Is There a Best Day to Send a Proposal? What Your Open Times Can and Cannot Prove
Once you can see when your documents get opened, a hypothesis forms almost immediately: Tuesday mornings are good, Friday afternoons are dead, nobody reads anything in August. It feels like data because it came from data.
Almost none of it survives contact with how many documents you actually send. This is the article about what a timestamp is worth, and it is mostly a warning — with two genuinely usable patterns at the end.
A timestamp is a fact about your clock, not theirs
The most common error is the easiest to make. You see a visit at 23:10 and you picture somebody reading your proposal in bed. You do not know that. The reader’s local time is not recorded anywhere, because a browser does not have to tell you and nothing here asks.
You can infer a timezone from country, and country is the one location field worth leaning on: DB-IP, whose database supplies these lookups, puts country-level accuracy at 95–99%. City is a different matter — 55–80%, with an error that is systematic rather than random, because addresses commonly resolve to an ISP hub and a whole region’s readers can appear in one capital.
So “late-night reader” is a claim you can make about a reader in a country you know, and not otherwise. The geolocation accuracy question has the detail.
Why your day-of-week theory is probably noise
Say you send four proposals a month. Over a quarter that is twelve documents. If seven of the twelve were first opened on a Tuesday or a Wednesday, that feels like a pattern.
Three things are wrong with it, and they compound:
- The sample is tiny. Twelve observations spread across five weekdays cannot distinguish a real effect from a coin landing the way coins do. You do not need statistics training to feel this — you need only imagine reshuffling the same twelve and seeing a different “pattern” appear.
- You caused it. If you send on Monday evenings, of course the opens cluster on Tuesday. The strongest predictor of when a document is opened is when it was sent, and it swamps everything else. Any day-of-week claim built without accounting for send day is a claim about your own calendar.
- The documents are not comparable. A proposal after a warm call and a cold follow-up to a stranger are different events. Pooling them and sorting by weekday hides the only variable that mattered.
There is no honest published figure for the best day to send a proposal in your line of work, and the ones circulating come from other people’s populations, other people’s industries, and often from email opens rather than document opens — which is a different measurement entirely.
How to test it properly, if you want to
The method is not complicated; it is just slower than reading a statistic off a chart.
- Fix everything except the thing you are testing. Same kind of document, same kind of recipient, same covering message.
- Alternate deliberately rather than sending when you happen to be ready. Alternation is what turns your sends into a test rather than a diary.
- Measure one clearly defined outcome — first open within 24 hours, say — and decide it in advance, before you see any of the numbers.
- Keep going far longer than feels necessary. Dozens of sends, not a dozen, before the result means anything.
Most people, on reading that, correctly decide it is not worth their time. That is a fine conclusion. The wrong conclusion is to skip the work and keep the belief.
The two timing patterns that are actually usable
Both work on a single document, which is why they escape the sample-size problem entirely — they are not statistical claims about a population, they are observations about one deal.
Time-to-first-open
The gap between sending and the first visit is meaningful because it is a comparison inside one relationship. A client who normally opens within the hour and has not opened in two days has told you something. The same two days from a client who always takes a week has told you nothing.
This is a per-relationship baseline, and it is worth keeping loosely in your head for the handful of people whose documents matter most.
Clustering within a single document
Several visits inside one working day, from readers who do not look like each other, is the shape of a document being circulated rather than read. That is a real event with a real response, and it is worth reading carefully — what a burst of opens means covers it, including the innocent explanations.
One quiet detail that keeps the chart honest
A day with no visits is drawn as a day with no visits — the daily series includes every calendar day in the period, including the empty ones. It matters more than it sounds: a chart that omits quiet days compresses a fortnight of silence into a line that looks busy, and silence is one of the more informative things in this dataset.
What you must not conclude
- Not “they read it at 11pm, they must be keen”. You have your clock, not theirs, unless you know the country.
- Not “Thursdays are best for us”. Not from twelve sends, and not while you send on Wednesdays.
- Not “August is dead”. Seasonal claims need years of data and a fixed pipeline; you have neither.
What a timestamp is genuinely good for is the interval afterwards — what an open with no reply means at each interval turns it into something to act on. The list of signals recorded at all is on the features page.