Can we split the 'all' category into meaningful buckets? It's currently too heterogeneous to interpret.
Confirmed on my side too.
Should GDP be in absolute terms for this comparison to mean anything?
Fixed. (edited to fix a unit)
Consistent with the published SLA. See /u/plg_avraham_beridze/p/plot-0015.
Makes sense, thanks.
The axis label is a bit small.
The baseline marking is unclear.
Well done.
I do not follow — is mobile subscriptions going up or down here?
The color convention isn't obvious.
Fine by me.
Perhaps naïve, but: would rotating the entire view 90° make the trends easier to read, or are we constrained by how it renders on mobile and in print? See /u/plg_avraham_beridze/p/plot-0015.
Needs a legend.
Good point there. (edited to fix a unit)
Missing the category remapping.
Needs a legend.
Need the définition of active status before we slice further?
Roger that.
That's the unadjusted figure.
Thanks , that resolves it.
The análisis looks solid.
Yes.
Which source is
observed coming from?Confirmed on my side too.
Do we have this in the Q1 pack yet?
Makes sense, thanks.
Works for me.
I think there's a calculation error here.
Same conclusion here.
Worth pointing out that we've historically underestimated the variance in this metric, so flagging the spike as potentially meaningful rather than noise is the conservative call.
Aligns with how the system prioritizes.
Lan Brandybuck Definitely.
Can we label the tick labels? Hard to read otherwise.
The gridlines are distracting.
Can we label the left axis? Hard to read otherwise.
Balin Mayene Well reasoned.
That holds up.
The percentage sign is missing.
What's the minimum sample size?
Matches our numbers too.
The units are missing from the secondary axis.
Very good.
Which vintage?
The way you've structured the lookback window to exclude the migration period is clever — it gives us clean apples-to-apples comparisons without having to build a manual adjustment factor.
I'm concerned about the outlier in Q3. Should we exclude it or investigate the cause?
Good catch.
Roger that.
Tāne Yates Should we be applying a holdout or control group adjustment here, or is the causal inference already baked into the metric definition in a way that accounts for selection bias?
Which vintage?
Good work.
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.
Right — headcount was the part I missed.
The 2020 break is a reporting lag — it shows up in every series from that source.
Elendil Raza 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.
Is the 1995 gap a reporting lag, or is the series genuinely missing?
Duplicated.
Not comparable.
What happened to Poland around 2015?
Which source is
baseline coming from?I'd suggest we document the assumptions and edge cases in a separate methodology guide that lives alongside the dashboard — it'll save us from answering the same questions about data quality over and over.
What's driving the spike in March? (edited to fix a unit)
Confirmed on my side too.
Tāne Yates Yep.
Same as before.
Is this the latest version?
Checks out.
Avraham Beridze 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.
That matches what I had.
Stale.
Can you split Q2 out?
Can we label the left axis? Hard to read otherwise.
Supported by the data.
The units are missing from the secondary axis.
Yes, exactly this.
The timing matches the network upgrade.
Minor: life expectancy is off in the tooltip.
Stale.
Should this be a log scale?
Are we filtering outliers?
Can we get a review from someone who wasn't involved in building the original metric definition — it's easy to miss subtle issues when you've been living in the code for months?
Right — urban share was the part I missed.
Nice — the Q4 view helps. (edited to fix a unit)
The rollup is dropping edge cases. (edited to fix a unit)
What happened to LatAm around 2017?
Consider rotating the labels 45 degrees.
Balin Mayene Are duplicates flagged?
Yes, exactly this.
Good catch.
What's driving the spike in March?
For background: Chile changed reporting in 1999, which is why H1 looks odd.
Can you split Q3 out?
Is this index-linked, or raw?
Makes sense, thanks.
I'm convinced by your argument that the regional patterns are statistically significant rather than noise — that's a crucial insight for how we should interpret the trend.
Supported by the data.
Can you split Q1 out? (edited to fix a unit)
Confirmed on my side too.
For background: Japan changed reporting in 2015, which is why H2 seems odd.
Which source is
Region coming from?Units?
Yep.
Fine by me.
Am I reading the secondary axis wrong? India looks stale to me.
This is much clearer, thanks.
This lines up with the other chart if you put retention in absolute terms.
Lan Brandybuck Are these cumulative or rolling?
Thanks for catching that détail.
Elendil Raza The análisis looks solid.
This lines up with the other chart if you put unemployment index-linked.
Definitely. See /u/plg_avraham_beridze/p/plot-0015.
Can we make the units clearer?
Context for anyone new: retention is only comparable in constant terms.
Should this go in the monthly summary?
What is the refresh cadence on this?
That's fair.
Fully aligned.
Sorry, lost me. What is the baseline?
We flagged this discrepancy in the August postmortem — the root cause was an incomplete migration of historical data that left a gap for approximately two weeks in early June across three regions.
Elendil Raza 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.
The metodología we're using here — is it documented somewhere? I'd like to understand the précis of how nulls are handled in the aggregation layer before we finalize this.
Correlates with the external market report. (edited to fix a unit)
Elendil Raza Confirmed. See /u/plg_avraham_beridze/p/plot-0015.
Not comparable.
Duplicated.
Is the forecasting model trained on the historical period that includes this anomaly, or did we exclude it as an outlier — and if so, doesn't that make the forecast artificially conservative?
raised this on the other chart too, same a definition change.
Can someone add this to the tracker?
The units are missing from the left axis.
Fair enough.
Matches our numbers too.
Avraham Beridze That's pre-deduplication.
Is the 2022 gap a coverage gap, or is the series genuinely missing?
Can you split H2 out?
That's fair. See /u/plg_avraham_beridze/p/plot-0015.
This is much clearer, thanks.
adjusted should probably be in absolute terms.The annotation arrows point to the correct values, but they're positioned so they obscure the peak — moving them outside the plot area or using a table for specific outliers would improve clarity.
Wait, which of these is Nigeria?
Lan Brandybuck Are these deduplicated?
Makes sense, thanks.
Should use thousands separator.
Not comparable.
Balin Mayene 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.
Not quite — the 2023 figure is a rebasing.
The pattern we're seeing mirrors what happened during the transition to the new billing system three months ago — initial shock, then steady recovery as users learned the new flow.
Colour order does not match the legend order.
What's the denominator here?
Confirmed on my side too.
I'd agree with this direction.