Weekly reporting sources and the phantom spike
Some music data only moves once a week. Charted daily it produces a spike every seventh day that looks like a marketing win. How to detect the pattern and what to do about it.
Danny Starr
Co-founder, Backline · 18 July 2026 · 3 min read
In short
- Several music data sources advance their cumulative totals roughly once a week rather than daily, in a batch.
- Computing daily differences across them produces six zeros and a seventh day carrying the week, which reads as an event.
- Detect it from your own stored snapshots: look for change dates falling on the same weekday, exactly seven days apart, across unrelated projects.
- One source we checked drifts a little most days and takes a large reconciliation weekly, which defeats the obvious flat-day detection.
- The fix is to treat those sources as genuinely weekly: bucket into seven day periods, label the chart weekly, and compare week against week.
A chart shows a large spike every Thursday. The team looks for what happens on Thursdays.
Nothing happens on Thursdays. The source updates on Thursdays.
The mechanism
The pipeline snapshots a cumulative total every day and derives a daily figure by subtraction. If the upstream total only advances weekly, six of those differences are zero and the seventh contains the entire week.
Nothing is technically wrong. Every number reconciles, the weekly total is correct, and the daily chart is a lie about when the activity happened.
One week of a weekly source, three ways
IllustrativeDetecting it in your own data
Do not rely on documentation. Look at your stored snapshots and ask on which dates each source's value actually changed.
The clean case. Flat for six days, a jump on the seventh, repeating. If the jump lands on the same weekday across unrelated projects, that is a batch reporting schedule rather than a coincidence about your audience.
We confirmed one source this way where two separately tracked tracks advanced only on two dates, identical for both, exactly seven days apart.
The case that fools you. One source we checked drifts a small amount most days and then takes a much larger reconciliation on the same weekday each week. The daily-flat detector never fires, because it is never flat, and the naive daily chart shows a real trickle plus a weekly cliff.
We found it by looking at the size of daily changes rather than their presence: a floor of small movement with a periodic jump many times larger, on the same weekday, on two unrelated projects.
The trap in the middle. Checking during a flat week produces a false negative. We recorded one source as daily on that basis and had to revise it later. Check across at least three weeks before concluding anything.
The fix
Treat weekly sources as weekly, everywhere.
Bucket the series. Each point is the cumulative total at the end of a seven day window minus the total seven days earlier. That telescopes correctly, needs no interpolation and is exactly right regardless of when in the week the batch lands.
Label the chart. A weekly bar, not a daily line. The label is doing real work: it tells the reader the granularity is genuinely coarser rather than implying a smooth daily series.
Change the comparison window. Momentum on a daily source compares seven days against the previous seven. On a weekly source it compares one point against the previous point. Using a seven point window on weekly data compares seven weeks against seven weeks, which is a different and much slower question.
Do not emit daily events downstream. If the pipeline sends dated figures anywhere, send one weekly figure when the increase lands, with the period length attached, rather than seven interpolated days.
Classifying and handling a source
Illustrative- 1Look at change dates in your own snapshotsAcross at least three weeks and more than one project. A single flat week produces a false negative.
- 2Check change sizes, not only presenceA source can drift daily and still take a weekly batch reconciliation many times larger.
- 3If weekly, add it to the shared listOne constant imported by the chart builder, the sync, the summaries and the AI tools.
- 4Bucket into seven day windowsTotal at window end minus total seven days earlier. Telescopes correctly, no interpolation.
- 5Label it weekly and change the comparisonWeekly bars, and momentum compares one point against the previous point rather than seven against seven.
Keep one list
The set of weekly sources should be one named constant that every consumer imports: the chart builder, the sync, the summary generator, the AI tool layer.
The alternative is a chart that buckets weekly next to an assistant that reports daily deltas for the same source, and a customer who now has two different stories about the same number.
What not to do
Do not interpolate into daily. Spreading a weekly total across seven days invents a granularity the source does not have. If you do it anyway for chart smoothness, every generated point has to be flagged as estimated, and at that point the weekly bar was simpler and more honest.
Do not apply a lag correction instead. A weekly source is coarse rather than late. Shifting it moves the spike to a different day without removing it, which is the worst of both.
Do not assume it is stable. Providers change pipelines. The cadence you verified in spring can be different by autumn, so it is worth re-checking when a chart starts looking odd rather than assuming the classification is permanent.
The general lesson
Cadence is a property of the data that no API we work with states, and it is only discoverable from your own accumulated snapshots. That is a good argument for storing raw readings with their dates rather than only storing derived values: the derivation can be redone when you learn something new about the source, and the raw history is what let us diagnose all three of the cases above.
Common questions
- Why does my streaming chart spike on the same day every week?
- Because that source only advances its cumulative total once a week, in a batch, while the pipeline computes daily differences. Six days read as zero and the seventh carries the whole week. The tell is that the spike lands on the same weekday, exactly seven days apart, and often on unrelated projects at the same time.
- How do you detect a weekly reporting source?
- From your own stored snapshots, by looking at which dates each source's value actually changed, across at least three weeks and more than one project. Also check the size of daily changes, because some sources drift a little every day and then take a large same-weekday reconciliation, which defeats a flat-day check.
- Should weekly data be interpolated into daily values?
- No. Spreading a weekly total across seven days invents granularity the source does not have. Bucket into seven day periods and label the chart as weekly. If interpolation is used for smoothing anyway, every generated point must be flagged as estimated.
Sources
- 1Spotify for Artists, Reviewed August 2026. Data in Spotify for Artists
- 2Music Ally, January 2026. 5.1tn annual music streams, but 120.5m tracks had 10 or fewer
Danny Starr
Co-founder, Backline
Danny Starr is a co-founder of Backline and builds the platform. He writes about the data engineering behind music analytics: ingestion, identity, honesty in charts, and the AI layer on top of it.
Backline does this for the projects you run
Streaming, audience, social, advertising, website, search, ticketing and press data in one dashboard per project, with an AI assistant that answers questions about your own connected data. Invite-only.
Keep reading
Data engineering
Reporting lag: the two days nobody accounts for
Third-party streaming totals describe a day that already passed. Uncorrected, every chart and every correlation is offset, and the error is invisible.
Danny Starr · 3 min read
Data engineering
Cumulative counters, daily deltas, and the data you can lose forever
Music data arrives as a running total, so daily numbers have to be derived. The three bugs that derivation produces, one of which destroys data permanently.
Danny Starr · 3 min read
Analytics foundations
When a spike is not a hit
Six things that produce a chart spike with no underlying audience gain, how to tell them apart in an afternoon, and what to check before anyone reforecasts a campaign.
Danny Angove · 3 min read
