Northern Lights Air · Loyalty Program Dashboard

Overview

Where the program stands

★ Net active members

14,670

▲+19.2% YoY

vs 12,305 at Dec 2017

✓ ≥ +15% · HIT

Gross adds FY2018

3,010

▲+21.0% YoY

2,365 net of cancellations

best year on record

Campaign lift

+41%

▲+281 incremental members

vs trend-adjusted counterfactual

✓ ≥ +25% · HIT

CAC per incremental member

$332

▲4% of a member's $8,047 CLV

expected-cost basis

✓ ≤ $800 · HIT

Annual churn rate

5.24%

▲+0.34 pp YoY

645 cancellations

✓ ≤ 5.5% · HIT

Monthly active member rate

48.2%

▲+5.1 pp YoY

members flying in a given month

✗ ≥ 50% · MISS

Flights per active member

2.95

▲+7.8% YoY

per active member per month

CLV under management

$117M

▲+19.5% YoY

$7,969 per active member

Points redemption rate

1.54%

●98.5% breakage

782M points outstanding

watch

Points liability (expected)

$2.2M

●$141M nominal

at today's redemption rate

stress-test

★ North Star — Net Active Members

Members enrolled and not cancelled, 2012–2018

The program closes 2018 at 14,670 active members, +19.2% on the year and clear of the +15% target. Every other KPI on this dashboard is a lever on this one line.

Table view · Monthly member base, 2016–2018

Year	Month	Active members	Gross adds	Cancellations	Net adds
2,016	1	8,460	205	40	165
2,016	2	8,607	181	34	147
2,016	3	8,785	208	30	178
2,016	4	8,978	214	21	193
2,016	5	9,175	231	34	197
2,016	6	9,351	210	34	176
2,016	7	9,514	211	48	163
2,016	8	9,683	207	38	169
2,016	9	9,834	190	39	151
2,016	10	9,995	198	37	161
2,016	11	10,147	190	38	152
2,016	12	10,324	211	34	177
2,017	1	10,469	188	43	145
2,017	2	10,633	196	32	164
2,017	3	10,835	237	35	202
2,017	4	10,994	196	37	159
2,017	5	11,182	231	43	188
2,017	6	11,338	200	44	156
2,017	7	11,501	217	54	163
2,017	8	11,665	208	44	164
2,017	9	11,837	210	38	172
2,017	10	12,006	206	37	169
2,017	11	12,148	191	49	142
2,017	12	12,305	207	50	157
2,018	1	12,446	192	51	141
2,018	2	12,695	295	46	249
2,018	3	12,972	330	53	277
2,018	4	13,276	346	42	304
2,018	5	13,473	244	47	197
2,018	6	13,699	272	46	226
2,018	7	13,890	231	40	191
2,018	8	14,030	218	78	140
2,018	9	14,163	185	52	133
2,018	10	14,321	213	55	158
2,018	11	14,504	246	63	183
2,018	12	14,670	238	72	166

Growth engine — gross adds vs cancellations

Monthly, 2016–2018

2018 added 3,010 members and lost 645, a net +2,365 — the best year on record. The campaign spike in Feb–Apr survives the churn offset, which is the question management actually asked (gross and net).

Table view · Annual member flow, 2012–2018

Year	Gross adds	Cancellations	Net adds	Members at year end
2,012	1,686	0	1,686	1,686
2,013	2,397	43	2,354	4,040
2,014	2,370	181	2,189	6,229
2,015	2,331	265	2,066	8,295
2,016	2,456	427	2,029	10,324
2,017	2,487	506	1,981	12,305
2,018	3,010	645	2,365	14,670

Campaign impact

Did the Feb–Apr 2018 promotion work?

Gross adds — the campaign months break the trend line

New enrolments per month, 2017 vs 2018

Feb–Apr 2018 delivered 971 sign-ups against 629 in the same window of 2017 — +54.4%. Note that the nine non-campaign months run only slightly ahead of 2017: the step is confined to the promotion window, which is what a working campaign looks like.

Table view · New enrolments by month

Month	2017	2018	Change
Jan	188	192	4
Feb	196	295	99
Mar	237	330	93
Apr	196	346	150
May	231	244	13
Jun	200	272	72
Jul	217	231	14
Aug	208	218	10
Sep	210	185	-25
Oct	206	213	7
Nov	191	246	55
Dec	207	238	31

…and ~281 of those members were genuinely incremental

Counterfactual = 2017 baseline × the +9.7% growth observed in the nine non-campaign months

Comparing to last year alone would claim +342 members. Stripping out the growth the program was making anyway leaves +281 (+40.7%) — the honest number, and still comfortably past the +25% success threshold.

Table view · Campaign lift calculation

Measure	Value
Feb–Apr 2017 actual	629
Underlying 2018 growth trend	×1.097
Feb–Apr 2018 counterfactual	690
Feb–Apr 2018 actual	971
Incremental members	+281
Lift vs counterfactual	+40.7%
Lift vs last year	+54.4%

Reverse-engineering the offer: the campaign is a permanent 1.5× earn rate

Share of flown member-months earning 1.5 points per km

The data dictionary says when the promotion ran but never what it offered. Points ÷ kilometres gives the answer: all 924 campaign members who ever flew earn 1.5 pts/km, zero standard members ever do, and no member is ever on both rates. Commercially this matters — the offer is not a signing bonus but an annuity of discount that keeps costing on every future kilometre.

Table view · Members by enrolment type × maximum earn rate observed

Enrolment type	1.0	1.5	never flew
2018 Promotion	0	924	47
Standard	14,243	0	1,523

Response was geographic, not demographic

Campaign share of each province's members ÷ the national rate

Among provinces with a base large enough to act on, Québec (1.22) and British Columbia (1.09) pulled above the national rate while Manitoba (0.52), Nova Scotia (0.63) and Alberta (0.69) lagged. Yukon and PEI show the most extreme indices but sit on 110 and 66 members — they are noise, and are greyed out here rather than quietly dropped. Meanwhile every demographic cut — gender, education, marital status, tier — moves less than 3 pp. The offer appealed across the board; only geography discriminated, so the next wave should be bought regionally.

Table view · Campaign response and profile by province

Province	Members	Campaign members	Campaign rate	Mean CLV	Churn rate	Response index
Yukon	110	9	8.2%	$6,772	10.9%	1.41
Quebec	3,300	233	7.1%	$8,161	12.5%	1.22
British Columbia	4,409	280	6.4%	$7,994	11.8%	1.09
Newfoundland	258	16	6.2%	$8,025	14.7%	1.07
Saskatchewan	409	24	5.9%	$8,076	12.2%	1.01
Ontario	5,404	297	5.5%	$7,914	12.5%	0.95
New Brunswick	636	33	5.2%	$8,154	10.5%	0.89
Alberta	969	39	4.0%	$7,753	12.6%	0.69
Nova Scotia	518	19	3.7%	$7,983	11.2%	0.63
Manitoba	658	20	3.0%	$8,067	15.2%	0.52
Prince Edward Island	66	1	1.5%	$7,704	16.7%	0.26

Demographics barely moved

Campaign mix minus standard mix, percentage points

Everything here sits inside ±3 pp. For a promotion this is a good result — it means the 1.5× offer has broad appeal and does not need demographic targeting. It also means any segment-based media plan would have been spending money to find a difference that is not there.

Table view · Campaign vs standard member mix by demographic

Dimension	Segment	Standard %	Campaign %	Difference (pp)
Gender	Female	50.2	50.9	0.67
Gender	Male	49.8	49.1	-0.67
Education	High School or Below	4.6	5.1	0.51
Education	College	25.4	24.5	-0.86
Education	Bachelor	62.4	65.1	2.66
Education	Master	3.1	2.0	-1.14
Education	Doctor	4.5	3.3	-1.16
Marital status	Divorced	15.0	16.0	0.97
Marital status	Married	58.2	57.5	-0.74
Marital status	Single	26.8	26.6	-0.23
Card tier	Star	45.7	44.6	-1.10
Card tier	Nova	33.9	34.0	0.11
Card tier	Aurora	20.4	21.4	0.99

Campaign members fly ~5× more per head

Flights per member per month, 2018

Campaign members are 5.8% of the base but flew 15.7% of all 2018 flights. In July they averaged 8.7 flights per member against 1.9 for everyone else. The 1.5× rate did not buy dormant sign-ups — it bought high-frequency travellers.

Table view · 2018 monthly engagement, campaign vs standard members

Segment	Month	Flights per member	Active member rate
Standard	Jan	0.84	45.2%
Standard	Feb	0.83	45.3%
Standard	Mar	1.19	48.1%
Standard	Apr	0.94	45.3%
Standard	May	1.33	48.4%
Standard	Jun	1.63	48.2%
Standard	Jul	1.86	51.9%
Standard	Aug	1.60	49.7%
Standard	Sep	1.22	47.5%
Standard	Oct	1.14	47.7%
Standard	Nov	1.10	48.6%
Standard	Dec	1.52	52.0%
Campaign	Feb	0.40	21.7%
Campaign	Mar	0.87	34.2%
Campaign	Apr	0.95	44.7%
Campaign	May	6.41	60.1%
Campaign	Jun	8.33	58.4%
Campaign	Jul	8.69	62.4%
Campaign	Aug	7.68	60.2%
Campaign	Sep	2.79	54.3%
Campaign	Oct	2.65	54.3%
Campaign	Nov	2.36	53.5%
Campaign	Dec	5.00	55.8%

…and more of them fly in any given month

Monthly active member rate, 2018

The campaign cohort peaks at a 62% monthly active rate against roughly 50% for standard members, and stays above them all year. This is the single strongest argument for re-running the promotion: it recruits people who actually fly.

What the 1.5× mechanic really costs

Headline liability vs expected cost vs cash paid — log scale

The mechanic issued 33.7M bonus points in 2018 — a headline $6.1M. But members redeem only 1.54% of what they earn, so the expected cost is $93,301 and the cash NLA actually paid campaign members in 2018 was $67,141. Against 281 incremental members that is a CAC of $332 — about 4% of a campaign member's mean CLV of $8,047. Caveat: the 1.5× rate is permanent, so this is an annuity of cost, not a one-off.

Table view · Campaign return on investment

Campaign economics	Value
Incremental members won	281
Mean CLV of a campaign member	$8,047
Incremental CLV acquired	$2,258,856
Bonus points issued (2018)	33,653,246 pts
— notional cost @ $0.18	$6,057,584
— expected cost @ 1.54% redemption	$93,301
— cash actually paid in 2018	$67,141
CAC per incremental member (expected-cost basis)	$332
CAC per incremental member (full-liability basis)	$21,578
Return on incremental CLV (expected-cost basis)	24×

Engagement

Do members actually fly?

Existing members flew more after the campaign

Flights per month · balanced panel (the 971 campaign joiners are excluded)

Because campaign members have no 2017 history, leaving them in would manufacture a +16% year-on-year jump out of cohort mix alone. Excluding them makes this a genuine like-for-like read on the pre-existing base — and it still shows a clear separation opening from May.

Table view · Flights by month, balanced panel

Month	2017 flights	2018 flights	YoY %
Jan	13,031	13,285	1.9
Feb	13,335	13,097	-1.8
Mar	18,324	18,737	2.3
Apr	15,416	14,863	-3.6
May	18,629	21,029	12.9
Jun	23,416	25,651	9.5
Jul	26,201	29,367	12.1
Aug	22,893	25,232	10.2
Sep	17,382	19,234	10.7
Oct	16,387	18,044	10.1
Nov	15,686	17,327	10.5
Dec	21,786	23,887	9.6

Difference-in-differences: −0.3% before, +10.7% after

Year-on-year change in flights, same members both years

Flights ran -0.3% in Jan–Apr and +10.7% in May–Dec — a +11.0 pp difference-in-differences among members who were already enrolled. Their monthly active rate rose from 43.0% to 48.2% over the same period. Read with care: one airline, no untreated control market — seasonality is controlled by comparing like months a year apart, but a mid-2018 travel-demand shift cannot be ruled out. This is strongly suggestive, not causal proof.

Seasonality — July runs 44% above an average month, January 32% below

Flights per month ÷ average month × 100 · 2017–2018 · balanced panel

Summer (Jun–Aug) carries 33% of the year's flights in 25% of the months, and December spikes on holiday travel. Any campaign aimed at flight volume should land in April–May so new members are activated before the peak — which is precisely what the Feb–Apr timing achieved.

Table view · Seasonality index, balanced panel

Month	Flights	Index
Jan	26,316	68.3
Feb	26,432	68.6
Mar	37,061	96.2
Apr	30,279	78.6
May	39,658	103.0
Jun	49,067	127.4
Jul	55,568	144.3
Aug	48,125	124.9
Sep	36,616	95.1
Oct	34,431	89.4
Nov	33,013	85.7
Dec	45,673	118.6

Half the base is barely engaged

Distribution of active months per member, 2017–2018

The monthly active member rate is 48.2% — short of the 50% target and the clearest untapped lever on the dashboard. Dormancy, not churn, is where the volume is: moving the rate by one point is worth roughly 6,000 incremental flights a year, and these members are already enrolled, so there is no acquisition cost to pay.

Table view · Members by number of active months

Active months	Members	Mean CLV	Churn rate	Total flights
0.00	1,570.00	$8,361	60.6%	0.00
1.00	409.00	$7,829	19.6%	1,125.00
2.00	423.00	$8,541	22.7%	2,332.00
3.00	448.00	$7,982	24.8%	4,551.00
4.00	491.00	$8,086	34.8%	8,298.00
5.00	499.00	$7,876	32.5%	12,037.00
6.00	527.00	$7,624	27.9%	16,005.00
7.00	476.00	$8,463	18.5%	15,560.00
8.00	445.00	$7,303	9.4%	14,098.00
9.00	478.00	$7,383	9.0%	13,043.00
10.00	646.00	$7,846	5.6%	17,774.00
11.00	871.00	$8,177	3.8%	25,638.00
12.00	1,213.00	$7,715	1.8%	38,790.00
13.00	1,582.00	$7,850	2.0%	55,133.00
14.00	1,643.00	$7,899	0.7%	61,947.00
15.00	1,633.00	$8,273	1.0%	65,383.00
16.00	1,386.00	$7,919	1.1%	59,277.00
17.00	958.00	$7,994	0.8%	43,773.00
18.00	564.00	$8,218	0.4%	27,145.00
19.00	301.00	$7,695	0.3%	15,304.00
20.00	111.00	$8,049	0.0%	6,139.00
21.00	44.00	$8,017	0.0%	2,559.00
22.00	16.00	$8,426	0.0%	945.00
23.00	2.00	$18,984	0.0%	120.00
24.00	1.00	$6,632	0.0%	70.00

Retention & churn

Where the base leaks

Half of all churn happens on one date

Tenure at cancellation, all 2,067 cancellations

1,048 of 2,067 cancellations (51%) occur at exactly month 8 of membership — 28× the neighbouring months. A spike that precise is not customer behaviour, it is a contract term expiring: NLA's loyalty program evidently carries an eight-month introductory period, and a large share of members leave the moment it lapses. In 2018 alone this single moment cost $2.67M of CLV across 325 members. It is the highest-leverage retention target in the business, because the date is known in advance for every member.

Table view · The month-8 cliff by enrolment cohort

Enrolment cohort	Cohort size	Cancelled at month 8	Month-8 churn rate	Total cancellations	Month 8 as % of cohort churn
2,012.00	1,686.00	0.00	0.0%	0.00	—
2,013.00	2,397.00	54.00	2.3%	362.00	14.9%
2,014.00	2,370.00	131.00	5.5%	387.00	33.9%
2,015.00	2,331.00	204.00	8.8%	404.00	50.5%
2,016.00	2,456.00	246.00	10.0%	390.00	63.1%
2,017.00	2,487.00	287.00	11.5%	365.00	78.6%
2,018.00	3,010.00	126.00	4.2%	159.00	79.2%

…and it is program-wide, not a campaign defect

Share of each enrolment cohort that cancels at exactly month 8

This is the correction that matters. The campaign cohort's cancellations look dramatic — 101 of 115 (88%) fall on month 8 — but that ratio is inflated by right-censoring: those members have only been enrolled 8–11 months, so month 8 is nearly the only tenure at which they could churn. Comparing rates instead of ratios, 10.4% of campaign members cancelled at month 8 versus 11.5% of the 2017 cohort and 10.0% of the 2016 cohort. The campaign did not create the cliff and its members are not unusually fragile — the cliff is a program design flaw affecting every cohort since 2013, and fixing it is worth far more than fixing it for one campaign.

Table view · Month-8 churn rate, like-for-like

Cohort	Month-8 churn rate
2013	2.3%
2014	5.5%
2015	8.8%
2016	10.0%
2017	11.5%
2018	4.2%
2018 campaign cohort	10.4%

Every cohort falls off the same month-8 cliff

Share of each cohort still enrolled, m months after joining

All three curves are flat to month 7 and then drop together — the campaign cohort, the 2018 standard cohort and the 2017 cohort alike. The campaign cohort is in fact the best of the three at every duration, ending at 88.2% against 85.6% for 2018 standard. The lesson is not that the promotion attracted fragile members; it is that every member hits the same wall at eight months. A tapered benefit step-down plus a save offer fired in month 7 — applied program-wide, not just to campaign joiners — is the single highest-return retention intervention available.

Table view · Retention curves, % still enrolled

Months since enrolment	2017 Standard	2018 Promotion	2018 Standard
0.00	99.88	100.00	99.95
1.00	99.72	99.59	99.89
2.00	99.60	99.38	99.74
3.00	99.40	99.38	99.55
4.00	99.20	99.18	99.39
5.00	99.08	99.07	98.94
6.00	98.91	98.97	98.45
7.00	98.67	98.87	98.39
8.00	87.13	88.47	85.94
9.00	86.89	88.48	84.90
10.00	86.73	88.47	84.90
11.00	86.57	nan	84.38

Every cohort decays at the same steady rate

Cohort retention heat-map, 2012–2018

Roughly 3 percentage points lost per year, remarkably stable across seven annual cohorts. There is no deteriorating vintage and no improving one — which tells us retention is driven by the program design, not by who was recruited in a given year.

Table view · Retention (%) at 6-month intervals

Cohort	0	6	12	18	24	30	36	42	48	54	60
2,012.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00	100.00
2,013.00	99.90	98.60	95.10	93.90	93.00	92.00	90.90	89.70	88.50	86.90	85.90
2,014.00	99.90	98.40	92.00	90.50	89.50	88.40	87.50	86.10	84.90	84.60	nan
2,015.00	100.00	98.70	88.70	87.20	86.20	85.00	83.70	82.70	nan	nan	nan
2,016.00	99.90	98.90	87.70	86.60	85.10	84.80	nan	nan	nan	nan	nan
2,017.00	99.90	98.90	86.40	86.00	nan	nan	nan	nan	nan	nan	nan
2,018.00	100.00	98.70	nan	nan	nan	nan	nan	nan	nan	nan	nan

Churn is structural, not segmental

Lifetime churn rate by segment (groups with n ≥ 50)

Every demographic cut sits inside an 11.6%–13.1% band around the 12.3% base rate, and members who leave are worth slightly more than those who stay ($8,132 vs $7,969 mean CLV). There is no bad segment to stop recruiting. Retention spend should target moments — the month-8 cliff, the first dormant quarter — not demographic profiles. That is also the cheaper intervention.

Table view · Churn and value by every segment

Dimension	Segment	Members	Churn rate	Mean CLV
Card	Star	7,637	11.8%	$6,742
Card	Nova	5,671	12.6%	$8,046
Card	Aurora	3,429	13.1%	$10,673
Education	High School or Below	782	11.9%	$7,707
Education	College	4,238	12.4%	$7,595
Education	Bachelor	10,475	12.3%	$8,207
Education	Master	508	12.8%	$7,441
Education	Doctor	734	13.1%	$7,833
Marital status	Divorced	2,518	11.6%	$8,201
Marital status	Married	9,735	12.5%	$8,058
Marital status	Single	4,484	12.5%	$7,719
Gender	Female	8,410	12.5%	$7,982
Gender	Male	8,327	12.2%	$7,995
Salary band	< $50k	926	12.2%	$8,478
Salary band	$50-75k	5,718	12.4%	$8,150
Salary band	$75-100k	4,334	12.2%	$8,229
Salary band	$100-150k	1,047	12.8%	$7,445
Salary band	$150k+	474	11.8%	$7,621
Province	Alberta	969	12.6%	$7,753
Province	British Columbia	4,409	11.8%	$7,994
Province	Manitoba	658	15.2%	$8,067
Province	New Brunswick	636	10.5%	$8,154
Province	Newfoundland	258	14.7%	$8,025
Province	Nova Scotia	518	11.2%	$7,983
Province	Ontario	5,404	12.5%	$7,914
Province	Prince Edward Island	66	16.7%	$7,704
Province	Quebec	3,300	12.5%	$8,161
Province	Saskatchewan	409	12.2%	$8,076
Province	Yukon	110	10.9%	$6,772

Value & segments

Who is worth defending

The book is top-heavy

Share of total CLV held by each decile of members

The top decile holds 31% of portfolio value and the top quintile 46%. That concentration is what makes a targeted retention programme worth building: defending the top 20% of members defends nearly half the book.

Table view · CLV deciles

Decile	Members	Mean CLV	Total CLV	Flights	Churn rate	% of value
1.00	1,674.00	$2,449	$4,100,254	29.7	11.5%	3.1
2.00	1,675.00	$2,954	$4,948,259	30.4	11.7%	3.7
3.00	1,672.00	$3,979	$6,653,379	30.5	12.4%	5.0
4.00	1,675.00	$4,814	$8,063,299	30.4	12.7%	6.0
5.00	1,682.00	$5,435	$9,142,152	30.1	13.2%	6.8
6.00	1,665.00	$6,477	$10,784,100	30.7	13.7%	8.1
7.00	1,673.00	$7,787	$13,027,039	30.3	11.2%	9.7
8.00	1,677.00	$9,050	$15,176,532	30.6	11.6%	11.4
9.00	1,670.00	$12,476	$20,835,580	30.3	12.5%	15.6
10.00	1,674.00	$24,480	$40,979,567	29.9	13.0%	30.6

Aurora punches above its weight

Share of members vs share of portfolio value, by card tier

Aurora is 20% of members but 27% of value, with a median CLV +70% above Star. Yet flights per member are identical across tiers (30.2 / 30.4 / 30.4) — tier is not earned by flying in this program. That makes tier upgrade a pricing and merchandising decision, not a reward for volume. Migrating 5% of Star members to Nova is worth roughly $0.5M of CLV under management.

Table view · Profile by card tier

Card tier	Members	Mean CLV	Median CLV	Total CLV	Flights per member	Churn rate
Star	7,637	$6,742	$4,787	$51,486,832	30.2	11.8%
Nova	5,671	$8,046	$5,799	$45,626,688	30.4	12.6%
Aurora	3,429	$10,673	$8,140	$36,596,641	30.4	13.1%

Tier shifts the whole distribution, but the spread within a tier is wider

CLV distribution by card tier · boxes are the interquartile range

Medians step cleanly upward — Star $4,787 → Nova $5,799 → Aurora $8,140 — but the boxes overlap heavily, so tier predicts value on average without determining it for any individual member. That nuance matters for targeting: a tier-based campaign will reach plenty of low-value Aurora members and miss plenty of high-value Star ones. This chart is also the evidence behind treating CLV as a tier-linked score rather than observed spend — CLV correlates ≈0 with distance flown (−0.00), flights (−0.01), tenure (0.00) and salary (−0.02). Every value figure on this dashboard is therefore CLV under management, never booked revenue.

Table view · CLV distribution statistics by tier

Card tier	Members	Mean	P25	Median	P75	Max
Star	7,637.00	$6,742	$2,725	$4,787	$7,897	$83,325
Nova	5,671.00	$8,046	$4,015	$5,799	$8,962	$74,229
Aurora	3,429.00	$10,673	$6,063	$8,140	$11,895	$73,226

RFM segmentation — where value and headcount diverge

Recency / Frequency / Monetary scoring on the flying population

Segments where the bar runs past the dot hold more value than headcount. Three readings matter. Champions (11% of members, 16% of value, 45 flights each, 1% churn) are the core to protect. At risk (high value) — 12% of members but 18% of value, with the highest mean CLV in the book at $12,497 — have lapsed despite flying heavily; this is where a save contact has the highest expected return. And 1,570 members have never taken a single flight, of whom 61% have already cancelled — an acquisition-quality problem that no retention campaign can fix, and the one segment worth screening out at sign-up rather than winning back.

Table view · RFM segment profile

RFM segment	Members	Mean CLV	Total CLV	Mean flights	Churn rate	Median recency (months)	% of members	% of value
Slipping away	3,659	$9,200	$33,662,506	29.5	14.2%	1.0	21.9	25.2
At risk (high value)	1,965	$12,497	$24,556,202	43.6	3.1%	1.0	11.7	18.4
Champions	1,867	$11,598	$21,654,219	45.4	0.7%	0.0	11.2	16.2
Big spenders	1,414	$11,377	$16,086,480	23.8	0.4%	0.0	8.4	12.0
Never flew	1,570	$8,361	$13,126,488	0.0	60.6%	99.0	9.4	9.8
Loyal flyers	2,429	$3,968	$9,637,281	44.6	0.5%	0.0	14.5	7.2
Hibernating	1,959	$3,913	$7,666,240	21.8	25.6%	2.0	11.7	5.7
Promising	1,874	$3,906	$7,320,747	23.5	0.2%	0.0	11.2	5.5

Points economics

What the program costs

Accrual dwarfs redemption every single month

Points accrued vs redeemed, 2017–2018

Across 24 months members earned 794M points and redeemed 12.2M — a redemption rate of 1.54% and breakage of 98.46%. The loyalty program is, in cash terms, funded almost entirely by points that are never used.

Table view · Monthly points flow

Year	Month	Points accrued	Points redeemed	Cash cost (CAD)	Redemption rate
2,017.00	1.00	19,735,330.00	351,079.00	$63,214	1.78%
2,017.00	2.00	19,777,954.00	356,485.00	$64,182	1.80%
2,017.00	3.00	27,379,770.00	443,731.00	$79,876	1.62%
2,017.00	4.00	23,176,724.00	415,371.00	$74,778	1.79%
2,017.00	5.00	27,713,330.00	467,855.00	$84,217	1.69%
2,017.00	6.00	35,341,513.00	575,017.00	$103,518	1.63%
2,017.00	7.00	39,086,938.00	584,114.00	$105,157	1.49%
2,017.00	8.00	34,346,465.00	544,887.00	$98,086	1.59%
2,017.00	9.00	26,001,387.00	474,227.00	$85,371	1.82%
2,017.00	10.00	24,525,813.00	438,563.00	$78,949	1.79%
2,017.00	11.00	23,690,277.00	410,276.00	$73,876	1.73%
2,017.00	12.00	32,547,125.00	546,821.00	$98,444	1.68%
2,018.00	1.00	19,943,433.00	427,708.00	$77,003	2.14%
2,018.00	2.00	20,515,407.50	420,067.00	$75,625	2.05%
2,018.00	3.00	30,225,349.50	532,753.00	$95,915	1.76%
2,018.00	4.00	24,356,433.50	484,437.00	$87,211	1.99%
2,018.00	5.00	45,414,472.00	555,249.00	$99,968	1.22%
2,018.00	6.00	56,342,946.00	659,298.00	$118,671	1.17%
2,018.00	7.00	63,614,826.00	694,223.00	$124,990	1.09%
2,018.00	8.00	54,539,659.00	652,358.00	$117,433	1.20%
2,018.00	9.00	35,284,250.00	543,768.00	$97,903	1.54%
2,018.00	10.00	32,777,965.00	514,307.00	$92,580	1.57%
2,018.00	11.00	31,011,432.00	495,619.00	$89,221	1.60%
2,018.00	12.00	46,936,417.00	645,633.00	$116,252	1.38%

The liability nobody has priced

Cost of the 782M outstanding points under different redemption scenarios

At today's 1.54% the outstanding bank is worth $2.2M. At the 10–25% redemption rates typical of airline programmes it would be $14M–$35M. Two things follow. (1) The nominal liability — 782M points × $0.18 = $141M — massively overstates the cash exposure, so it should never be quoted raw. (2) A redemption rate this low is itself a warning sign: it means members do not see rewards as reachable, which is the condition that precedes disengagement. Low redemption flatters the P&L today and erodes the program tomorrow.

Table view · Points liability sensitivity

Redemption scenario	Expected cost (CAD)	vs today
1.5%	$2,167,846	1.0×
5.0%	$7,038,462	3.2×
10.0%	$14,076,925	6.5×
15.0%	$21,115,387	9.7×
20.0%	$28,153,849	13.0×
25.0%	$35,192,312	16.2×

Campaign members hoard their points

Redemption rate by enrolment type and card tier

Campaign members redeem at 0.37% against 1.71% for standard members — roughly 4× less, despite earning 50% faster. In the short term that makes the campaign even cheaper than modelled. In the medium term it is the risk: a large, fast-growing balance held by the most engaged flyers, unredeemed. If that cohort ever redeems in bulk, the cost lands all at once.

Table view · Points economics by segment

Dimension	Segment	Points accrued	Points redeemed	Cash cost	Redemption rate
Enrolment type	Campaign	100,959,739.50	372,894	$67,141	0.37%
Enrolment type	Standard	693,325,477.00	11,860,952	$2,135,299	1.71%
Card tier	Star	358,969,038.00	5,610,623	$1,010,097	1.56%
Card tier	Nova	270,499,520.00	4,095,183	$737,252	1.51%
Card tier	Aurora	164,816,658.50	2,528,040	$455,091	1.53%

Geography & mix

Where members are

Members by province

Ontario, British Columbia and Québec are 78% of the base

The member base mirrors Canada's population distribution, so there is no obvious under-served region to attack. The interesting geography is in response rates, not headcount — see the Campaign tab.

Table view · Full provincial profile

Province	Members	Mean CLV	Total CLV	Flights per member	Churn rate	Campaign rate
Ontario	5,404	$7,914	$42,765,917	30.3	12.5%	5.5%
British Columbia	4,409	$7,994	$35,244,344	30.5	11.8%	6.4%
Quebec	3,300	$8,161	$26,931,156	30.2	12.5%	7.1%
Alberta	969	$7,753	$7,512,452	30.3	12.6%	4.0%
Manitoba	658	$8,067	$5,307,794	29.0	15.2%	3.0%
New Brunswick	636	$8,154	$5,186,060	31.3	10.5%	5.2%
Nova Scotia	518	$7,983	$4,135,352	29.8	11.2%	3.7%
Saskatchewan	409	$8,076	$3,303,237	29.9	12.2%	5.9%
Newfoundland	258	$8,025	$2,070,471	29.4	14.7%	6.2%
Yukon	110	$6,772	$744,882	33.3	10.9%	8.2%
Prince Edward Island	66	$7,704	$508,496	27.6	16.7%	1.5%

Top 14 cities

Toronto, Vancouver and Montréal alone are 48% of members

Extreme urban concentration. Combined with the provincial response index, this points to a very cheap next campaign: Montréal and Vancouver metro cover the two highest-responding provinces and roughly a third of the entire member base.

Table view · All 29 cities

City	Members	Province	Mean CLV	Flights per member	Churn rate
Toronto	3,351	Ontario	$7,862	30.6	12.3%
Vancouver	2,582	British Columbia	$7,847	30.3	11.9%
Montreal	2,059	Quebec	$8,318	30.1	12.7%
Winnipeg	658	Manitoba	$8,067	29.0	15.2%
Whistler	582	British Columbia	$8,769	30.4	11.9%
Halifax	518	Nova Scotia	$7,983	29.8	11.2%
Ottawa	509	Ontario	$8,034	31.0	10.6%
Trenton	486	Ontario	$7,788	29.4	13.8%
Edmonton	486	Alberta	$7,920	30.9	12.3%
Quebec City	485	Quebec	$7,898	29.8	10.5%
Dawson Creek	444	British Columbia	$8,033	30.1	12.4%
Fredericton	425	New Brunswick	$8,213	31.4	10.4%
Regina	409	Saskatchewan	$8,076	29.9	12.2%
Kingston	401	Ontario	$7,806	29.5	13.7%
Tremblant	398	Quebec	$7,745	30.7	13.6%
Victoria	389	British Columbia	$8,160	31.2	11.8%
Hull	358	Quebec	$8,074	30.7	13.1%
West Vancouver	324	British Columbia	$7,354	32.1	11.4%
St. John's	258	Newfoundland	$8,025	29.4	14.7%
Thunder Bay	256	Ontario	$8,501	29.9	11.3%
Sudbury	227	Ontario	$8,287	28.8	15.0%
Moncton	211	New Brunswick	$8,037	31.0	10.9%
Calgary	191	Alberta	$7,746	30.4	11.5%
Banff	179	Alberta	$7,274	27.2	14.5%
London	174	Ontario	$7,804	29.4	13.8%
Peace River	113	Alberta	$7,802	32.5	12.4%
Whitehorse	110	Yukon	$6,772	33.3	10.9%
Kelowna	88	British Columbia	$8,598	33.4	10.2%
Charlottetown	66	Prince Edward Island	$7,704	27.6	16.7%

Province × tier — tier dominates, geography does not

Mean CLV by province and card tier

Reading across any row, CLV rises sharply from Star to Aurora. Reading down any column, it barely moves. This is the two-way version of the finding on the Value tab: tier is the value lever, geography is the acquisition lever, and they should be managed by different teams with different budgets.

Table view · Mean CLV by province × tier

Province	Star	Nova	Aurora
Alberta	6,446.00	7,764.00	10,237.00
British Columbia	6,642.00	8,358.00	10,430.00
Manitoba	6,891.00	7,518.00	11,318.00
New Brunswick	6,194.00	8,580.00	12,268.00
Newfoundland	7,756.00	6,882.00	10,561.00
Nova Scotia	6,611.00	8,170.00	10,710.00
Ontario	6,809.00	7,835.00	10,430.00
Prince Edward Island	6,564.00	8,950.00	9,058.00
Quebec	6,924.00	8,077.00	11,184.00
Saskatchewan	6,347.00	8,331.00	11,093.00
Yukon	6,089.00	6,875.00	8,905.00

Enrolment by year

The campaign is what makes 2018 the record year

Standard enrolments in 2018 (2,039) actually ran below 2017 (2,487) — because Feb–Apr sign-ups were all classified as campaign. The record total is entirely attributable to the promotion, which is exactly why the counterfactual on the Campaign tab matters: it separates the campaign's contribution from the program's underlying momentum.

Table view · Enrolment by year and type

Year	Standard	Campaign	Total
2,012	1,686	0	1,686
2,013	2,397	0	2,397
2,014	2,370	0	2,370
2,015	2,331	0	2,331
2,016	2,456	0	2,456
2,017	2,487	0	2,487
2,018	2,039	971	3,010

Method & data

How the numbers were made

Data lineage

Three source extracts → one analysis-ready star schema

Customer Loyalty History16,737 rows × 16 cols one row per member

Customer Flight Activity392,936 rows × 8 cols one row per member-month

Calendar2,557 rows × 4 cols 2012-01-01 → 2018-12-31

↓ clean · de-duplicate · derive

members16,737 × 58 + tenure, churn, RFM, activity roll-up

activity389,065 × 13 rectangular monthly panel

member_base · cohorts · monthly84 · 302 · 24 rows KPI aggregates

The join is a clean 1:N on Loyalty Number — no orphan facts, no members without a dimension row, no duplicate member keys. All three extracts carry a UTF-8 BOM on the first header, which silently corrupts the first column name if it is not stripped on read.

Cleaning log

Every decision taken on the raw data, and why

Step	Issue found	Decision	Rows
Flight activity	1,922 byte-identical rows and 3,871 rows breaking the (member, year, month) grain	Kept one record per member-month (highest distance). Verified the result is an exactly rectangular panel: 15,766 standard x 24 months + 971 promotion x 11 months = 389,065 rows	3,871
Flight activity	54.5% of member-months show 0 flights / 0 km / 0 points	Kept as genuine dormant months -- these are the denominator of the engagement KPIs, not missing data. Flagged with `flew` so activity rates can exclude them on demand	211,092
Flight activity	Promotion members carry only 11 months (Feb-Dec 2018); no 2017 history	Treated as structural, not missing. All year-on-year and pre/post comparisons of existing member behaviour exclude promotion members so the panel is balanced	10,681
Members	20 members with a negative salary (min -58,486)	Took the absolute value -- a leading-minus data-entry error, the magnitudes are otherwise in the normal range	20
Members	4,238 missing salaries (25.3%) -- these are exactly the 4,238 College-educated members, 0% missing in every other education level	Structurally missing (not random), so imputing would fabricate a College salary distribution. Left as NULL, excluded from salary cuts, and flagged with `salary_known` so no chart silently drops a quarter of the base	4,238
Members	88% NULL in Cancellation Year / Month	NULL means 'still a member', not missing. Converted to a boolean `churned` flag rather than dropped or imputed	14,670
Members	0 members with a cancellation date before their enrolment date	None found -- the enrolment/cancellation sequence is internally consistent	0
Members	Country is 'Canada' for all 16,737 rows; City maps 1:1 to Province	Dropped Country from the analysis (zero variance) and used Province/City as the geography hierarchy	16,737
Members	CLV correlates ~0 with distance flown (-0.004), flights (-0.006), tenure (0.002) and salary (-0.022); it is driven almost entirely by card tier	Treated CLV as a static member-value score attached to tier, NOT as observed spend. Volume KPIs are measured in flights and km; program cost in real redemption dollars	16,737

Assumptions & filters

Everything a reader needs in order to reproduce or challenge these numbers

Item	Rule applied	Why
Observation window	Flight activity 2017-01 → 2018-12 (24 months)	The only period the fact table covers. Enrolment history reaches back to 2012.
Grain	One row per member per month	3,871 rows broke this key. Tested summing vs de-duplicating: summing inflated affected members to 20.2 flights/year against 14.1 for an otherwise identical control group. De-duplicating restores an exactly rectangular panel (15,766 × 24 + 971 × 11 = 389,065 rows), so these are source-system re-writes.
Campaign window	2018-02 → 2018-04	Given by the data dictionary and independently confirmed by the earn-rate switch.
Campaign mechanic	A permanent 1.5× points earn rate	Not documented anywhere — derived from points ÷ kilometres. All 924 campaign members who flew earn 1.5 pts/km; zero standard members ever do.
Point value	1 point = $0.18 CAD	Derived from the data: Dollar Cost Points Redeemed ÷ Points Redeemed is constant to 4 decimals.
Balanced panel	Campaign members excluded from every year-on-year and pre/post comparison	They have no 2017 history. Leaving them in would manufacture a +16% YoY jump from cohort mix alone.
Counterfactual	2017 baseline × the +9.7% growth seen in the nine non-campaign months of 2018	Separates the campaign's contribution from the program's underlying momentum.
Salary	Kept as NULL for the 4,238 members without one; never used in a headline KPI	Missingness is exactly the College population and 0% everywhere else — imputation would invent a College income distribution and then let us 'discover' it.
Negative salaries	20 values replaced with their absolute value	Magnitudes are in the normal range; only the sign is wrong — a leading-minus entry error.
Cancellation date	NULL treated as 'still a member', not as missing	Becomes the churned flag that drives every retention KPI.
CLV	Treated as a static, tier-linked member-value score — never as booked revenue	CLV correlates ≈0 with distance (−0.00), flights (−0.01), tenure (0.00) and salary (−0.02), but is almost fully determined by card tier.
Country	Dropped from the analysis	All 16,737 members are in Canada — zero variance.

What this analysis cannot say

Stated up front rather than discovered in questions

No revenue in the data — The fact table carries no fare or ticket value, and CLV proves to be a tier-linked score rather than observed spend. Every value figure here is CLV under management, never booked revenue.
No control group — One airline, no untreated market. The +11.0 pp difference-in-differences on existing members is suggestive, not causal — seasonality is controlled by comparing like months a year apart, but a mid-2018 travel-demand shift cannot be ruled out.
A 24-month activity window — Cohort retention beyond month 24 is measured on progressively smaller and older cohorts; the early months are the only part measured on the full base.
Salary is 25% missing by construction — It is used only as a secondary cut and never in a headline KPI.
Synthetic data — Northern Lights Air is a fictitious carrier. The structure is realistic and internally consistent, but relationships that would exist in real data — CLV against flying, for one — are absent, and that shapes what can honestly be claimed.

Insights & actions

What to do next

The eight findings

Each one traceable to a chart on this dashboard

The campaign worked — and the gain was not borrowed.

971 sign-ups in Feb–Apr 2018 against a trend-adjusted counterfactual of 690: +281 incremental members, +40.7%. The classic failure mode of an acquisition promotion is pull-forward — harvesting next quarter's sign-ups early. That did not happen: the eight months after the campaign still ran +10.6% ahead of the same months in 2017.

Response was geographic, not demographic.

Québec indexed at 1.22 and British Columbia 1.09 against the national response rate, while Manitoba (0.52), Nova Scotia (0.63) and Alberta (0.69) lagged. Every demographic cut — gender, education, marital status, tier — moved less than 3 pp. The offer has broad appeal; only geography discriminates.

Existing members flew more too.

On a balanced panel that excludes campaign joiners, flights ran −0.3% in Jan–Apr and +10.7% in May–Dec: a +11.0 pp difference-in-differences, with the active member rate rising from 43.0% to 48.2%. Valuing the campaign at only its 281 new members materially undercounts it.

Campaign members are the best flyers, not discount-seekers.

They are 5.8% of the base but flew 15.7% of all 2018 flights, peaking at a 62% monthly active rate against ~50%. In July they averaged 8.7 flights per member versus 1.9.

Half of all churn happens on a single, predictable date.

1,048 of 2,067 cancellations (51%) fall on exactly month 8 of membership — 28× the neighbouring months, and it cost $2.67M of CLV in 2018 alone. This looks at first like a campaign defect, but it is not: it affects every cohort since 2013, and the campaign cohort's month-8 rate (10.4%) is lower than the 2017 cohort's (11.5%). An eight-month introductory term is expiring program-wide.

Churn is structural, not segmental.

Every demographic cut sits in an 11.6%–13.1% band around the 12.3% base rate, and churners carry higher CLV than stayers ($8,132 vs $7,969). There is no bad segment to stop recruiting — retention must target moments, not profiles.

The book is top-heavy and tier is the lever.

The top 20% of members hold 46% of value. Aurora's median CLV is 70% above Star, yet flights per member are identical across tiers — tier is not earned by flying, so upgrading it is a pricing decision the airline fully controls.

The program is funded by breakage — and that is also its risk.

Members redeem 1.54% of what they earn; 98.5% is never used. That is what makes the campaign cheap. But 1.5% is far below the 10–25% typical of airline programmes, which signals members do not see rewards as reachable — the condition that precedes disengagement.

Recommended actions

What the loyalty division should do next, and what it is worth

Action	Grounded in	Expected effect	Priority
Re-run the promotion, bought regionally Concentrate spend in Québec and British Columbia; cut Manitoba and Alberta weight.	Insights 1, 2	Same ~+40% lift at materially lower media cost	High
Kill the month-8 cliff — program-wide Taper the introductory benefit across months 8–12 instead of ending it, and fire a save offer in month 7 to every member approaching the date.	Insight 5	Recovering half the 325 month-8 leavers of 2018 ≈ +$1.3M CLV a year	High
Attack dormancy, not demographics Trigger a win-back after three consecutive dormant months, aimed at moments rather than segments.	Insights 3, 6	48.2% → 50%+ clears the target; ~6,000 incremental flights a year	High
Merchandise tier upgrades directly Tier drives CLV but is not earned by flying — so sell it rather than waiting for it to be earned.	Insight 7	Moving 5% of Star to Nova ≈ +$0.5M CLV under management	Medium
Publish the points liability monthly, with a stress test Report expected cost at 1.5%, 10% and 25% redemption alongside the nominal figure.	Insight 8	A move to 10% redemption would cost ~$12M; the exposure needs a number	Medium
Instrument the next campaign as a real experiment Hold out matched regions so the lift can be measured causally rather than inferred.	Insight 3's caveat	Turns a before/after read into a defensible causal one	Medium

Conclusion

The Feb–Apr 2018 promotion is the clearest success in the program's history: +281 incremental members at a CAC of $332 — roughly 4% of what a campaign member is worth — with no pull-forward, and a +11.0 pp lift in flying among members who were already enrolled. It should be repeated.

Two things should change when it is. Buy it regionally — Québec and British Columbia responded well above the national rate while demographics did not discriminate at all, so the media plan can be materially cheaper for the same lift. And instrument it as a real experiment, with matched hold-out regions, so the next read is causal rather than inferred.

Separately — and worth more than the campaign itself — fix the month-8 cliff. Half of every cancellation the program has ever recorded lands on exactly the eighth month of membership, across every cohort since 2013. That is an introductory term expiring, not customer behaviour, and in 2018 alone it cost $2.67M of CLV. Because the date is known in advance for every single member, it is also the most tractable retention problem the division has.

Underneath the campaign sit two structural facts the division should manage deliberately. Dormancy is the largest untapped lever — 48.2% of members fly in a given month, and these people are already acquired. And the program is funded by breakage: at a 1.54% redemption rate the points bank costs almost nothing today, but that number is far below industry norms, and it should be read as a signal about perceived reward reachability, not as a windfall.

Month	2017	2018	Change
Jan	188	192	4
Feb	196	295	99
Mar	237	330	93
Apr	196	346	150
May	231	244	13
Jun	200	272	72
Jul	217	231	14
Aug	208	218	10
Sep	210	185	-25
Oct	206	213	7
Nov	191	246	55
Dec	207	238	31

Month	2017	2018	Change
Jan	188	192	4
Feb	196	295	99
Mar	237	330	93
Apr	196	346	150
May	231	244	13
Jun	200	272	72
Jul	217	231	14
Aug	208	218	10
Sep	210	185	-25
Oct	206	213	7
Nov	191	246	55
Dec	207	238	31

Month	2017	2018	Change
Jan	188	192	4
Feb	196	295	99
Mar	237	330	93
Apr	196	346	150
May	231	244	13
Jun	200	272	72
Jul	217	231	14
Aug	208	218	10
Sep	210	185	-25
Oct	206	213	7
Nov	191	246	55
Dec	207	238	31