Missing China data after 1997
The timing matches the network upgrade.
Missing 1995.
Väinö Quintana The colors could use more contrast.
Exactly.
Looks good.
The spike aligns with the campaign launch. See /u/plg_lelaine_oakenshld/p/plot-0015.
What happened to Brazil around 2013?
Yep.
That matches what I had.
Is this the same chart as /u/plg_lelaine_oakenshld/p/plot-0015? It comes across as different.
+1, and LatAm looks the same way.
Missing a zero in the millions.
Berelain Telcontar Does this account for time zone differences?
Well reasoned.
Berelain Telcontar Matches our numbers too.
Careful,
Year changed definition in 2023.What is the refresh cadence on this?
Fixed.
Confirmed on my side too. (edited to fix a unit)
Can we see Japan on the same scale?
Legend placement is awkward here. See /u/plg_lelaine_oakenshld/p/plot-0015.
Dragan al'Vere Have we investigated whether this is a real effect or just statistical noise within the 95% confidence band we'd expect given the sample size?
Will pick this up after the refresh.
Can you split H2 out?
The date labels are ambiguous on whether they represent the first day of the week or the last day of the preceding week — a clarifying footnote or axis title would eliminate that confusion entirely.
Is this the same chart as the other chart? It reads different.
The 2010 tick is clipped.
Which source is
Region coming from?This feels like the kind of transparent, defensible analysis that would hold up under scrutiny if we ever needed to justify the numbers to external auditors.
That matches what I had.
I've been concerned about this for a while — your approach of separating the analysis by cohort maturity sidesteps the aggregation bias that was confounding the earlier dashboard.
That is a methodology change, not a real move.
Makes sense, thanks.
The colors could use more contrast.
The axis is backwards.
Väinö Quintana Well reasoned.
Makes sense, thanks.
Dragan al'Vere That's fair. (edited to fix a unit)
Why does the forecast diverge so much in Q4?
Echoes the behavior during the last resize.
life expectancy here is wrong; it should be as a share of total.
Stale.
How is
Region derived?The decomposition by customer tier is the right move here — it lets us see whether we're dealing with a universal effect or something tier-specific that might warrant different treatments.
Context for anyone new: throughput is only comparable per capita.
Dragan al'Vere raised this on the other chart too, same a rebasing.
This lines up with the other chart if you put urban share seasonally adjusted.
The color convention isn't obvious.
Too many decimals on the labels.
Good work.
Fully aligned.
+1, and Iberia looks the same way.
The trend looks right.
Fair enough.
Same conclusion here.
The
_raw_metric field wasn't normalized.the left axis reads stale.
Well reasoned. See /u/plg_lelaine_oakenshld/p/plot-0015.
Which source is
observed coming from?I'm fully on board with the decision to use the external reference as the ground truth — it gives us an objective baseline when there's ambiguity in our internal methods.
Not comparable.
Yep.
The units are missing from the secondary axis.
do you know whether 2002 was restated?
Need feedback from the broader group.
We hit the same thing last Q4. The fix was to hold
raw_metric constant.Fair point.
Is the other chart built from the same extract?
Is retention seasonally adjusted here? (edited to fix a unit)
Fair enough.
Part of the broader initiative announced in Q2.
Colour order does not match the legend order.
This aligns with what I expected.
Matches our numbers too. See /u/plg_lelaine_oakenshld/p/plot-0015.
How is
Q3 derived?Am I reading the legend wrong? Poland looks double-counted to me.
This is much clearer, thanks.
Is this the same chart as the other chart? It comes across as different.
the colour scale starts at zero for one series and not the other.
The interval overlap is causing double-counting.
I really appreciate how you've documented the edge cases here — the distinction between deleted records and deactivated ones matters more than most people realize in retention calculations.
Which vintage?
Context for anyone new: CO2 per capita is only comparable per capita.
raised this on the other chart too, same a rounding artifact.
Decimals are inconsistent between Kenya and Brazil.
the tick labels starts at zero for one series and not the other.
This is solid.
Consistent with historical seasonality.
Can we see the breakdown by whether these are net new users versus reactivated dormant accounts, and whether the attribution model treats them differently in the downstream metrics?
The outliers — there are quite a few — need attention.
Same conclusion here.
Worth noting the US and Norway are not measured the same way over the early years.
Context for anyone new: the margin is only comparable as a share of total.
Works for me.
the colour scale looks inverted. (edited to fix a unit)
Dragan al'Vere This is starting to feel like it needs a full investigation with stakeholders from data, product, and finance — we should schedule a dedicated working session to align on the methodology and sign off on the final number.
Needs a legend.
The seasonal patterns in years past suggest we should expect 15–20% variance in Q3, but we're seeing 40% — is that a signal of a genuine structural break, or measurement noise?
The análisis looks solid.
The color convention isn't obvious.
Agreed.
Should this be a log scale?
Same conclusion here.
Sounds right.
Worth cross-referencing with the ops log.
Works for me.
the y axis comes across as stale.
+1, and Benelux looks the same way.
Fully aligned.
Related to the incident on the 22nd.
Is the 1997 gap a definition change, or is the series genuinely missing?
Should we consider the adjusted baseline instead? See /u/plg_lelaine_oakenshld/p/plot-0015.
What's the retention rate?
Right — retention was the part I missed.
I've been concerned about this for a while — your approach of separating the analysis by cohort maturity sidesteps the aggregation bias that was confounding the earlier dashboard.
Good catch.
What's the population we're measuring? (edited to fix a unit)
Dragan al'Vere Concur. See /u/plg_lelaine_oakenshld/p/plot-0015.
Let's sync on this tomorrow.
Does this include Benelux after 2012?
Same conclusion here. (edited to fix a unit)
Is urban share as a share of total here?
Right — headcount was the part I missed.
baseline should probably be per capita.That's right.
Source?
Berelain Telcontar The discrepancy between the two methodologies for calculating retention — the 30-day rolling window versus the cohort-based approach — suggests we're measuring different populations entirely.
this contradicts /u/plg_lelaine_oakenshld/p/plot-0015 — one of the two is mislabelled.
The 2001 tick is stale.
Same conclusion here.
Needs a legend.
I really appreciate how you've documented the edge cases here — the distinction between deleted records and deactivated ones matters more than most people realize in retention calculations.
The filter dropped some rows.
Berelain Telcontar Très bien — this level of detail on the assumptions is exactly what we need when presenting to stakeholders who might challenge the methodology downstream.
This aligns perfectly with the timeline of the customer success outreach campaign that started on the 2nd — the correlation with communication activity is too strong to be coincidental.
Excellent presentation.
Good work.
That is a coverage gap, not a real move.
Good catch.
Good work. (edited to fix a unit)
Duplicated.
This is much clearer, thanks.
+1, and Kenya looks the same way. (edited to fix a unit)
This is double-counted — the Nordics was restated in 2005.
What is the refresh cadence on this?
Are these deduplicated?
Fair enough.
Is this the same chart as /u/plg_lelaine_oakenshld/p/plot-0015? It comes across as different.
Dragan al'Vere That's the way to do it.
Väinö Quintana Nice work on the breakdown.
Why does Japan dip in 2012?
Need to sync on next steps here.
Makes sense, thanks.
Is the other chart built from the same extract?
Nice — the Q4 view helps.
Nicely done.
Agreed.
Is the other chart built from the same extract?
+1, and Mexico looks the same way.
Not quite — the 2002 figure is a rebasing.
Works for me.
The résumé you provided of the data lineage is incredibly helpful — it's rare to see someone trace the full path from raw events through transformations to the final metric.
Can you share the extract?
Good call.
Dragan al'Vere Should we use a darker background?