This should give you correct results, but won't be the fastest thing on the planet on your large dataset. 

Indexing could help quite a bit, and/or a computed column with an index on the computation.  In particular, a outletid+productid+weekdate composite index.

--use a second table to identify where the breaks > 7 days are

create

table #ptemp (WeekDate datetime NOT NULL, OutletId nvarchar(50) NOT NULL, ProductId nvarchar(50) NOT NULL)

insert

#ptemp (outletid,productID,weekdate)

select

p.outletid,p.productID,p.weekdate

from

#PromoHistory p left outer join #PromoHistory p2

on

p.outletid=p2.outletID and p.productID=p2.productid and p2.weekdate=dateadd(dd,-7,p.weekdate)

where

p2.weekdate is null

--add an index to help performance some

create

index ptmp on #ptemp(outletid,productid)

--get the data

select

ranktmp,min(WeekDate) startdate , max(dateadd(dd,7,weekdate)) endweek, OutletId, ProductId, Price, Cost,PromotionType,AdType,DisplayType,[X Value],[Y Value]

from

(select ph.*, count(pt.weekdate) ranktmp from #promohistory ph inner join #ptemp pt on ph.OutletId=pt.outletid and ph.ProductId=pt.productid

where

pt.weekdate<=ph.weekdate

group

by

ph.WeekDate, ph.OutletId, ph.ProductId, Price, Cost,PromotionType,AdType,DisplayType,[X Value],[Y Value]

)

ptmp

group

by OutletId, ProductId, Price, Cost,PromotionType,AdType,DisplayType,[X Value],[Y Value], ranktmp

order

by outletid, productid,ranktmp, min(WeekDate),max(weekdate)

--clean up and remove secondary table

drop

table #ptemp