Bidirectional relationships and ambiguity

[@teunkruijer]

I can't even begin to describe how happy I am that i found this video. I was about to redesign my whole model to only use one fact table, and all I had to do to get the filters to sync was add a measure. Thank you so much! Great video.

[@karannchew2534]

Hi, Would you mind sharing more please? What type of measure did you create to get the filters to sync please?

[@atanasprodatanasov9233]

Brutal! I spent 4 sleepless nights because of that... thank you so much!

[@hasnainhaider8191]

man wow, how beautifully he explains and all the tools he uses. a true master.

[@harfshortshayari3340]

could you please help me in understanding why it worked after applying the visual filter?

[@27sreeni]

Really Really Superb explanation, I love your videos .. one request if any monthly enable SQLBI its really helpful to me

[@Jackassik]

My company created a bunch of data models where every possible relationship is bi-directional before hiring me... I told them that I won't touch them and I won't take responsibility for the numbers they display and your video will be a very good example why you need people who knows how to build data models as it's not simply drag and drop. Thank you very much for a very informative and well presented video :)

[@rick_.]

"Welcome to Hell!" I was laughing so much that I had to pause the video to put in this comment! I think the real key to understand why adding the filter to the slicer works is that a slicer is just another visual. All the usual cross-visual filtering rules apply. It's one of those things that is obvious if you think about it at all, but it is easy to fall into the trap of thinking a slicer is just a different way to show a filter. I did.

[@birasafabrice]

this has definetely earned you a new sub, you really nailed every part of the clarification needed throughout the tutorial. thanks a lot for this helpful tutorial. I have to learn from you a lot!

[@mau_lopez]

Great video, very clear and well done, thanks Mr. Ferrari. I would just emphasise that that the problem presented doesn't come only from using the bidirectional relationship, but it comes form using the bidirectional relationship IN a loop, because in a purely star schema no ambiguity is created when using bidirectional relationships. This would be just to reinforce what is already a great explanation!

[@SQLBI]

True - but you have to be sure you'll not add other fact tables to the same model sharing the same dimensions... in the future

[@kevinlabore1726]

in Summation WOW -- the really explains why you should ALMOST never use bi-directional relationships for purposes of filtering and why you probably should ALMOST NEVER trust a model that does .. in the back of my mind I was trying to figure out how the ambiguity impact things -- great walk through THANK YOU

[@JawadAli-nv9mr]

I always had issues with ambiguous models. and what I concluded from my report is to never use bi directional filter. I never knew what was the reason. now I got I must appreciate your efforts 👍👍

[@bactran7799]

While looking for a solution for my model that has a lot of bidirectional relationships, I meet Alberto with the amazing video and the answer for my big question. Thank you for your great works, I always appreciate your contribution to make DAX easier to use

[@vithaitang6492]

Thank you Mr. Ferrari, I love the way you raise the problem, analyze it and give the solution. Very clear and useful!

[@brianligat9493]

Just as I was looking at this, I found this video and it answered everything. Thanks Alberto. I think PBI should give a warning if it detects alternate fllter paths. If it would be difficult to explain such ambiguity to a non-technical person, then the design should be changed. The stakeholders/consumers will lose trust and confidence if something is happening that cannot be understood.

[@rawalkarnikita2947]

Thank you so much Sir for posting this valuable video. I can't express my gratitude towards you that how much I was needed this video. Thank you again 😊

[@manuelcometto13]

Sono rimasto stupefatto dalla chiarezza della spiegazione. Complimenti davvero

[@diederikvankolfschoten47]

This is a very neat trick to avoid ambiguity!! Thanks! SQLBI is on top of my top 5 resources!

[@dbszepesi]

Such an important video. I keep seeing bi-directional relationships in model and model and my concerns usually fall on deaf ears. Great to have a very detailed video to point people to now. Something not discussed here but I thought I remembering seeing in the video training that SQLBI does - the bidirectional relationships can also slow performance because they tend to keep the model from being able to choose the optimal path.

[@SQLBI]

yes bidirectional filters can also affect performance, but first of all you need to get right numbers!

[@sachin.tandon]

Very elegantly explained. Using the side Filter Pane would not have occured to me. I would have edited the Sales Measure by using CROSSFILTER. But I think the Filter Pane, does the job just as well!

[@alessandrocadenelli5849]

yes, but you have to do anyway the 'trick of filter pane' if you want to "synchronize" the two slicers...so two birds with one stone..

[@preethul1]

Excellent. So well explained. I would have never understood the consequences of bidirectional if I didn't bump into this video. Thanks a lot.

[@Bharath_PBI]

Amazing explanation. Self service BI report developers MUST watch this video. I had to explain this to lot of people and now I can just guide them here 👍 Thank you.

[@huseyinburaktasci1638]

I noticed the name of the final tab called "Welcome to hell" :D Clearly explains the issue! Thanks for the content, very useful!

[@lovebites4098]

I was facing the same issue and was stuck trying to figure out how bidirectional works from the last 2 days. I tried to check if it works like sql joins but still the result was ambiguous showing a greater value than that in sql. Thanks a lot for making this content. For slicer u used 'is not blank' filter. but i have direct column from 2 unrelated table(not connected directly, although connected through other tables, but flow non-unidirectional) and that too numerical(sum of amount). But now i hope i will figure it out how is it working in my case. thanks a lot lot sir. its much much informative and helpful.

[@dsfshshfhgsghs]

What a good teacher you are, Alberto! Fantastic explanation! Tante grazie!

[@CharudattaAmbekar]

Thanks! This quick solution of visual filters is possible only in Power Bi. In Excel, we use 'Crossfilter' in the measure formula.

[@Istion2991]

Amazing explanation. Fixed the problem I didn't even know how to frame a question around.

[@Broadmoore]

This is the best video I’ve ever seen on ambiguity and relationships. Thank you!

[@albinlabelle]

Thank you for this clear explanation (love the handwritten arrows on top of the modeling view !). Aside from the ambiguity arising from the use of a bidirectional relationship (which I manage to keep away from), I discovered a neat way to show only relevant choices in slicers !

[@ongkamsiong]

You provided much better insight about bidirectional filter than Microsoft documentation / training.

[@MichelMetran]

I discovered the channel today and... congratulations on such quality content! Learning a lot here!

[@nevermore17011990]

this video islike a movie of DAX !

[@ooooFISCHERoooo]

Cool. And yet so simple. Love the sheet name “Welcome to hell” 😂

[@williamrodriguez2694]

Literally busted out laughing when I saw that!

[@tuliofernandopinillosalegr3799]

Excelente exposición!!! Eres lo máximo Alberto Ferrari.

[@StephanOnisick]

Excellent Example of Crossfilter

[@user-ht4bl8fm1t]

Great Explanation Sir!! Keep it up👍

[@whdea3426]

So, the filter actually applies the crossfilter at the backend, this was not known to me , thanks a lot

[@TrungNguyen-jy2bc]

Amazing sir! The solution is so elegant that I have to clap when I saw it!

[@dhiraj3933]

This is a great! I need to try this next time I come across situation like this.

[@rafaelmiralha5995]

Thank you very much for these videos Alberto and SQLBI! Learning a lot about DAX!

[@Hurairahgiri]

Wow. I am taken aback by the explanation. Much love!

[@ThePasserby76]

Hello Alberto. Have just found your channel. Awesome video. Just one request - pleaseeee do more examples on ambiguity. Really helps us the beginners.

[@click-ue3kc]

Awesome! I am working on a model and I am happy I found this.

[@florin-danielpop3183]

Really easy to understand, as always great job. Thank you SQLBI team for your amazing work.

[@dariuszspiewak5624]

Hi Alberto. I think it would stand to reason to also mention that if someone uses USERELATIONSHIP and/or CROSSFILTER, they should always make sure that such a manipulation does not transforms the model into an ambiguous one.

[@nishanthvarmas8568]

Oh my goodness, such a beautiful explanation. Wow Man!

[@MahmoudSamir7oda]

Thank you Alberto i have been looking at that trick for weeks since i have similar business problem. Glad i came a croos the video. 🤗

[@bluesage1528]

Fantastic explanation and presentation! Thank you so much.

[@AnhTuan-fv2kn]

Thank you very much for these videos Alberto and SQLBI

[@fadwa2413]

I LOVE YOU Alberto !! YOU ARE THE LEGEND

[@RonaldPostelmans]

Nice explanation, i now finally understand ambiguity and now why i can not use bidrectional filters

[@user-qu3he1ed9e]

I couldn't understand why filtering could make the data sync. Can you give me a simple example? By the way, Great video, thanks a lot.

[@danielhowe4460]

Thank you for providing this detailed video on ambiguous models. There isn't a lot of information out there, so this is very much appreciated. Though, I only see a problem if you have a model with two fact tables. If you keep things simple (single fact table, multiple dimension tables), there should be no opportunity for multiple paths, correct? Therefore the model wouldn't be ambiguous?

[@joseaugustocarvalhofilho3506]

Excellent explanation! Congrats for such a great work!

[@Zenith40450]

Really loved the tab you named "Welcome to Hell" 😀😀😀😀

[@DanielTavares29]

Ciao, friends! Enjoy DAX! Thanks for sharing

[@goolamfareedbangie510]

This video is great but I'm not allowed to apply the measure on a visual if the visual is a card

[@Ronnet]

Would've liked to see the "welcome to hell" calculated metrics before and after removing the both directions relationship.

[@Robay146]

This is a classic lesson. Thank you! Big up!

[@raitup00]

Omg! Good problem, great explanation and excellent solution

[@MohamedHassan-fo3no]

Finally understood it right!!

[@Nura853]

Thank you for clear and precise explanation!

[@ExplorandoDatos]

Fantástico! Thank you MAESTRO!

[@tomreborns5739]

Thanks you to make it so clear and easy ! without avoiding to dig in detail :) Make times fresher to me haha

[@giacofigueroa]

Unfortunately this cannot be done when the user uses the filter pane to filter the report

[@SQLBI]

Yes, Microsoft didn't implement a UI for that. But if you enable bidirectional filters for that, you know the risk!

[@gborka]

Thank you a lot. Really happy to find this channel.

[@oliveroshea5765]

Alberto is the "data monster"!

[@rachitgoyal521]

Amazingly explained such a complex scenario.. Kudos!

[@berkayozkan5509]

Very clear. Thanks!!

[@kashifhaque2085]

This is very helpful, thanks!

[@barishazar5908]

Useful content, poetic explanation...Suberb teacher

Love this one, a true best pratice that can change life and model design

[@madroneroc]

Great, thanks! Would be interesting if you have a video on when is a good time to use bi-directional relationships?

[@SQLBI]

Beside limited cases (like implementing security patterns or physical implementation of M2M cardinality relationships - see sqlbits.com/Sessions/Event18/Different_types_of_many-to-many_relationships_in_Power_BI), it's almost never a good idea to enable bidirectional filters in the model, while it makes perfect sense in specific measure where you can control the side effects of it.

[@ovidlwwebsco]

awesome explanation, thank U

[@wasserslasken]

Thank you for a great video. I'm a newbie and am a bit confused as to how the direction of filtering goes when you apply a filter on the visual in the end of the video. It seems the cross filter direction does not affect that, since a filter on sales has an effect on product.

[@inerasmodellfornyttokalkyl1201]

Fantastic video!

[@ahamedyasar4722]

wonderful explanation... thank you...😀

[@tameraziz2104]

Thanks for sharing...

[@RoORou]

Great video, thanks!

[@RACHIMSAJA]

Very insightful! Thanks for the knowledge master🙏

[@evalang4435]

Absolutely perfect, thank you so much for this video

[@mabl4367]

I guess sales of zero amounts are uncommon but would it not be better so have a measure called "sales exists" and filter on that instead?

[@monicaosetkowski1542]

Mind blown! I need to try this. Thank you

[@vijaynrao2499]

Great! Really really great examples and explanation!

[@gpdoc4699]

Very well explained. Thanks Alberto.

[@Mrglasshalfempty]

Holy crap I never would have thought of that in a million years! Thanks!!

[@frankgovers8420]

Great, thx, great explanation... there are dax teachers who tell their students to put all the crossfilter in relations in both directions. Here is the reason why you never do this. Great alberto. ( who rides in a maserati! 😄)

[@SQLBI]

It hurts to know that someone promotes bad practices teaching that!

[@ajkrishna001]

Excellent very useful

[@wexwexexort]

Great explanation. You just earned a new sub sir.

[@gregg4636]

Presented Model is wrong or oversimplified. You should have relation field between Sales and Purchase. Then Product should be connected to Purchases only (can't Sell if not Purchased), and two separate copies of Calendar table one connected to Sales and second to Purchases for timeline of each Fact.

[@sukoi2113]

Very nicely explained. Thank you so much.

[@karannchew2534]

Notes for my future revision 6:05 A bi-directional relationship created two paths that link Date and Purchases. Yellow and Red. That's why the model is ambiguous. Yellow: direct link. Red: through Sales and Product. // A table can flow to another table by relationship via only one path. Multiple paths create ambiguity. // Here, Purchase table has relationships to both Date and Product. //These relationships affect filter going into Purchase, which is usually OK and is a typical star scheme. // However, with the bi-direction link into Product, Date now also reaches the Purchase table through another path (Date-Sales-Product). There are two separate paths from Date to Purchase. IF ANY TABLE REACH ANY ANOTHER TABLE THROUGH MORE THAN ONE PATH, THE DATA MODEL RISK BEING AMBIGUOUS 8:10 Sales is filtered by Customer and Dates. Which one being used? 8:38 Purchase can be calculated by including: 1. Customer and indirect Date filter 2. Direct Date filter 3. Customer and indirect Date filter (with direct date filter removed)

[@khaledtellopalacios3072]

Great video

[@olgaitskova8542]

This is amazing, thank you so much for this video!

[@bandaralqahtani4356]

thank you so much for this incridable video :)

[@phoenixsandy74]

Wow. This is so informative and useful.

[@harfshortshayari3340]

I do not understand how it worked, how it is filtering colour? Can anyone please help

[@future62]

Thanks a ton for this

[@rezag.7349]

Thank you for this useful video. Would it be fair to say that the ambiguity issue which you describe would not be an issue in a Star Schema w. one fact table (where Dimensional tables only reach to each other via the Fact table)?

[@SQLBI]

Yes, but it will appear the day you add another fact table, and at that point removing the bidirectional filter would break existing reports.

[@nesmeeclarensdomercant2569]

Continent is from Customer Table, it gives information about which continent the customers are from. Why would you want to filter purchase table by Continent in the Customer table? What is the relation between customer continent and purchase? Help me understand

[@SQLBI]

Indeed, that's the problem introduced by the bidirectional relationship - you don't want that to happen at all!

[@OneNI83]

Good video.Actually what happen in the behind when you add sales measure to the filters as not blank.Can you elaborate?

[@kg1610]

Thanks for explaining this, much appreciated.

[@kg1610]

Alberto, can you read this thread on the same topic. If you have any thoughts please comment. www.reddit.com/r/PowerBI/comments/kqsy0v/filter_direction/ Thanks for all the insight you have been sharing.

[@SQLBI]

You should avoid bidirectional filters in the model, use them in the measures only if possible. Best practices, workarounds, and explanations here: www.sqlbi.com/articles/syncing-slicers-in-power-bi/ www.sqlbi.com/articles/bidirectional-relationships-and-ambiguity-in-dax/