest category rather than the loudest one.

Disqualifying property. Separate inboxes that cannot be read together. Email in one place, messages in another, and phone notes nowhere is the normal state, and it means the largest category is never visible.

What the exit record must let you do

Capture a row the moment somebody cancels, whether or not they say anything, and hold free text when they do.

It must be easy to cancel. That is a design requirement before it is a courtesy. Where billing recurs, the Federal Trade Commission enforces the Restore Online Shoppers' Confidence Act, 15 U.S.C. 8401, which requires clear disclosure, express consent, and a simple way to stop the charges. The FTC's Negative Option Rule, 16 CFR Part 425, covers negative option offers as well.

A hard cancellation path also destroys the record, because people who give up and dispute the charge tell you nothing at all.

Disqualifying property. Cancellation that requires contacting you. It inflates the retention figure and removes the only group who can explain your ceiling.

What the money record must let you do

Show, for one customer, what came in and what it cost to serve them, including the hours somebody actually spent.

Two properties matter. Recurring costs must be separable from one-time costs, because only recurring costs affect whether the thing works at volume.

Your time must be enterable at a rate, since unpaid founder hours are the largest input in this period and will not scale.

Disqualifying property. Anything that reports revenue without the cost of delivery beside it. In a manually delivered service that figure is close to meaningless. The identity these two figures sit inside, and the reason a recovery period matters more than a ratio early on, is set out in the market metrics.

The claim register, which is not software

A list of every factual statement on your live pages, with the evidence and the date checked. Claims accumulate fastest in the first weeks and get reviewed least, and the standard is that a reasonable basis exists before a claim runs.

If you start quoting customers, the register needs a column for the connection between you and the person quoted, because the Federal Trade Commission's Endorsement Guides, 16 CFR Part 255, require an endorsement to reflect the endorser's real experience and any material connection to be disclosed. A shared document is the right shape for this. Nothing is gained by putting it in a system.

Jobs no tool does

It does not decide which question this week is for. It does not tell you whether a small count means anything, which is a judgment covered in the discovery counts. It does not talk to the people who left. And it does not produce reach, which is the constraint most ventures actually hit in this period.

The specific risk is the week spent configuring analytics for nineteen customers. At that size everything worth knowing fits on one screen and most of it came from conversations. Time spent on instrumentation beyond the four fields is time not spent on the two uncomfortable jobs, which is a pattern founder readiness treats as a warning sign.

A setup that takes an afternoon

Whatever handles sign-up captures arrival date, route, and warm or cold. Whatever handles billing captures the cancellation row. One text file holds every conversation, appended, never overwritten. One spreadsheet holds the weekly counts with denominators.

That is enough to answer every question in the first hundred days, and it will outlast any tool you choose.

Common questions

Why are no products named?

Because a named list ages badly and because the capability list is what determines whether you can answer anything. Every requirement here can be checked against software you already have.

Is analytics software worth it this early?

For the purchase path, yes, because you need to see where people stop. For behavior inside the product, usually not: you cannot know which action predicts staying until enough people have stayed.

What about a customer relationship system?

Useful once you are talking to more people than you can hold in your head. Before that it mostly adds fields nobody fills in.

How do I record warm and cold honestly?

Ask at sign-up in one question, and record your own answer too. If you can name the person, they are warm, however they arrived.

What if I already launched without the four fields?

Add them now, and mark the date you added them. Everything before that date is one undifferentiated cohort, and saying so in the record is better than pretending the split exists.