How this instrument works
Week-over-week growth measures how far a number moved between two adjacent seven-day windows, expressed as a percentage of where it started. The formula subtracts last week's value from this week's, then divides by last week's — anchoring to the earlier window is what turns a raw difference into a rate comparable across metrics of any size, whether the underlying number is 400 signups or 4 million dollars of orders.
Growth and marketing teams at early-stage products lean on this reading because it is the fastest cadence available: a seven-day window closes long before a calendar month does, so a launch, a price change or an ad campaign shows its effect here first. It differs from month-over-month, which averages an anomaly across roughly thirty days and so reacts more slowly but steadier, and from year-over-year, which compares the same calendar week twelve months apart and cancels seasonal and day-of-week effects entirely at the cost of reporting the news a year late.
The trade for that speed is noise. A seven-day sample is small enough that one unusual day — a viral mention, a site outage, a single large order, a payday landing inside the window — can swing the whole week's total on its own, without any real shift in the underlying trend. A common misreading is comparing a current week still in progress against a full, completed prior week; the partial total mechanically understates growth because it has not had seven days to accumulate.
- Enter the current seven-day total in This week's value — revenue, signups, sessions, or any metric tracked daily.
- Enter the same metric for the seven days before that in Last week's value.
- Read the result in Week-over-week growth, % — a positive number means the metric rose, a negative one means it fell.
- Recompute at the same point in the cycle each time (say, every Monday morning) so consecutive readings compare weeks of equal length.
- Watch several readings in a row rather than reacting to one, since any single week can be skewed by an unusual day.
Worked example — a metric climbs from 10,000 to 11,000
Take a product tracking weekly active users: last week closed at 10,000 and this week closed at 11,000. Feeding lastWeek = 10,000 and thisWeek = 11,000 into the formula gives (11,000 − 10,000) ⁄ 10,000 × 100 = 10.0%, a swing this instrument would surface days before a monthly report ever confirms it.
That single 10.0% reading is worth checking against the week before it, since week-over-week is the noisiest cadence in common use. If the prior week also read close to 10%, the product is compounding steadily; if it read closer to 2%, this week's jump may be one unusual day — a press mention, a competitor's outage, a marketing push — rather than a genuine change in direction.
Questions
Why does week-over-week swing so much more than monthly figures?
A week is a small sample — seven days — so one unusual day, such as a viral mention, a site outage or a single large order, can move the whole total on its own. A month averages that same anomaly across roughly thirty days, diluting its effect. Week-over-week is the fastest signal available, but it trades stability for speed, so treat one reading as a lead rather than a verdict.
What if this week isn't finished yet?
Comparing a partial week already in progress against a full prior week mechanically understates growth, because the numerator has not had seven days to accumulate. Either wait until the week closes before reading the percentage, or compare an equal number of days from each week — Monday-through-Wednesday against Monday-through-Wednesday, for instance.
Why did my result come out negative?
A negative result means this week's value is smaller than last week's — the subtraction runs negative before the division happens. A metric slipping from 10,000 to 8,000, for example, returns exactly −20%. The sign is informative rather than an error; report it alongside the size of the move instead of dropping it.
What happens if last week's value is zero?
The formula is undefined at zero, since division by zero has no result, so this instrument requires last week's value to be greater than zero. If the metric genuinely started from nothing, such as a brand-new product's launch week, report the raw increase directly instead of a percentage, which cannot express growth from an empty base.
How is this different from month-over-month or year-over-year growth?
Week-over-week compares two adjacent seven-day windows, the fastest cadence in common use, catching a shift within days rather than waiting a month or a year. Month-over-month smooths some of that single-day noise over a longer window; year-over-year compares the same calendar week twelve months apart, cancelling seasonal and day-of-week effects but reporting the news slowest of the three.
Can this track any metric, or only revenue?
Yes — the formula only needs the same metric measured at two points seven days apart. Weekly active users, orders, support tickets, ad spend and site sessions all work; enter whichever number is tracked as This week's value and Last week's value. What must stay fixed between the two entries is the definition of what is being counted.
References
- U.S. Small Business Administration — Manage your business finances
- Federal Reserve — Statistical releases and historical data
Read this first: This instrument shows arithmetic, not advice. Real offers add fees, taxes and terms that vary by lender and place — verify the figures against your actual paperwork before deciding anything.